1 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3 * config/i386/i386.md (rex64namesuffix): New mode attribute.
4 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
5 Merge insn pattern from sse_cvtsi2ss<round_name> and
6 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
7 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
8 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
9 using SWI48 mode iterator.
10 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
11 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
12 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
13 pattern from sse_cvttss2si<round_saeonly_name>
14 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
15 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
16 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
17 using SWI48 mode iterator.
18 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
19 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
20 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
21 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
22 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
23 using SWI48 mode iterator.
24 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
25 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
26 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
27 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
28 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
30 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
31 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
32 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
33 pattern from sse_cvttsd2si<round_saeonly_name>
34 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
36 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
38 * config/nds32/nds32-md-auxiliary.c
39 (nds32_valid_smw_lwm_base_p): Refine.
40 (nds32_output_smw_single_word): Refine.
41 (nds32_output_smw_double_word): New.
42 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
44 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
46 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
47 (nds32_output_stack_pop): Refine.
48 (nds32_expand_unaligned_load): Refine.
49 (nds32_expand_unaligned_store): Refine.
51 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
52 Chung-Ju Wu <jasonwucj@gmail.com>
54 * config/nds32/constants.md: Add TP_REGNUM constant.
55 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
56 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
58 * config/nds32/nds32-doubleword.md: Consider flag_pic.
59 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
60 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
61 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
62 and PIC code generation.
63 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
65 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
67 * config/nds32/nds32.md: Support TLS and PIC.
68 * config/nds32/nds32.c: Support TLS and PIC.
69 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
70 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
73 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
75 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
78 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
79 Chung-Ju Wu <jasonwucj@gmail.com>
81 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
82 * config/nds32/nds32-md-auxiliary.c
83 (symbolic_reference_mentioned_p): New.
84 (nds32_legitimize_ict_address): New.
85 (nds32_expand_ict_move): New.
86 (nds32_indirect_call_referenced_p): New.
87 (nds32_symbol_binds_local_p): Delete.
88 (nds32_long_call_p): Modify.
89 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
90 * config/nds32/nds32-protos.h
91 (symbolic_reference_mentioned_p): Declare.
92 (nds32_legitimize_ict_address): Declare.
93 (nds32_expand_ict_move): Declare.
94 (nds32_indirect_call_referenced_p): Declare.
95 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
96 (nds32_relax_group): Use nds32_ict_const_p as condition.
97 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
98 (nds32_asm_file_start): Output ict_model directive in asm code.
99 (nds32_legitimate_address_p): Consider indirect call.
100 (nds32_print_operand): Consider indirect call.
101 (nds32_print_operand_address): Consider indirect call.
102 (nds32_insert_attributes): Handle "indirect_call" attribute.
103 (TARGET_LEGITIMATE_ADDRESS_P): Define.
104 (TARGET_LEGITIMATE_CONSTANT_P): Define.
105 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
106 (TARGET_DELEGITIMIZE_ADDRESS): Define.
107 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
108 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
109 (TARGET_ICT_MODEL_SMALL): Define.
110 (TARGET_ICT_MODEL_LARGE): Define.
111 * config/nds32/nds32.md (movsi): Consider ict model.
112 (call, call_value): Consider ict model.
113 (sibcall, sibcall_value): Consider ict model.
114 * config/nds32/nds32.opt (mict-model): New option.
115 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
118 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
119 Monk Chiang <sh.chiang04@gmail.com>
120 Jim Wilson <jimw@sifive.com>
122 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
123 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
124 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
125 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
126 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
127 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
128 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
129 compute save_libcall_adjustment properly.
130 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
131 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
132 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
133 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
134 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
135 (ABI_SPEC): Handle mabi=ilp32e.
136 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
138 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
139 <-march>: Add rv32e as an example.
141 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
144 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
145 (intra_create_variable_infos): Handle C++ constructors.
147 2018-05-18 Martin Liska <mliska@suse.cz>
149 * passes.def: Remove a redundant pass.
151 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
154 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
156 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
158 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
160 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
162 (armv2, armv2a, armv3, armv3m): Delete architectures.
163 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
164 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
165 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
167 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
168 (*mulsidi3adddi): Likewise.
169 (mulsidi3): Likewise.
170 (*mulsidi3_nov6): Likewise.
171 (umulsidi3): Likewise.
172 (umulsidi3_nov6): Likewise.
173 (umaddsidi4): Likewise.
174 (*umulsidi3adddi): Likewise.
175 (smulsi3_highpart): Likewise.
176 (*smulsi3_highpart_nov6): Likewise.
177 (umulsi3_highpart): Likewise.
178 (*umulsi3_highpart_nov6): Likewise.
179 * config/arm/arm.h (arm_arch3m): Delete.
180 * config/arm/arm.c (arm_arch3m): Delete.
181 (arm_option_override_internal): Update armv3-related comment.
182 (arm_configure_build_target): Delete use of isa_bit_mode32.
183 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
184 (arm_rtx_costs_internal): Delete check of arm_arch3m.
185 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
188 * config/arm/arm-protos.h (arm_arch3m): Delete.
189 * config/arm/arm-tables.opt: Regenerate.
190 * config/arm/arm-tune.md: Likewise.
191 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
192 deleted architectures.
194 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
196 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
197 (armv5t, armv5te): New features.
198 (ARMv5, ARMv5e): Delete fgroups.
199 (ARMv5t, ARMv5te): Adjust for above changes.
201 (armv5, armv5e): Delete arches.
202 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
204 (*call_reg_arm): Likewise.
205 (*call_value_reg_armv5): Likewise.
206 (*call_value_reg_arm): Likewise.
207 (*call_symbol): Likewise.
208 (*call_value_symbol): Likewise.
209 (*sibcall_insn): Likewise.
210 (*sibcall_value_insn): Likewise.
212 (prefetch): Likewise.
213 (define_split and define_peephole2 dependent on arm_arch5):
215 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
217 (TARGET_ARM_QBIT): Likewise.
218 (TARGET_DSP_MULTIPLY): Likewise.
219 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
220 (arm_arch5, arm_arch5e): Delete.
221 (arm_arch5t, arm_arch5te): Declare.
222 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
223 (arm_arch5t): Declare.
224 (arm_option_reconfigure_globals): Update for the above.
225 (arm_options_perform_arch_sanity_checks): Update comment, replace
226 use of arm_arch5 with arm_arch5t.
227 (use_return_insn): Likewise.
228 (arm_emit_call_insn): Likewise.
229 (output_return_instruction): Likewise.
230 (arm_final_prescan_insn): Likewise.
231 (arm_coproc_builtin_available): Likewise.
232 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
233 arm_arch5e with arm_arch5t and arm_arch5te.
234 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
235 (arm_arch5t, arm_arch5te): Declare.
236 * config/arm/arm-tables.opt: Regenerate.
237 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
238 * config/arm/t-multilib: Likewise.
239 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
240 instead of arm_arch5.
241 (*call_reg_thumb1): Likewise.
242 (*call_value_reg_thumb1_v5): Likewise.
243 (*call_value_reg_thumb1): Likewise.
244 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
246 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
248 2018-05-18 Martin Liska <mliska@suse.cz>
250 PR gcov-profile/84846
251 * doc/gcov.texi: Document -t option of gcov tool.
253 2018-05-18 Martin Liska <mliska@suse.cz>
255 PR gcov-profile/84846
256 * gcov.c (print_usage): Add new -t option.
257 (process_args): Handle the option.
258 (generate_results): Use stdout as output when requested by
261 2018-05-18 Martin Liska <mliska@suse.cz>
263 PR gcov-profile/84846
264 * coverage.c (coverage_init): Write PWD to .gcno file.
265 * doc/gcov.texi: Document how working directory is printed.
266 * gcov-dump.c (dump_gcov_file): Print PWD.
267 * gcov.c (output_intermediate_file): Likewise.
268 (read_graph_file): Read PWD string.
269 (output_lines): Print PWD.
271 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
274 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
275 for retval and return false if all args to phi are zero.
277 2018-05-18 Richard Biener <rguenther@suse.de>
279 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
281 (evrp_dom_walker::before_dom_children): Call it.
283 2018-05-18 Richard Biener <rguenther@suse.de>
285 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
286 results when processing array refs with variable index.
288 2018-05-18 Toon Moene <toon@moene.org>
290 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
291 directly after that of -floop-interchange. Indicate that both
292 options are enabled by default when specifying -O3.
294 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
296 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
297 iterator. Delete separate integer-mode vec_set<mode> expander.
298 (aarch64_simd_vec_setv2di): Delete.
299 (vec_setv2di): Delete.
300 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
301 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
302 the "w, r" alternative.
304 2018-05-18 Martin Liska <mliska@suse.cz>
306 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
307 * tree-pass.h (make_pass_lower_switch_O0): New function.
308 * tree-switch-conversion.c (node_has_low_bound): Remove.
309 (node_has_high_bound): Likewise.
310 (node_is_bounded): Likewise.
311 (class pass_lower_switch): Make it a template type and create
313 (pass_lower_switch::execute): Add template argument.
314 (make_pass_lower_switch): New function.
315 (make_pass_lower_switch_O0): New function.
316 (do_jump_if_equal): Remove.
317 (emit_case_nodes): Simplify to just handle all 3 cases and leave
318 all the hard work to tree optimization passes.
320 2018-05-18 Martin Liska <mliska@suse.cz>
322 * dbgcnt.c (limit_low): Renamed from limit.
323 (limit_high): New variable.
324 (dbg_cnt_is_enabled): Check for upper limit.
325 (dbg_cnt): Adjust dumping.
326 (dbg_cnt_set_limit_by_index): Add new argument for high
328 (dbg_cnt_set_limit_by_name): Likewise.
329 (dbg_cnt_process_single_pair): Parse new format.
330 (dbg_cnt_process_opt): Use strtok.
331 (dbg_cnt_list_all_counters): Remove 'value' and add
333 * doc/invoke.texi: Document changes.
335 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
337 * doc/sourcebuild.texi (scalar_all_fma): Document.
338 * tree.def (FMA_EXPR): Delete.
339 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
340 * internal-fn.c (ternary_direct): New macro.
341 (expand_ternary_optab_fn): Likewise.
342 (direct_ternary_optab_supported_p): Likewise.
343 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
344 * builtins.c (fold_builtin_fma): Delete.
345 (fold_builtin_3): Don't call it.
346 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
347 * expr.c (expand_expr_real_2): Likewise.
348 * fold-const.c (operand_equal_p): Likewise.
349 (fold_ternary_loc): Likewise.
350 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
351 * gimple.c (DEFTREECODE): Likewise.
352 * gimplify.c (gimplify_expr): Likewise.
353 * optabs-tree.c (optab_for_tree_code): Likewise.
354 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
355 * tree-eh.c (operation_could_trap_p): Likewise.
356 (stmt_could_throw_1_p): Likewise.
357 * tree-inline.c (estimate_operator_cost): Likewise.
358 * tree-pretty-print.c (dump_generic_node): Likewise.
359 (op_code_prio): Likewise.
360 * tree-ssa-loop-im.c (stmt_cost): Likewise.
361 * tree-ssa-operands.c (get_expr_operands): Likewise.
362 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
363 * fold-const-call.h (fold_fma): Delete.
364 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
365 CFN_FNMA and CFN_FNMS.
367 * genmatch.c (combined_fn): New enum.
368 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
369 (commutative_op): New function.
370 (commutate): Use it. Handle more than 2 operands.
371 (dt_operand::gen_gimple_expr): Use commutative_op.
372 (parser::parse_expr): Allow :c to be used with non-binary
373 operators if the commutative operand is known.
374 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
375 CFN_FMS, CFN_FNMA and CFN_FNMS.
376 (backprop::process_assign_use): Remove FMA_EXPR handling.
377 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
378 (gen_hsa_fma): New function.
379 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
380 IFN_FNMA and IFN_FNMS.
381 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
382 * gimple-fold.h (follow_all_ssa_edges): Declare.
383 * gimple-fold.c (follow_all_ssa_edges): New function.
384 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
385 gimple_build interface and use follow_all_ssa_edges to fold the result.
386 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
387 instead of checking for optabs directly.
388 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
389 rather than FMA_EXPRs.
390 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
391 call to IFN_FMA instead of an FMA_EXPR.
393 2018-05-17 Jim Wilson <jimw@sifive.com>
395 * expr.c (do_tablejump): When converting index to Pmode, if we have a
396 sign extended promoted subreg, and the range does not have the sign bit
397 set, then do a sign extend.
399 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
400 test, check for sign extended subreg and/or constant operands, and
401 do a sign extend in that case.
403 2018-05-17 Steve Ellcey <sellcey@cavium.com>
405 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
406 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
408 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
409 Change logics_shift_reg to logics_shift_imm.
410 (thunderx2t99_fp_loadpair_basic): Delete.
411 (thunderx2t99_fp_storepair_basic): Delete.
412 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
413 (thunderx2t99_asimd_polynomial): Delete.
414 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
415 and neon_fp_mul_d_scalar_q.
416 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
417 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
418 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
419 (thunderx2t99_asimd_lut): Add missing tbl types.
420 (thunderx2t99_asimd_ext): Delete.
421 (thunderx2t99_asimd_load1_1_mult): Delete.
422 (thunderx2t99_asimd_load1_2_mult): Delete.
423 (thunderx2t99_asimd_load1_ldp): New.
424 (thunderx2t99_asimd_load1): New.
425 (thunderx2t99_asimd_load2): Add missing *load2* types.
426 (thunderx2t99_asimd_load3): New.
427 (thunderx2t99_asimd_load4): New.
428 (thunderx2t99_asimd_store1_1_mult): Delete.
429 (thunderx2t99_asimd_store1_2_mult): Delete.
430 (thunderx2t99_asimd_store2_mult): Delete.
431 (thunderx2t99_asimd_store2_onelane): Delete.
432 (thunderx2t99_asimd_store_stp): New.
433 (thunderx2t99_asimd_store1): New.
434 (thunderx2t99_asimd_store2): New.
435 (thunderx2t99_asimd_store3): New.
436 (thunderx2t99_asimd_store4): New.
438 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
440 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
441 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
443 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
444 Segher Boessenkool <segher@kernel.crashing.org>
447 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
450 2018-05-17 Richard Biener <rguenther@suse.de>
452 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
453 for pruning loop and prune defs feeding only already visited PHIs.
455 2018-05-17 Richard Biener <rguenther@suse.de>
457 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
459 2018-05-17 Bin Cheng <bin.cheng@arm.com>
460 Richard Biener <rguenther@suse.de>
462 PR tree-optimization/85793
463 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
464 for VMAT_ELEMENTWISE.
466 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
468 * internal-fn.h (lookup_internal_fn): Declare
469 * internal-fn.c (lookup_internal_fn): New function.
470 * gimple.c (gimple_build_call_from_tree): Handle calls to
472 * gimple-pretty-print.c (dump_gimple_call): Print "." before
473 internal function names.
474 * tree-pretty-print.c (dump_generic_node): Likewise.
475 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
477 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
479 * gimple-fold.h (gimple_build): Make the function forms take
480 combined_fn rather than built_in_function.
481 (gimple_simplify): Likewise.
482 * gimple-match-head.c (gimple_simplify): Likewise.
483 * gimple-fold.c (gimple_build): Likewise.
484 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
485 rather than gimple_build_call_internal.
486 (get_initial_defs_for_reduction): Likewise.
487 (vect_create_epilog_for_reduction): Likewise.
488 (vectorizable_live_operation): Likewise.
490 2018-05-17 Martin Liska <mliska@suse.cz>
492 * gimple-ssa-sprintf.c (format_directive): Do not use
493 space in between 'G_' and '('.
495 2018-05-17 Jakub Jelinek <jakub@redhat.com>
498 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
499 even if the mask is not all ones.
502 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
504 (ix86_gimple_fold_builtin): Likewise.
507 * config/i386/i386.c: Include tree-vector-builder.h.
508 (ix86_vector_shift_count): New function.
509 (ix86_fold_builtin): Fold shift builtins by scalar count.
510 (ix86_gimple_fold_builtin): Likewise.
512 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
513 _mm512_setzero): New intrinsics.
515 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
516 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
518 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
519 code generation for cases where splatting a value is not useful.
520 * simplify-rtx.c (simplify_ternary_operation): Simplify
521 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
522 mode to a vec_concat.
524 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
526 * config.gcc: Support "goldmont-plus".
527 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
529 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
530 PROCESSOR_GOLDMONT_PLUS.
531 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
532 (processor_target_table): Add "goldmont-plus".
533 (PTA_GOLDMONT_PLUS): Define.
534 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
535 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
536 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
537 (fold_builtin_cpu): Add "goldmont-plus".
538 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
539 (ix86_option_override_internal): Add "goldmont-plus".
540 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
541 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
542 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
543 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
545 2018-05-17 Richard Biener <rguenther@suse.de>
547 PR tree-optimization/85757
548 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
549 remove defs that only feed that PHI from further processing.
551 2018-05-16 Jim Wilson <jimw@sifive.com>
553 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
555 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
557 2018-05-16 Mark Wielaard <mark@klomp.org>
559 * dwarf2out.c (count_index_strings): New function.
560 (output_indirect_strings): Call count_index_strings and generate
561 header for dwarf_version >= 5.
563 2018-05-16 Mark Wielaard <mark@klomp.org>
565 * dwarf2out.c (dwarf_FORM): New function.
566 (set_indirect_string): Use dwarf_FORM.
567 (reset_indirect_string): Likewise.
568 (size_of_die): Likewise.
569 (value_format): Likewise.
570 (output_die): Likewise.
571 (add_skeleton_AT_string): Likewise.
572 (output_macinfo_op): Likewise.
573 (index_string): Likewise.
574 (output_index_string_offset): Likewise.
575 (output_index_string): Likewise.
576 (count_index_strings): Likewise.
578 2018-05-16 Carl Love <cel@us.ibm.com>
580 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
581 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
583 2018-05-16 Martin Jambor <mjambor@suse.cz>
585 * ipa-prop.c (ipa_free_all_edge_args): Remove.
586 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
588 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
590 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
591 (fnma<mode>4): Likewise.
592 (fms<mode>4): Likewise.
593 (fnms<mode>4): Likewise.
594 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
595 (aarch64_fnma<mode>4): Likewise.
596 (aarch64_fms<mode>4): Likewise.
597 (aarch64_fnms<mode>4): Likewise.
598 (aarch64_fnmadd<mode>4): Likewise.
600 2018-05-16 Jason Merrill <jason@redhat.com>
602 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
604 2018-05-16 Richard Biener <rguenther@suse.de>
606 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
607 (dump_stmt_cost): Declare.
608 (add_stmt_cost): Dump cost we add.
609 (add_stmt_costs): New function.
610 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
612 (vect_analyze_stmt): Adjust prototype.
613 (vectorizable_condition): Likewise.
614 (vectorizable_live_operation): Likewise.
615 (vectorizable_reduction): Likewise.
616 (vectorizable_induction): Likewise.
617 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
618 cost vector to pass to vectorizable_ and record afterwards.
619 (vect_model_reduction_cost): Take cost vector argument and adjust.
620 (vect_model_induction_cost): Likewise.
621 (vectorizable_reduction): Likewise.
622 (vectorizable_induction): Likewise.
623 (vectorizable_live_operation): Likewise.
624 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
625 SLP_TREE_NUMBER_OF_VEC_STMTS.
626 (vect_analyze_slp_cost_1): Remove.
627 (vect_analyze_slp_cost): Likewise.
628 (vect_slp_analyze_node_operations): Take visited args and
629 a target cost vector. Avoid processing already visited stmt sets.
630 (vect_slp_analyze_operations): Use a local cost vector to gather
631 costs and register those of non-discarded instances.
632 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
633 (vect_schedule_slp_instance): Remove copying of
634 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
636 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
637 adding cost. Record cost entry location.
638 (vect_prologue_cost_for_slp_op): Function to compute cost of
639 a constant or invariant generated for SLP vect in the prologue,
640 split out from vect_analyze_slp_cost_1.
641 (vect_model_simple_cost): Make static. Adjust for SLP costing.
642 (vect_model_promotion_demotion_cost): Likewise.
643 (vect_model_store_cost): Likewise, make static.
644 (vect_model_load_cost): Likewise.
645 (vectorizable_bswap): Add cost vector arg and adjust.
646 (vectorizable_call): Likewise.
647 (vectorizable_simd_clone_call): Likewise.
648 (vectorizable_conversion): Likewise.
649 (vectorizable_assignment): Likewise.
650 (vectorizable_shift): Likewise.
651 (vectorizable_operation): Likewise.
652 (vectorizable_store): Likewise.
653 (vectorizable_load): Likewise.
654 (vectorizable_condition): Likewise.
655 (vectorizable_comparison): Likewise.
656 (can_vectorize_live_stmts): Likewise.
657 (vect_analyze_stmt): Likewise.
658 (vect_transform_stmt): Adjust calls to vectorizable_*.
659 * tree-vectorizer.c: Include gimple-pretty-print.h.
660 (dump_stmt_cost): New function.
662 2018-05-16 Richard Biener <rguenther@suse.de>
664 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
665 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
666 * tree-ssa-dse.c: Include tree-ssa-loop.h.
667 (check_name): New callback.
668 (dse_classify_store): Track cycles via a visited bitmap of PHI
669 defs and simplify handling of in-loop and across loop dead stores
670 and properly fail for loop-variant refs. Handle byte-tracking with
671 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
674 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
676 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
677 (vect_get_mask_type_for_stmt): Likewise.
678 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
680 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
681 to determine the statement's vector type and the vector type that
682 should be used for calculating nunits. Deal with cases in which
683 the type has to be deferred.
684 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
685 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
686 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
687 (vect_determine_vf_for_stmt): New functions, split out from...
688 (vect_determine_vectorization_factor): ...here.
689 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
690 (vect_get_mask_type_for_stmt): New functions, split out from
691 vect_determine_vectorization_factor.
693 2018-05-16 Richard Biener <rguenther@suse.de>
695 * tree-cfg.c (verify_gimple_assign_ternary): Properly
696 verify the [VEC_]COND_EXPR embedded comparison.
698 2018-05-15 Martin Sebor <msebor@redhat.com>
700 PR tree-optimization/85753
701 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
702 RECORD_TYPE in addition to ARRAY_TYPE.
704 2018-05-15 Martin Sebor <msebor@redhat.com>
707 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
709 2018-05-15 Richard Biener <rguenther@suse.de>
711 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
712 add by_clobber_p one. Change algorithm to collect all defs
713 representing uses we need to walk and try reducing them to
714 a single one before failing.
715 (dse_dom_walker::dse_optimize_stmt): Adjust.
717 2018-05-13 Mark Wielaard <mark@klomp.org>
719 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
720 (size_of_loc_descr): Likewise.
721 (output_loc_operands): Likewise.
722 (output_loc_operands_raw): Likewise.
723 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
724 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
725 (hash_loc_operands): Likewise.
726 (compare_loc_operands): Likewise.
728 2018-05-14 Mark Wielaard <mark@klomp.org>
730 * dwarf2out.c (count_index_addrs): New function.
731 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
733 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
735 PR tree-optimization/83648
736 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
737 return value as malloc candidate.
739 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
742 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
743 as true in call to suggest_attribute.
745 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
747 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
750 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
752 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
754 (aarch64_crypto_aesd_fused): Likewise.
756 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
758 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
759 (movsi_aarch64): Likewise.
760 (load_pairsi): Likewise.
761 (load_pairdi): Likewise.
762 (store_pairsi): Likewise.
763 (store_pairdi): Likewise.
764 (load_pairsf): Likewise.
765 (load_pairdf): Likewise.
766 (store_pairsf): Likewise.
767 (store_pairdf): Likewise.
768 (zero_extend): Likewise.
769 (trunc): Swap alternatives.
770 (fcvt_target): Add '?' to prefer w over r.
772 2018-05-14 Jakub Jelinek <jakub@redhat.com>
775 * config/i386/i386.md: Disallow non-commutative arithmetics in
776 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
777 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
778 in the peephole2 before it.
780 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
782 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
783 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
784 (ix86_handle_option): Handle -mcldemote.
785 * config.gcc: New header.
786 * config/i386/cldemoteintrin.h: New file.
787 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
788 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
790 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
791 OPTION_MASK_ISA_CLDEMOTE.
792 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
793 (ix86_valid_target_attribute_inner_p): Ditto.
794 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
795 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
796 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
797 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
798 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
800 * config/i386/i386.opt: Add -mcldemote.
801 * config/i386/x86intrin.h: New header.
802 * doc/invoke.texi: Add -mcldemote.
804 2018-05-14 Richard Biener <rguenther@suse.de>
806 * doc/match-and-simplify.texi: Adjust :s documentation.
808 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
810 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
811 intended memcpy size.
812 (REORDER_45): Likewise.
814 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
817 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
818 * vec.c (qsort_chk): Use gcc_qsort.
819 * Makefile.in (OBJS-libcommon): Add sort.o.
820 (build/sort.o): New target. Use it...
821 (BUILD_RTL): ... here, and...
822 (build/gencfn-macros): ... here, and...
823 (build/genmatch): ... here.
825 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
826 Chung-Ju Wu <jasonwucj@gmail.com>
828 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
829 * config/nds32/nds32-graywolf.md: New file.
830 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
831 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
833 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
834 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
835 * config/nds32/nds32.md (pipeline_model): Add graywolf.
836 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
837 * config/nds32/pipelines.md: Include n15 settings.
839 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
840 Chung-Ju Wu <jasonwucj@gmail.com>
842 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
843 * config/nds32/nds32-n13.md: New file.
844 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
845 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
847 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
848 * config/nds32/nds32.md (pipeline_model): Add n13.
849 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
850 * config/nds32/pipelines.md: Include n13 settings.
852 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
853 Chung-Ju Wu <jasonwucj@gmail.com>
855 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
856 * config/nds32/nds32-n10.md: New file.
857 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
858 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
860 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
861 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
862 * config/nds32/nds32.md (pipeline_model): Add n10.
863 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
864 * config/nds32/pipelines.md: Include n10 settings.
866 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
867 Kito Cheng <kito.cheng@gmail.com>
868 Chung-Ju Wu <jasonwucj@gmail.com>
870 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
871 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
872 Add enum values for DSP extension instructions.
873 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
875 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
876 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
878 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
879 * config/nds32/nds32-dspext.md: New file for DSP implementation.
880 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
881 * config/nds32/nds32-intrinsic.md: Likewise.
882 * config/nds32/nds32_intrinsic.h: Likewise.
883 * config/nds32/nds32-md-auxiliary.c: Likewise.
884 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
885 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
886 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
887 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
888 * config/nds32/nds32-protos.h: New declarations for DSP extension.
889 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
890 TYPE_DMAC in switch statement.
891 * config/nds32/nds32.c: New checking and implementation for DSP
892 extension instructions.
893 * config/nds32/nds32.h: Likewise.
894 * config/nds32/nds32.md: Likewise.
895 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
896 * config/nds32/predicates.md: Implement new predicates for DSP
899 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
901 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
902 Reformat alternatives and attributes so it is easier to identify
903 which constraints/attributes go with which instruction.
904 (mov<mode>_hardfloat32, FMOVE64): Likewise.
905 (mov<mode>_softfloat32, FMOVE64): Likewise.
906 (mov<mode>_hardfloat64, FMOVE64): Likewise.
907 (mov<mode>_softfloat64, FMOVE64): Likewise.
909 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
911 * doc/extend.texi (PowerPC Built-in Functions): Rename this
913 (Basic PowerPC Built-in Functions): The new name of the
914 subsection previously known as "PowerPC Built-in Functions".
915 (Basic PowerPC Built-in Functions Available on all Configurations):
917 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
918 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
919 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
920 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
922 2018-05-11 Martin Jambor <mjambor@suse.cz>
925 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
928 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
931 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
933 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
935 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
936 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
937 (ix86_handle_option): Handle -mwaitpkg.
938 * config.gcc: New header.
939 * config/i386/cpuid.h (bit_WAITPKG): New bit.
940 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
941 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
943 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
944 OPTION_MASK_ISA_WAITPKG.
945 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
946 (ix86_option_override_internal): Add PTA_WAITPKG.
947 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
948 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
950 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
951 __builtin_ia32_umwait and __builtin_ia32_tpause.
952 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
953 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
954 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
955 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
956 UNSPECV_TPAUSE): New.
957 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
958 * config/i386/i386.opt: Add -mwaitpkg.
959 * config/i386/waitpkgintrin.h: New file.
960 * config/i386/x86intrin.h: New header.
961 * doc/invoke.texi: Add -mwaitpkg.
963 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
966 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
968 (cortex-m0): Use armv6s-m isa.
969 (cortex-m0plus): Likewise.
970 (cortex-m1): Likewise.
971 (cortex-m0.small-multiply): Likewise.
972 (cortex-m0plus.small-multiply): Likewise.
973 (cortex-m1.small-multiply): Likewise.
975 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
976 Jakub Jelinek <jakub@redhat.com>
978 PR tree-optimization/85692
979 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
980 source permute as well.
982 2018-05-11 Martin Liska <mliska@suse.cz>
985 * doc/extend.texi: Document LLVM style format for no_sanitize
988 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
990 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
991 mode_supports_vsx_dform_quad to mode_supports_dq_form.
992 (mode_supports_vsx_dform_quad): Likewise.
993 (mode_supports_vmx_dform): Move these functions to be next to the
994 other mode_supports functions.
995 (mode_supports_dq_form): Likewise.
996 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
997 mode_supports_dq_form.
998 (reg_offset_addressing_ok_p): Likewise.
999 (offsettable_ok_by_alignment): Likewise.
1000 (rs6000_legitimate_offset_address_p): Likewise.
1001 (legitimate_lo_sum_address_p): Likewise.
1002 (rs6000_legitimize_address): Likewise.
1003 (rs6000_legitimize_reload_address): Likewise.
1004 (rs6000_secondary_reload_inner): Likewise.
1005 (rs6000_preferred_reload_class): Likewise.
1006 (rs6000_output_move_128bit): Likewise.
1008 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1010 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1011 Generate SImode target register for null target.
1012 <case IX86_BUILTIN_XGETBV>: Ditto.
1013 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1014 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1016 2018-05-10 Carl Love <cel@us.ibm.com>
1018 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1019 dcbtt and dcbtstt if operands[2] is 0.
1021 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1024 * config/i386/sse.md (usadv64qi): New expander.
1026 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1028 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1029 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1030 -maltivec=be support.
1031 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1032 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1033 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1034 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1035 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1036 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1037 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1038 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1039 altivec_vsumsws): Adjust.
1040 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1041 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1042 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1044 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1045 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1046 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1047 (altivec_lve<VI_char>x): Delete expand.
1048 (*altivec_lve<VI_char>x_internal): Rename to...
1049 (altivec_lve<VI_char>x): ... this.
1050 (altivec_lvxl_<mode>): Delete expand.
1051 (*altivec_lvxl_<mode>_internal): Rename to ...
1052 (altivec_lvxl_<mode>): ... this.
1053 (altivec_stvxl_<mode>): Delete expand.
1054 (*altivec_stvxl_<mode>_internal): Rename to ...
1055 (altivec_stvxl_<mode>): ... this.
1056 (altivec_stve<VI_char>x): Delete expand.
1057 (*altivec_stve<VI_char>x_internal): Rename to ...
1058 (altivec_stve<VI_char>x): ... this.
1059 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1060 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1061 reduc_plus_scal_<mode>): Adjust.
1062 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1064 (rs6000_cpu_cpp_builtins): Adjust.
1065 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1066 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1067 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1068 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1069 -maltivec=be support.
1070 (rs6000_split_vec_extract_var): Adjust.
1071 (rs6000_split_v4si_init): Adjust.
1072 (swap_selector_for_mode): Delete.
1073 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1074 altivec_expand_stvex_be): Delete.
1075 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1076 -maltivec=be support.
1077 (rs6000_gimple_fold_builtin): Ditto.
1078 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1080 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1081 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1082 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1083 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1084 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1085 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1086 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1087 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1088 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1089 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1090 anonymous split): Adjust.
1091 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1092 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1094 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1096 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1097 when --with-gxx-include-dir is also specified.
1098 * configure: Regenerate.
1100 2018-05-09 Jim Wilson <jimw@sifive.com>
1103 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1104 * config/riscv/t-withmultilib: New.
1105 * config/riscv/withmultilib.h: New.
1106 * doc/install.texi: Document RISC-V --with-multilib-list support.
1108 2018-05-09 Richard Biener <rguenther@suse.de>
1110 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1112 (vect_bb_vectorization_profitable_p): Adjust. Compute
1113 actual scalar cost using the cost vector and the add_stmt_cost
1116 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1118 PR rtl-optimization/85645
1119 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1120 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1122 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1124 PR rtl-optimization/85645
1125 * shrink-wrap.c (spread_components): Return a boolean saying if
1126 anything was changed.
1127 (try_shrink_wrapping_separate): Iterate spread_components until
1128 nothing changes anymore.
1130 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1132 PR rtl-optimization/85645
1133 * regrename.c (build_def_use): Also kill the chains that include the
1134 destination of a REG_CFA_REGISTER note.
1136 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1138 PR rtl-optimization/85645
1139 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1140 insn that has a REG_CFA_REGISTER note.
1142 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1144 * cfgexpand.c (expand_clobber): New function.
1145 (expand_gimple_stmt_1): Use it.
1146 * tree-vect-stmts.c (vect_clobber_variable): New function,
1148 (vectorizable_simd_clone_call): ...here.
1149 (vectorizable_store): Emit a clobber either side of an
1150 IFN_STORE_LANES sequence.
1151 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1153 2018-05-09 Tom de Vries <tom@codesourcery.com>
1156 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1157 (define_insn "trap_if_false"): Add exit after trap.
1159 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1161 PR rtl-optimization/85638
1162 * bb-reorder.c: Include common/common-target.h.
1163 (create_forwarder_block): New function extracted from...
1164 (fix_up_crossing_landing_pad): ...here. Rename into...
1165 (dw2_fix_up_crossing_landing_pad): ...this.
1166 (sjlj_fix_up_crossing_landing_pad): New function.
1167 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1168 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1169 from both partitions and exit the loop after one iteration.
1171 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1174 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1176 (Basic PowerPC Built-in Functions): The new name of the
1177 subsection previously known as "PowerPC Built-in Functions".
1178 (Basic PowerPC Built-in Functions Available on all Configurations):
1180 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1182 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1184 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1186 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1189 2018-05-08 Jim Wilson <jimw@sifive.com>
1191 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1192 (LD_EMUL_SUFFIX): New.
1193 (LINK_SPEC): Use it.
1195 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1197 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1199 (Basic PowerPC Built-in Functions): The new name of the
1200 subsection previously known as "PowerPC Built-in Functions".
1201 (Basic PowerPC Built-in Functions Available on all Configurations):
1203 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1205 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1207 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1209 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1212 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1215 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1216 after cmpelim optimization.
1218 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1220 * config.gcc: Support "goldmont".
1221 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1222 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1224 * config/i386/i386.c (m_GOLDMONT): Define.
1225 (processor_target_table): Add "goldmont".
1226 (PTA_GOLDMONT): Define.
1227 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1228 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1229 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1230 (fold_builtin_cpu): Add "goldmont".
1231 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1232 (ix86_option_override_internal): Add "goldmont".
1233 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1234 (processor_type): Add PROCESSOR_GOLDMONT.
1235 * config/i386/i386.md: Add CPU "glm".
1236 * config/i386/glm.md: New file.
1237 * config/i386/x86-tune.def: Add m_GOLDMONT.
1238 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1240 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1243 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1245 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1246 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1247 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1250 * config/i386/i386.c (ix86_fold_builtin): Handle
1251 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1254 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1255 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1257 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1260 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1261 (check_arch): Likewise.
1262 (check_fpu): Return the result rather than printing it.
1263 (end arch): Fix operator precedence.
1264 (end cpu): Likewise.
1265 (END): Print the result from check_fpu.
1267 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1268 Alan Hayward <alan.hayward@arm.com>
1269 David Sherwood <david.sherwood@arm.com>
1271 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1272 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1273 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1274 (*fcmuo<mode>_and): New patterns.
1276 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1278 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1279 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1280 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1281 (cmp_op, sve_imm_con): New code attributes.
1282 (SVE_COND_INT_CMP, imm_con): Delete.
1283 (cmp_op): Remove above unspecs from int attribute.
1284 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1286 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1287 comparison-specific unspecs.
1288 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1289 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1290 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1291 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1292 (*vec_fcm<cmp_op><mode>): Rename to...
1293 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1294 (*vec_fcmuo<mode>): Rename to...
1295 (*fcmuo<mode>): ...this and adjust likewise.
1296 (*pred_fcm<cmp_op><mode>): New pattern.
1297 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1298 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1300 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1302 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1303 (aarch64_emit_sve_predicated_cond): New function.
1304 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1305 (aarch64_emit_unspec_cond_or): Replace with...
1306 (aarch64_emit_sve_or_conds): ...this new function. Use
1307 aarch64_emit_sve_ptrue_op for the individual comparisons and
1308 aarch64_emit_binop to OR them together.
1309 (aarch64_emit_inverted_unspec_cond): Replace with...
1310 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1311 aarch64_emit_sve_ptrue_op for the comparison and
1312 aarch64_emit_unop to invert the result.
1313 (aarch64_expand_sve_vec_cmp_float): Update after the above
1314 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1316 2018-05-07 Nathan Sidwell <nathan@acm.org>
1318 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1319 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1320 (Backwards Compatibility): Likewise.
1322 2018-05-07 Luis Machado <luis.machado@linaro.org>
1326 2018-05-07 Luis Machado <luis.machado@linaro.org>
1328 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1329 <prefetch_dynamic_strides>: New const bool field.
1330 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1331 prefetch_dynamic_strides.
1332 (exynosm1_prefetch_tune): Likewise.
1333 (thunderxt88_prefetch_tune): Likewise.
1334 (thunderx_prefetch_tune): Likewise.
1335 (thunderx2t99_prefetch_tune): Likewise.
1336 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1337 (aarch64_override_options_internal): Update to set
1338 PARAM_PREFETCH_DYNAMIC_STRIDES.
1339 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1340 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1341 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1342 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1343 prefetch-dynamic-strides setting.
1345 2018-05-07 Luis Machado <luis.machado@linaro.org>
1347 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1348 <minimum_stride>: New const int field.
1349 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1350 minimum_stride field.
1351 (exynosm1_prefetch_tune): Likewise.
1352 (thunderxt88_prefetch_tune): Likewise.
1353 (thunderx_prefetch_tune): Likewise.
1354 (thunderx2t99_prefetch_tune): Likewise.
1355 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1356 (aarch64_override_options_internal): Update to set
1357 PARAM_PREFETCH_MINIMUM_STRIDE.
1358 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1359 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1360 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1361 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1362 stride is constant and is below the minimum stride threshold.
1364 2018-05-07 Luis Machado <luis.machado@linaro.org>
1366 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1369 2018-05-07 Luis Machado <luis.machado@linaro.org>
1371 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1372 <prefetch_dynamic_strides>: New const bool field.
1373 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1374 prefetch_dynamic_strides.
1375 (exynosm1_prefetch_tune): Likewise.
1376 (thunderxt88_prefetch_tune): Likewise.
1377 (thunderx_prefetch_tune): Likewise.
1378 (thunderx2t99_prefetch_tune): Likewise.
1379 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1380 (aarch64_override_options_internal): Update to set
1381 PARAM_PREFETCH_DYNAMIC_STRIDES.
1382 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1383 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1384 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1385 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1386 prefetch-dynamic-strides setting.
1388 2018-05-07 Luis Machado <luis.machado@linaro.org>
1390 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1391 <minimum_stride>: New const int field.
1392 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1393 minimum_stride field.
1394 (exynosm1_prefetch_tune): Likewise.
1395 (thunderxt88_prefetch_tune): Likewise.
1396 (thunderx_prefetch_tune): Likewise.
1397 (thunderx2t99_prefetch_tune): Likewise.
1398 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1399 (aarch64_override_options_internal): Update to set
1400 PARAM_PREFETCH_MINIMUM_STRIDE.
1401 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1402 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1403 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1404 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1405 stride is constant and is below the minimum stride threshold.
1407 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1410 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1411 the type is addressable. Don't force op into register if it has
1414 2018-05-05 Roland McGrath <mcgrathr@google.com>
1417 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1418 any section for which we don't know a specific type it should have,
1419 regardless of name. Previously this was done only for the exact
1420 names ".init_array", ".fini_array", and ".preinit_array".
1421 (default_elf_asm_named_section): Add comment about
1422 relationship with default_section_type_flags and SECTION_NOTYPE.
1423 (get_section): Don't consider it a type conflict if one side has
1424 SECTION_NOTYPE and the other doesn't, as long as neither has the
1425 SECTION_BSS et al used in the default_section_type_flags logic.
1427 2018-05-05 Tom de Vries <tom@codesourcery.com>
1430 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1431 (workaround_barsyncs): New function.
1432 (nvptx_reorg): Use workaround_barsyncs.
1433 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1434 (define_expand "nvptx_membar_cta"): New define_expand.
1435 (define_insn "*nvptx_membar_cta"): New insn.
1437 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1439 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1440 To improve optimization opportunities.
1441 * builtin-types.def: The new needed builtin types for the above.
1443 2018-05-04 Richard Biener <rguenther@suse.de>
1445 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1446 * gimple-ssa-store-merging.c
1447 (imm_store_chain_info::output_merged_store): Remove redundant create,
1448 release split_store vector contents on failure.
1449 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1450 scalar stmt vector on cache hit.
1452 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1454 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1456 * config.gcc (powerpc-xilinx-eabi*): Remove.
1457 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1459 (fusion_addis_mem_combo_load): Ditto.
1460 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1462 (rs6000_cpu_cpp_builtins): Ditto.
1463 * config/rs6000/rs6000-linux.c
1464 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1465 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1466 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1468 (rs6000_setup_reg_addr_masks): Ditto.
1469 (rs6000_init_hard_regno_mode_ok): Ditto.
1470 (rs6000_option_override_internal): Ditto.
1471 (legitimate_lo_sum_address_p): Ditto.
1472 (rs6000_legitimize_address): Ditto.
1473 (rs6000_legitimize_reload_address): Ditto.
1474 (rs6000_legitimate_address_p): Ditto.
1475 (abi_v4_pass_in_fpr): Ditto.
1476 (setup_incoming_varargs): Ditto.
1477 (rs6000_gimplify_va_arg): Ditto.
1478 (rs6000_split_multireg_move): Ditto.
1479 (rs6000_savres_strategy): Ditto.
1480 (rs6000_emit_prologue_components): Ditto.
1481 (rs6000_emit_epilogue_components): Ditto.
1482 (rs6000_emit_prologue): Ditto.
1483 (rs6000_emit_epilogue): Ditto.
1484 (rs6000_elf_file_end): Ditto.
1485 (rs6000_function_value): Ditto.
1486 (rs6000_libcall_value): Ditto.
1487 * config/rs6000/rs6000.h: Ditto.
1488 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1489 (TARGET_MINMAX): ... this. New.
1490 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1491 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1492 (*movsi_internal1_single): Delete.
1493 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1494 mfpu=, mxilinx-fpu): Delete.
1495 * config/rs6000/singlefp.h: Delete.
1496 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1497 * config/rs6000/t-rs6000: Ditto.
1498 * config/rs6000/t-xilinx: Delete.
1499 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1500 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1501 (VStype_simple): Delete.
1502 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1503 * config/rs6000/xfpu.h: Delete.
1504 * config/rs6000/xfpu.md: Delete.
1505 * config/rs6000/xilinx.h: Delete.
1506 * config/rs6000/xilinx.opt: Delete.
1507 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1508 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1510 2018-05-04 Tom de Vries <tom@codesourcery.com>
1513 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1516 2018-05-04 Richard Biener <rguenther@suse.de>
1519 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1520 (expand_complex_div_wide): Likewise.
1521 (expand_complex_operations_1): Likewise.
1522 (expand_complex_libcall): Preserve EH info of the original stmt.
1523 (tree_lower_complex): Handle removed blocks.
1524 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1525 on complex multiplication and division libcall builtins.
1527 2018-05-04 Richard Biener <rguenther@suse.de>
1530 * fold-const.c (negate_expr_p): Restrict negation of operand
1531 zero of a division to when we know that can happen without
1533 (fold_negate_expr_1): Likewise.
1535 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1538 * real.h (real_nextafter): Declare.
1539 * real.c (real_nextafter): New function.
1540 * fold-const-call.c (fold_const_nextafter): New function.
1541 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1542 CASE_CFN_NEXTTOWARD.
1543 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1544 even when arg1_mode is different from arg0_mode.
1546 2018-05-03 Nathan Sidwell <nathan@acm.org>
1548 * doc/extend.texi (Deprecated Features): Remove
1550 (Backwards Compatibility): Likewise.
1551 * doc/invoke.texi (C++ Language Options): Likewise.
1552 (C++ Dialect Options): Likewise.
1554 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1557 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1558 _mm512_mask_mullox_epi64): New intrinsics.
1560 2018-05-03 Tom de Vries <tom@codesourcery.com>
1563 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1564 dump files): Add offload-tree.
1566 2018-05-03 Richard Biener <rguenther@suse.de>
1568 PR tree-optimization/85615
1569 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1570 to loops not nested in BBs loop father to avoid creating multi-entry
1573 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1575 PR tree-optimization/70291
1576 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1577 arguments. Change return type to tree. Emit libcall as a new
1578 statement rather than replacing existing one when inplace_p is true.
1579 (expand_complex_multiplication_components): New function.
1580 (expand_complex_multiplication): Expand floating-point complex
1581 multiplication using the above.
1582 (expand_complex_division): Rename inner_type parameter to type.
1583 Update expand_complex_libcall call-site.
1584 (expand_complex_operations_1): Update expand_complex_multiplication
1585 and expand_complex_division call-sites.
1587 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1590 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1591 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1592 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1593 the highest significant bit of the shift count mask is clear. In
1594 check whether and[sq]i3 is needed verify that all significant bits
1595 of the shift count other than the highest are set.
1597 2018-05-02 Tom de Vries <tom@codesourcery.com>
1600 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1601 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1602 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1603 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1604 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1605 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1606 __builtin_goacc_parlevel_size.
1608 2018-05-02 Richard Biener <rguenther@suse.de>
1610 PR tree-optimization/85597
1611 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1612 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1615 2018-05-02 Tom de Vries <tom@codesourcery.com>
1618 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1619 dump files): Add ltrans-tree.
1621 2018-05-02 Tom de Vries <tom@codesourcery.com>
1624 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1625 dump files): Add wpa-ipa.
1627 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1629 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1630 powerpc*-*-linux*paired* target.
1631 * config/rs6000/750cl.h: Delete.
1632 * config/rs6000/paired.h: Delete.
1633 * config/rs6000/paired.md: Delete.
1634 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1636 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1637 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1638 comment. Remove paired float support.
1639 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1640 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1642 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1643 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1645 * config/rs6000/rs6000.c: Remove paired float support.
1646 (paired_expand_vector_init, paired_expand_vector_move,
1647 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1648 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1649 paired_expand_builtin, paired_expand_predicate_builtin,
1650 paired_init_builtins): Delete.
1651 * config/rs6000/rs6000.h: Remove paired float support.
1652 * config/rs6000/rs6000.md: Remove paired float support.
1653 (move_from_CR_ov_bit): Delete.
1654 * config/rs6000/rs6000.opt (mpaired): Delete.
1655 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1656 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1658 2018-05-02 Richard Biener <rguenther@suse.de>
1661 * gimplify.c (gimplify_save_expr): When in SSA form allow
1662 SAVE_EXPRs to compute to SSA vars.
1664 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1667 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1668 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1669 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1670 clobber operands[2], instead use a new pseudo. Formatting fixes.
1672 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1674 PR tree-optimization/85586
1675 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1676 exit early for statements in the same group if the accesses are
1679 2018-05-02 Tom de Vries <tom@codesourcery.com>
1682 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1685 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1687 PR tree-optimization/85143
1688 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1690 2018-05-01 Tom de Vries <tom@codesourcery.com>
1693 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1694 not found" error message.
1696 2018-05-01 Tom de Vries <tom@codesourcery.com>
1699 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1700 * vec.c (test_ordered_remove_if): New function.
1701 (vec_c_tests): Call test_ordered_remove_if.
1702 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1703 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1704 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1705 VEC_ORDERED_REMOVE_IF.
1707 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1709 PR tree-optimization/82665
1710 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1711 pointer subtraction where arguments come from a memchr call.
1713 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1715 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1716 --push-state --as-needed and --pop-state instead of --as-needed and
1717 --no-as-needed if ld supports it.
1718 * configure: Regenerated.
1721 * doc/install.texi2html: Replace _002d with - and _002a with * in
1722 generated html files using sed.
1724 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1727 * gcc-rich-location.c (blank_line_before_p): New function.
1728 (use_new_line): New function.
1729 (gcc_rich_location::add_fixit_insert_formatted): New function.
1730 * gcc-rich-location.h
1731 (gcc_rich_location::add_fixit_insert_formatted): New function.
1733 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1735 * selftest.c (assert_streq): Rename "expected" and "actual" to
1736 "val1" and "val2". Extend NULL-handling to cover both inputs
1737 symmetrically, while still requiring both to be non-NULL for a pass.
1738 * selftest.h (assert_streq): Rename "expected" and "actual" to
1740 (ASSERT_EQ): Likewise.
1741 (ASSERT_EQ_AT): Likewise.
1742 (ASSERT_KNOWN_EQ): Likewise.
1743 (ASSERT_KNOWN_EQ_AT): Likewise.
1744 (ASSERT_NE): Likewise.
1745 (ASSERT_MAYBE_NE): Likewise.
1746 (ASSERT_MAYBE_NE_AT): Likewise.
1747 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1748 the assertion to pass.
1749 (ASSERT_STREQ_AT): Likewise.
1751 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1753 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1754 interaction with -pie.
1756 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1758 * selftest.h: Fix alphabetization of per-source-file selftest
1761 2018-04-30 Jason Merrill <jason@redhat.com>
1763 PR c++/61982 - dead stores to destroyed objects.
1764 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1767 2018-04-30 Jason Merrill <jason@redhat.com>
1769 * tree.c (build_clobber): New.
1770 * tree.h: Declare it.
1771 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1773 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1775 * diagnostic-show-locus.c (layout::layout): Update for
1776 location_get_source_line returning a char_span.
1777 (struct char_span): Move to input.h.
1778 (struct correction): Update for fields in char_span becoming
1780 (struct source_line): Update for location_get_source_line
1781 returning a char_span.
1782 (layout::print_line): Likewise.
1783 * edit-context.c (edited_file::print_content): Likewise.
1784 (edited_file::print_diff_hunk): Likewise.
1785 (edited_file::print_run_of_changed_lines): Likewise.
1786 (edited_file::get_num_lines): Likewise.
1787 (edited_line::edited_line): Likewise.
1788 * final.c (asm_show_source): Likewise.
1789 * input.c (location_get_source_line): Convert return type
1790 from const char * to char_span, losing the final "line_len"
1792 (dump_location_info): Update for the above.
1793 (get_substring_ranges_for_loc): Likewise. Use a char_span
1794 when handling the literal within the line.
1795 (test_reading_source_line): Update for location_get_source_line
1796 returning a char_span.
1797 * input.h (class char_span): Move here from
1798 diagnostic-show-locus.c, converting from a struct to a class.
1799 Make data members private.
1800 (char_span::operator bool): New.
1801 (char_span::length): New.
1802 (char_span::get_buffer): New.
1803 (char_span::operator[]): New.
1804 (char_span::subspan): Make const.
1805 (char_span::xstrdup): New.
1806 (location_get_source_line): Convert return type from const char *
1807 to char_span, losing the final "line_size" param.
1809 2018-04-30 Jan Hubicka <jh@suse.cz>
1811 * lto-wrapper.c (ltrans_priorities): New static var.
1812 (cmp_priority): New.
1813 (run_gcc): Read priorities and if doing parallel build order
1814 the Makefile by them.
1816 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1818 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1820 2018-04-30 Richard Biener <rguenther@suse.de>
1822 * tree-cfg.c (verify_address): Remove base argument, add
1823 flag whether to check TREE_ADDRESSABLE and do that.
1824 (verify_expr): Remove.
1825 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1826 (verify_gimple_assign_single): Likewise.
1827 (verify_gimple_switch): Likewise.
1828 (verify_expr_location_1): Dereference tp once. Add (disabled)
1829 piece from verify_expr.
1830 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1832 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1834 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1836 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1838 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1839 (small_data_pattern): Likewise.
1840 (arc_rewrite_small_data): Likewise.
1841 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1842 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1843 (get_symbol_alignment): New function.
1844 (legitimate_small_data_address_p): Likewise.
1845 (legitimate_scaled_address): Update, call
1846 legitimate_small_data_address_p.
1847 (output_sdata): New static variable.
1848 (arc_print_operand): Update how we handle small data operands.
1849 (arc_print_operand_address): Likewise.
1850 (arc_legitimate_address_p): Update, use
1851 legitimate_small_data_address_p.
1852 (arc_rewrite_small_data_p): Remove.
1853 (arc_rewrite_small_data_1): Likewise.
1854 (arc_rewrite_small_data): Likewise.
1855 (small_data_pattern): Likewise.
1856 (compact_sda_memory_operand): Update to use
1857 legitimate_small_data_address_p and get_symbol_alignment.
1858 (prepare_move_operands): Don't rewite sdata pattern.
1859 (prepare_extend_operands): Remove.
1860 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1862 (zero_extendqisi2): Likewise.
1863 (zero_extendhisi2): Likewise.
1864 (extendqihi2): Likewise.
1865 (extendqisi2): Likewise.
1866 (extendhisi2): Likewise.
1870 * config/arc/constraints.md (Usd): Change it to memory constraint.
1872 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1874 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1875 as source of std instructions.
1876 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1877 allow 6-bit constants as source for store instructions.
1878 (movdi_insn): Update instruction pattern to allow 6-bit constants
1879 as source for store instructions.
1881 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1883 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1885 2018-04-30 Nathan Sidwell <nathan@acm.org>
1886 Sandra Loosemore <sandra@codesourcery.com>
1888 * dumpfile.c (dump_open): Allow '-' for stdout.
1889 * doc/invoke.texi (Developer Options): Document dump filename
1890 determination early. Document stdin/stdout selection.
1892 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1894 Microblaze Target: PIC data text relative
1896 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1897 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1899 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1900 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1901 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1902 New addressing mode for data-text relative position indepenedent code.
1903 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1904 'ADDRESS_SYMBOLIC_TXT_REL'.
1905 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1906 (microblaze_legitimate_pic_operand): Exclude function calls from
1907 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1908 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1910 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1911 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1912 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1913 for 'address + offset'.
1914 (microblaze_expand_prologue): Add new function prologue call for
1916 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1917 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1918 table in case of TARGET_PIC_DATA_TEXT_REL.
1919 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1920 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1921 Add new macros 'UNSPEC_TEXT',
1922 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1923 + exclude function calls from 'UNSPEC_PLT' in case of data text
1925 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1926 new target hook for generating address diff vector tables in case of
1928 * doc/tm.texi : Regenerate.
1929 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1930 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1931 of addr diff vector generation.
1932 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1933 target hook definition.
1934 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1935 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1936 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1937 option for data text relative.
1939 2018-04-30 Richard Biener <rguenther@suse.de>
1941 * tree-chrec.h (evolution_function_is_constant_p): Remove
1943 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1945 2018-04-30 Richard Biener <rguenther@suse.de>
1948 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1950 2018-04-30 Richard Biener <rguenther@suse.de>
1952 PR tree-optimization/28364
1953 PR tree-optimization/85275
1954 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1955 copying first exit test.
1957 2018-04-28 Mark Wielaard <mark@klomp.org>
1959 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1961 (dwarf_AT): Handle DW_AT_addr_base.
1962 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1964 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1967 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1968 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1969 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1970 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1972 2018-04-28 Richard Biener <rguenther@suse.de>
1974 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1975 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1976 to reflect use. Only add interesting stmts.
1978 2018-04-27 Martin Jambor <mjambor@suse.cz>
1981 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1982 the jump function allows for passing through aggregate values.
1984 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1986 * input.h (in_system_header_at): Convert from macro to inline
1988 (from_macro_expansion_at): Likewise.
1989 (from_macro_definition_at): Likewise.
1991 2018-04-27 Jeff Law <law@redhat.com>
1993 * config.gcc: Mark tile* targets as deprecated/obsolete.
1995 2018-04-27 Richard Biener <rguenther@suse.de>
1997 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2000 2018-04-27 Richard Biener <rguenther@suse.de>
2002 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2004 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2006 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2007 with Yd constraint. Set "preferred_for_speed" attribute from
2008 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2010 (*movdi_internal): Ditto.
2011 (movti_interunit splitters): Remove
2012 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2013 (movdi_interunit splitters): Ditto.
2014 * config/i386/constraints.md (Ye): Remove.
2015 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2017 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2020 * config/aarch64/constraints.md (Usg): Limit to 31.
2023 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2025 PR tree-optimization/85529
2026 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2027 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2028 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2029 zero extension or masking of the MSB bit.
2030 (optimize_range_tests): Add FIRST_BB argument, pass it through
2031 to optimize_range_tests_var_bound.
2032 (maybe_optimize_range_tests, reassociate_bb): Adjust
2033 optimize_range_tests callers.
2035 2018-04-26 Richard Biener <rguenther@suse.de>
2036 Jakub Jelinek <jakub@redhat.com>
2038 * cgraph.h (symbol_table): Just declare debug method here.
2039 * symtab.c (symbol_table::debug): Define.
2041 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2043 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2045 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2047 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2048 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2049 (*movdi_internal): Substitute Yi and Yj constraint with x
2050 and Ym and Yn constraint with y constraint. Update "isa"
2051 attribute and set "preferred_for_speed" attribute from
2052 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2053 (*movsi_internal): Ditto.
2054 (*movdf_internal): Ditto.
2055 (*movsf_internal): Ditto.
2056 (*zero_extendsidi2): Ditto.
2057 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2058 (sse2_loadld): Ditto.
2059 (*vec_extract<ssevecmodelower>_0): Ditto.
2060 (*vec_extractv4si_0_zext_sse4): Ditto.
2061 (vec_concatv2di): Ditto.
2062 (*vec_dup<mode>): Ditto.
2063 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2064 * config/i386/constraints.md (Yi): Remove.
2069 2018-04-26 Nathan Sidwell <nathan@acm.org>
2071 * dumpfile.c (dump_open): New.
2072 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2073 (dump_finish): Detect stdio/stderr by value not name.
2075 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2077 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2079 2018-04-26 Tom de Vries <tom@codesourcery.com>
2082 * config/nvptx/nvptx.c (verify_neutering_jumps)
2083 (verify_neutering_labels): New function
2084 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2086 2018-04-26 Tom de Vries <tom@codesourcery.com>
2089 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2090 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2093 2018-04-26 Richard Biener <rguenther@suse.de>
2094 Tom de Vries <tom@codesourcery.com>
2097 * lto-streamer-out.c (output_function): Fixup loops if required to match
2098 discovery done in the reader.
2100 2018-04-26 Richard Biener <rguenther@suse.de>
2102 PR tree-optimization/85116
2103 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2104 have a loop exit from the single latch predecessor. Remove
2105 case of header with just condition.
2106 (ch_base::copy_headers): Exclude infinite loops from any
2108 (pass_ch::execute): Record exits.
2110 2018-04-26 Richard Biener <rguenther@suse.de>
2112 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2113 prologue cost vector and pass it to vect_get_load_cost.
2114 (vect_get_peeling_costs_all_drs): Likewise.
2115 (vect_peeling_hash_get_lowest_cost): Likewise.
2116 (vect_enhance_data_refs_alignment): Likewise.
2118 2018-04-26 Richard Biener <rguenther@suse.de>
2121 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2122 checking of integer<->pointer conversions.
2123 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2124 sign-/zero-extending pointer types.
2125 (expand_omp_for_static_chunk): Likewise.
2127 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2128 Jean Lee <xiaoyur347@gmail.com>
2130 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2131 (TARGET_ASAN_SHADOW_OFFSET): Define.
2132 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2133 true for -fsanitize=address.
2135 2018-04-25 Mark Wielaard <mark@klomp.org>
2137 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2140 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2142 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2143 than "alu", remove explicit "memory" and "imm_disp" attributes.
2144 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2147 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2148 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2149 gen_lowpart_no_emit.
2151 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2154 * config/i386/i386.c (ix86_expand_builtin): Change memory
2155 operand to XI, extend p0 to Pmode.
2156 * config/i386/i386.md: Change unspec volatile and operand
2157 1 mode to XI, change operand 0 mode to P.
2159 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2161 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2162 GET_MODE_MASK before any checking.
2163 (nds32_can_use_bset_p): Likewise.
2164 (nds32_can_use_btgl_p): Likewise.
2166 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2168 * config/nds32/nds32-doubleword.md: New define_split pattern for
2169 illegal register number.
2171 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2173 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2175 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2177 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2179 2018-04-25 Richard Biener <rguenther@suse.de>
2181 * lto-streamer.h (LTO_major_version): Bump to 8.
2183 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2185 * BASE-VER: Set to 9.0.0.
2187 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2189 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2190 in __abskf2 and __powikf2.
2192 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2195 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2196 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2197 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2198 Use the above on operand 2. Reindent.
2199 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2201 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2204 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2206 * config/i386/i386.opt (mcet): Removed.
2207 * doc/install.texi: Remove -mcet documentation.
2208 * doc/invoke.texi: Likewise.
2210 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2213 * doc/install.texi: Remove -mcet from bootstrap-cet.
2215 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2218 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2219 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2223 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2224 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2225 containing a CONST_VECTOR.
2227 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2229 * doc/install.texi: Update newlib dependency for nvptx.
2231 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2234 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2235 instead of INTVAL when shifting x left.
2237 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2239 PR tree-optimization/85478
2240 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2241 vect_grouped_store_supported for single element vectors.
2243 2018-04-24 Richard Biener <rguenther@suse.de>
2246 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2247 load cost increase to the case of non-constant step.
2249 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2252 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2253 destination if any_malformed_asm.
2255 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2258 * expr.c (store_field): In the bitfield case, if the value comes from
2259 a function call and is returned in registers by means of a PARALLEL,
2260 do not change the mode of the temporary unless BLKmode and VOIDmode.
2262 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2264 PR rtl-optimization/85423
2265 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2266 dependencies to debug insns when the previous insn is non-debug.
2268 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2270 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2271 enums into a single definition.
2272 (fls): Fix predicates and printing.
2275 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2277 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2278 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2279 and short u6 immediate.
2280 (check_if_valid_sleep_operand): Remove.
2281 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2283 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2285 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2286 flag_always_save_lp condition.
2287 * config/nds32/nds32.opt (malways-save-lp): New option.
2289 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2291 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2292 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2293 * config/nds32/nds32.h
2294 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2295 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2297 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2299 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2300 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2302 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2303 Chung-Ju Wu <jasonwucj@gmail.com>
2305 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2307 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2308 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2310 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2312 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2314 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2316 * config/nds32/nds32-protos.h (nds32_data_alignment,
2317 nds32_local_alignment): Declare.
2318 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2319 nds32_local_alignment): New functions.
2320 (TARGET_CONSTANT_ALIGNMENT): Define.
2321 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2323 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2325 * config/nds32/nds32.c
2326 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2327 (TARGET_MODES_TIEABLE_P): Likewise.
2329 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2331 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2334 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2335 Chung-Ju Wu <jasonwucj@gmail.com>
2337 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2338 for unaligned access.
2339 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2341 * config/nds32/nds32-intrinsic.md: Likewise.
2342 * config/nds32/nds32_intrinsic.h: Likewise.
2343 * config/nds32/nds32.h (nds32_builtins): Likewise.
2344 * config/nds32/nds32.opt (munaligned-access): New option.
2345 * config/nds32/nds32.c (nds32_asm_file_start): Display
2346 flag_unaligned_access status.
2348 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2350 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2351 -mno-relax is present.
2352 * config/riscv/linux.h (LINK_SPEC): Ditto.
2354 2018-04-20 Martin Sebor <msebor@redhat.com>
2357 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2359 (gimple_fold_builtin_stxcpy_chk): Same.
2360 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2362 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2365 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2366 __powikf2 when long double is IEEE 128-bit.
2368 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2370 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2371 step to make sure stack always aligned.
2373 2018-04-20 Carl Love <cel@us.ibm.com>
2376 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2377 size check for arg0.
2379 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2380 Tom de Vries <tom@codesourcery.com>
2383 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2384 Emit insns for calls too.
2385 (nvptx_find_par): Always look for worker-level predecessor insn.
2386 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2388 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2389 (nvptx_process_pars): Propagate frames for calls.
2391 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2394 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2396 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2397 (ix86_handle_option): Don't handle OPT_mibt.
2398 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2400 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2402 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2403 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2404 (ix86_target_macros): Define __CET__ with flag_cf_protection
2405 for -fcf-protection.
2406 * config/i386/i386.c (isa2_opts): Remove -mibt.
2407 * config/i386/i386.h (TARGET_IBT): Removed.
2408 (TARGET_IBT_P): Likewise.
2409 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2410 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2411 * config/i386/i386.opt (mcet): Update help message.
2414 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2415 -mcet as an alias for -mshstk.
2417 2018-04-20 Richard Biener <rguenther@suse.de>
2420 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2421 complexity by forcing a single use of the multiply operand.
2423 2018-04-20 Martin Jambor <mjambor@suse.cz>
2426 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2427 recursion dependency to only apply to non-clones.
2429 2018-04-20 Martin Jambor <mjambor@suse.cz>
2432 * ipa-cp.c (create_specialized_node): Check that clones of
2433 self-recursive edges exist during IPA-CP.
2435 2018-04-19 Toon Moene <toon@moene.org>
2437 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2440 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2442 PR tree-optimization/85467
2443 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2444 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2445 VECTOR_CST element to type.
2447 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2450 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2451 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2452 (builtin_longjmp): Likewise.
2453 (save_stack_nonlocal): New pattern.
2454 (restore_stack_nonlocal): Likewise.
2456 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2459 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2460 Replace ASM_OUTPUT_LABEL with fprintf.
2462 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2465 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2466 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2467 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2468 define __IBT__ and __SHSTK__ for -fcf-protection.
2469 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2471 (ix86_trampoline_init): Likewise.
2472 (x86_output_mi_thunk): Likewise.
2473 (ix86_notrack_prefixed_insn_p): Likewise.
2474 (ix86_option_override_internal): Don't disallow -fcf-protection.
2475 * config/i386/i386.md (rdssp<mode>): Also enable for
2477 (incssp<mode>): Likewise.
2478 (nop_endbr): Likewise.
2479 * config/i386/i386.opt (mcet): Change help message to built-in
2483 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2484 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2485 enable CET built-in functions.
2487 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2489 * common/config/i386/i386-common.c
2490 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2491 OPTION_MASK_ISA_MOVDIRI_UNSET,
2492 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2493 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2494 * config.gcc (movdirintrin.h): New header.
2495 * config/i386/cpuid.h (bit_MOVDIRI,
2496 bit_MOVDIR64B): New bits.
2497 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2499 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2500 (VOID, PVOID, PCVOID)): New function types.
2501 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2502 __builtin_ia32_directstoreu_u64,
2503 __builtin_ia32_movdir64b): New builtins.
2504 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2505 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2507 (ix86_valid_target_attribute_inner_p): Ditto.
2508 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2509 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2510 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2511 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2512 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2513 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2514 (movdiri<mode>, movdir64b_<mode>): New.
2515 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2516 * config/i386/immintrin.h: Include movdirintrin.h.
2517 * config/i386/movdirintrin.h: New file.
2518 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2520 2018-04-19 Richard Biener <rguenther@suse.de>
2523 * cfg.c (clear_bb_flags): When loop state says we have
2524 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2526 2018-04-19 Richard Biener <rguenther@suse.de>
2528 PR tree-optimization/84737
2529 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2530 copying restrict info.
2531 (vect_setup_realignment): Use it.
2532 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2533 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2534 the first DR to all generated stores.
2535 (vectorizable_load): Likewise for loads.
2537 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2539 PR tree-optimization/85446
2540 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2541 the integral and pointer types to have the same precision.
2543 * doc/install.texi: Document --disable-cet being the default and
2546 2018-04-18 Martin Liska <mliska@suse.cz>
2548 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2551 2018-04-18 Martin Liska <mliska@suse.cz>
2554 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2557 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2558 arguments if they are comparable.
2560 2018-04-18 Martin Liska <mliska@suse.cz>
2563 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2566 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2569 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2572 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2573 ENDBR after calling __morestack.
2575 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2578 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2579 by using gcc_base_ver to generate a gcc_driver_version, and use
2580 it when generating GCC_DRIVER_NAME.
2581 * configure: Regenerate.
2583 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2586 * config.gcc: Obsolete powerpc*-*-*spe*.
2588 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2591 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2592 (stabstr_D): Change type of unum from unsigned int to
2593 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2596 2018-04-17 Jim Wilson <jimw@sifive.com>
2599 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2600 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2601 Set arg_pointer_offset after using pretend_args_size.
2603 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2605 PR rtl-optimization/85431
2606 * dse.c (record_store): Ignore zero width stores.
2609 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2610 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2611 __builtin_stack_restore rather than after it.
2612 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2613 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2614 argument instead of virtual_dynamic_stack_rtx.
2616 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2618 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2620 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2621 Add note to error message to explain internal mapping of overloaded
2622 built-in function name to non-overloaded built-in function name.
2623 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2626 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2629 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2630 where the inputs overlap with the output.
2632 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2634 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2635 (=v, v) alternative and explicit "memory" attribute.
2636 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2637 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2639 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2640 "sselog1" type instead of "sselog".
2641 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2642 "sselog". Remove explicit "memory" attribute.
2643 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2644 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2646 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2647 "sselog1" type instead of "sselog". Remove explicit "memory"
2649 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2650 use "sselog1" type instead of "sselog". Remove explicit "memory"
2652 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2653 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2655 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2656 "sselog1" type instead of "sselog". Remove explicit "memory"
2658 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2659 use "sselog1" type instead of "sselog". Remove explicit "memory"
2663 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2666 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2669 2018-04-17 Martin Jambor <mjambor@suse.cz>
2672 * ipa-cp.c (create_specialized_node): Call
2673 expand_all_artificial_thunks if necessary.
2675 2018-04-17 Martin Liska <mliska@suse.cz>
2678 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2679 in message, remote space in between '_G' and '('.
2681 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2684 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2685 avx512f_vmcmp<mode>3<round_saeonly_name>,
2686 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2687 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2688 avx512f_rndscale<mode><round_saeonly_name>,
2689 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2690 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2691 Use %<iptr>2 instead of %2 for -masm=intel.
2692 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2693 avx512f_vcvttss2usi<round_saeonly_name>,
2694 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2696 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2697 avx512f_vcvttsd2usi<round_saeonly_name>,
2698 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2699 Use %q1 instead of %1 for -masm=intel.
2700 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2701 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2702 of %3 for -masm=intel.
2703 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2705 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2707 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2709 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2710 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2712 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2714 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2715 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2716 %g1 and one with %0 and %1.
2717 (avx512er_vmrcp28<mode><round_saeonly_name>,
2718 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2720 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2721 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2722 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2723 of %0 and %{%4%} for -masm=intel.
2724 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2725 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2726 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2727 order of %0 and %{%5%}%{z%} for -masm=intel.
2729 2018-04-17 Jan Hubicka <jh@suse.cz>
2732 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2734 2018-04-17 Martin Liska <mliska@suse.cz>
2737 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2738 for target_clone error message. Make default implementation
2739 clone to be a local declaration.
2740 (separate_attrs): Add new argument and check for an empty
2742 (expand_target_clones): Handle it.
2743 (ipa_target_clone): Make redirection just for target_clones
2746 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2747 Tom de Vries <tom@codesourcery.com>
2750 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2751 tiled basic blocks without omp continue statements.
2753 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2756 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2757 vec_extract expression as having side effects to make sure it gets
2760 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2763 * config/i386/i386.c (get_builtin_code_for_version): Check
2766 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2769 * gcc/config.gcc: Support "skylake".
2770 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2772 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2773 (processor_target_table): Add "skylake".
2774 (ix86_option_override_internal): Add "skylake".
2775 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2776 PROCESSOR_CANNONLAKE.
2777 (get_builtin_code_for_version): Fix priority for
2778 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2779 PROCESSOR_SKYLAKE-AVX512.
2780 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2781 (processor_type): Add PROCESSOR_SKYLAKE.
2783 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2784 Jason Merrill <jason@redhat.com>
2787 * convert.c (convert_to_integer_1): Use direct recursion for
2788 enumeral types and types with a precision less than the number
2789 of bits in their mode.
2791 2018-04-16 Julia Koval <julia.koval@intel.com>
2794 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2795 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2797 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2800 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2801 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2802 and -mno-direct-move.
2804 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2807 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2808 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2809 Ensure negative shifts result in {0}.
2811 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2813 PR rtl-optimization/79916
2814 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2815 regs (if any) to define how to gnerate SD moves when LRA is in
2818 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2820 PR rtl-optimization/85393
2821 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2822 * except.c (expand_dw2_landing_pad_for_region): Make static.
2823 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2824 a label and unconditional jump to old_bb, rather than
2825 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2828 PR rtl-optimization/85376
2829 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2830 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2831 instead of a specific value.
2833 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2834 Bin Cheng <bin.cheng@arm.com>
2836 PR tree-optimization/82965
2837 PR tree-optimization/83991
2838 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2839 by_profile_only parameter.
2840 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2841 information if the loop was predicted to iterate too many times.
2842 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2844 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2847 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2850 2018-04-13 Martin Liska <mliska@suse.cz>
2851 Jakub Jelinek <jakub@redhat.com>
2854 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2855 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2856 * builtins.c (expand_builtin_memory_copy_args): Use
2857 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2858 handle dest_addr == pc_rtx.
2860 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2863 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2864 asked to not generate direct moves.
2865 (fix_trunc<mode>si2_stfiwx): Similar.
2866 (fix_trunc<mode>si2_internal): Similar.
2868 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2871 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2872 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2873 lookup if dest in some wider mode is known to be const0_rtx and
2874 if so, record permanent equivalence for it to be ZERO_EXTEND of
2875 the narrower mode destination.
2877 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2879 * lto-streamer-out.c (output_function): Revert 259346.
2880 * omp-expand.c (expand_oacc_for): Likewise.
2882 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2884 PR rtl-optimization/85354
2885 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2886 * sel-sched.c (sel_global_init): ... here.
2888 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2891 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2892 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2893 mode for PE-COFF targets.
2894 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2895 (i386_pe_asm_lto_end): Likewise.
2896 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2897 (TARGET_ASM_LTO_END): Likewise.
2898 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2899 (i386_pe_asm_lto_start): New function.
2900 (i386_pe_asm_lto_end): Likewise.
2902 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2903 Richard Biener <rguenther@suse.de>
2906 * lto-streamer-out.c (output_function): Fix CFG loop state before
2908 * omp-expand.c (expand_oacc_for): Handle calls to internal
2909 functions like regular functions.
2911 2018-04-12 Richard Biener <rguenther@suse.de>
2914 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2915 for the early LTO debug to properly generate references to it
2916 during DIE emission. Do not re-use that for the skeleton for
2918 (dwarf2out_early_finish): Likewise.
2920 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2923 * config/i386/sse.md
2924 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2925 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2926 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2927 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2928 and output is a reg, avoid creating invalid lowpart subreg, but
2929 instead split into a 512-bit move. Don't split if not AVX512VL,
2930 input is xmm16+ reg and output is a mem.
2931 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2932 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2933 xmm16+ reg and output is a mem.
2935 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2937 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2938 also for flag_dwarf2_cfi_asm.
2940 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2942 PR rtl-optimization/85342
2943 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2944 a bool scalar var inside of the loop instead. Don't try to update
2945 recog_data.operand after failed apply_change_group.
2947 2018-04-12 Tom de Vries <tom@codesourcery.com>
2950 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2951 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2952 array with flexible array member as array without given dimension.
2953 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2954 argument for undefined param to true.
2956 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2959 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2960 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2961 from PowerPC section.
2962 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2963 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2964 help text that is too long.
2965 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2966 help text that is too long.
2967 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2968 help text that is too long.
2970 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2972 * config/alpha/alpha.md (stack_probe_internal): Rename
2973 from "probe_stack". Update all callers.
2975 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2977 PR rtl-optimization/84566
2978 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2979 sched_macro_fuse_insns.
2981 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2984 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2985 (compute_block_dependences): ... from here.
2987 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2989 PR tree-optimization/85331
2990 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2991 from int to HOST_WIDE_INT.
2993 2018-04-11 Martin Jambor <mjambor@suse.cz>
2996 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2997 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2998 not the same as the source val.
2999 (cgraph_edge_brings_value_p): New parameter.
3000 (gather_edges_for_value): Pass destination value to
3001 cgraph_edge_brings_value_p.
3002 (perhaps_add_new_callers): Likewise.
3003 (get_info_about_necessary_edges): Likewise and exclude values brought
3004 only by self-recursive edges.
3005 (create_specialized_node): Redirect only clones of self-calling edges.
3006 (+self_recursive_pass_through_p): New function.
3007 (find_more_scalar_values_for_callers_subset): Use it.
3008 (find_aggregate_values_for_callers_subset): Likewise.
3009 (known_aggs_to_agg_replacement_list): Removed.
3010 (decide_whether_version_node): Re-calculate known constants for all
3011 remaining context clones.
3013 2018-04-11 Richard Biener <rguenther@suse.de>
3016 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3017 from early DWARF output.
3018 (dwarf2out_early_finish): Output line info unconditionally into
3019 early DWARF and add reference to it.
3021 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3024 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3025 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3026 other than V2DFmode using iptr mode attribute.
3027 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3029 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3031 PR rtl-optimization/84659
3032 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3034 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3037 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3039 (output_loc_list): Pass address of a dummy size variable even in the
3040 locview handling loop.
3041 (index_location_lists): Add comment on why skip_loc_list_entry can't
3044 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3047 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3050 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3053 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3054 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3055 and -mstring-compare-inline-limit.
3057 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3060 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3061 for stack clash protection in a register whenever we need it to be in
3064 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3066 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3067 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3069 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3072 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3074 (mlong-double-): Ditto.
3075 * config/rs6000/sysv4.opt (msdata=): Ditto.
3076 (mtls-size=): Ditto.
3078 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3080 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3081 erroneous entries for
3082 "vector int vec_ldl (int, long int *)", and
3083 "vector unsigned int vec_ldl (int, unsigned long int *)".
3084 Add comments and entries for
3085 "vector bool char vec_ldl (int, bool char *)",
3086 "vector bool short vec_ldl (int, bool short *)",
3087 "vector bool int vec_ldl (int, bool int *)",
3088 "vector bool long long vec_ldl (int, bool long long *)",
3089 "vector pixel vec_ldl (int, pixel *)",
3090 "vector long long vec_ldl (int, long long *)",
3091 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3092 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3093 type tree bool_long_long_type_node and correct definition of
3094 bool_V2DI_type_node to make reference to this new type tree.
3095 (rs6000_mangle_type): Replace erroneous reference to
3096 bool_long_type_node with bool_long_long_type_node.
3097 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3098 comments to emphasize sign distinctions for char and int types and
3099 replace RS6000_BTI_bool_long constant with
3100 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3101 use of RS6000_BTI_pixel.
3102 (bool_long_type_node): Remove this macro definition.
3103 (bool_long_long_type_node): New macro definition
3105 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3107 PR rtl-optimization/85300
3108 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3109 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3110 simplify_unary_operation fails.
3112 2018-04-10 Martin Liska <mliska@suse.cz>
3114 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3115 cgraph_edge and ipa_ref.
3117 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3121 * config/i386/sse.md
3122 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3123 computation of the VEC_MERGE selector from mask.
3124 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3125 Fix decoding of the VEC_MERGE selector into mask.
3127 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3129 PR tree-optimization/85286
3130 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3132 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3134 * final.c (final_1): Set insn_last_address as well as
3135 insn_current_address.
3137 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3140 * explow.c (emit_stack_probe): Call validize_mem on memory location
3141 before passing it to gen_probe_stack. Create address operand and
3142 legitimize it for the probe_stack_address case.
3144 2018-04-09 Jan Hubicka <jh@suse.cz>
3147 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3148 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3149 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3151 (free_lang_data): Rebuild type inheritance graph.
3153 2018-04-09 Martin Sebor <msebor@redhat.com>
3155 * invoke.texi (-finline-small-functions): Mention other optimization
3157 (-findirect-inlining, -fpartial-inlining): Same.
3158 (-finline-functions-called-once): Same.
3159 (-freorder-blocks-and-partition): Same.
3161 2018-04-09 Jan Hubicka <jh@suse.cz>
3164 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3165 jumps; choose last target that matches the criteria (i.e.
3166 no partition changes for non-crossing jumps).
3167 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3168 support for redirecting crossing jumps to non-crossing.
3170 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3172 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3173 also for naked functions.
3175 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3177 * config/arc/arc.md (add_shift): New pattern.
3178 (add_shift2): Likewise.
3179 (sub_shift): Likewise.
3180 (sub_shift_cmp0_noout): Likewise.
3181 (compare_si_ashiftsi): Likewise.
3182 (xbfu_cmp0_noout): New combine pattern.
3183 (xbfu_cmp0"): Likewise.
3184 (movsi_set_cc_insn): Place the predicable variant first.
3185 (commutative_binary_cmp0_noout): Remove clobber.
3186 (commutative_binary_cmp0): New pattern.
3187 (noncommutative_binary_cmp0): Likewise.
3188 (noncommutative_binary_cmp0_noout): Likewise.
3189 (noncommutative_binary_comparison_result_used): Removed.
3190 (rsub_cmp0): New pattern.
3191 (rsub_cmp0_noout): Likewise.
3192 (extzvsi): Changed, keep only meaningful variants.
3193 (SQH, SEZ): New iterators.
3194 (SQH_postfix): New mode attribute.
3195 (SEZ_prefix): New code attribute.
3196 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3197 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3198 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3200 (noncommutative_operator): Check the availability of barrel
3203 2018-04-09 Richard Biener <rguenther@suse.de>
3205 PR tree-optimization/85284
3206 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3207 Only use the niter constraining form of simple_iv when the exit
3210 2018-04-09 Tom de Vries <tom@codesourcery.com>
3213 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3214 (define_expand "*memory_barrier"): New define_expand.
3215 (define_insn "memory_barrier"): New insn.
3217 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3219 PR rtl-optimization/80463
3220 PR rtl-optimization/83972
3221 PR rtl-optimization/83480
3223 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3224 correct producer for the insn.
3225 (tidy_control_flow): Fixup seqnos in case of debug insns.
3227 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3229 PR rtl-optimization/83913
3231 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3232 different sched-times when merging exprs.
3234 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3236 PR rtl-optimization/83962
3238 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3239 tidy_fallthru_edge and tidy_control_flow.
3241 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3243 PR rtl-optimization/83530
3245 * sel-sched.c (force_next_insn): New global variable.
3246 (remove_insn_for_debug): When force_next_insn is true, also leave only
3247 next insn in the ready list.
3248 (sel_sched_region): When the region wasn't scheduled, make another pass
3249 over it with force_next_insn set to 1.
3251 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3253 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3255 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3256 for interrupt control.
3257 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3258 functions for interrupt control.
3259 * config/nds32/nds32-intrinsic.md: Likewise.
3260 * config/nds32/nds32_intrinsic.h: Likewise.
3261 * config/nds32/nds32.h (nds32_builtins): Likewise.
3263 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3265 * config/nds32/nds32.c (nds32_init_machine_status,
3266 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3267 strict_aligned_p field.
3268 (nds32_expand_to_rtl_hook): New function.
3269 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3270 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3272 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3273 Chung-Ju Wu <jasonwucj@gmail.com>
3275 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3276 * config/nds32/nds32-n7.md: New file.
3277 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3278 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3280 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3281 * config/nds32/nds32.md (pipeline_model): Add n7.
3282 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3283 * config/nds32/pipelines.md: Include n7 settings.
3285 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3286 Chung-Ju Wu <jasonwucj@gmail.com>
3288 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3289 * config/nds32/nds32-e8.md: New file.
3290 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3291 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3293 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3294 * config/nds32/nds32.md (pipeline_model): Add e8.
3295 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3296 * config/nds32/pipelines.md: Include e8 settings.
3298 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3299 Chung-Ju Wu <jasonwucj@gmail.com>
3301 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3302 * config/nds32/nds32-n8.md: New file.
3303 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3304 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3306 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3307 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3308 * config/nds32/nds32.md (pipeline_model): Add n8.
3309 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3310 * config/nds32/pipelines.md: Include n8 settings.
3312 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3313 Chung-Ju Wu <jasonwucj@gmail.com>
3315 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3316 * config/nds32/nds32-n9-2r1w.md: New file.
3317 * config/nds32/nds32-n9-3r2w.md: New file.
3318 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3319 nds32_register_ports): New or modify for cpu n9.
3320 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3322 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3323 * config/nds32/nds32-utils.c: New file.
3324 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3325 TARGET_MUL_SLOW): Define.
3326 * config/nds32/nds32.md (pipeline_model): New attribute.
3327 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3328 New options that support cpu n9.
3329 * config/nds32/pipelines.md: Include n9 settings.
3330 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3332 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3334 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3335 information if necessary.
3336 (output_cond_branch_compare_zero): Likewise.
3337 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3338 (nds32_target_alignment): Refine for alignment.
3339 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3340 (FUNCTION_BOUNDARY): Modify.
3341 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3343 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3345 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3347 * config/nds32/constants.md (unspec_volatile_element): Add values for
3348 TLB operation and data prefetch.
3349 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3350 functions for TLB operation and data prefetch.
3351 * config/nds32/nds32-intrinsic.md: Likewise.
3352 * config/nds32/nds32_intrinsic.h: Likewise.
3353 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3354 (nds32_print_operand): Likewise.
3355 * config/nds32/nds32.h (nds32_builtins): Likewise.
3357 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3358 Andrew Pinski <pinsika@gcc.gnu.org>
3361 * match.pd: Use constant_boolean_node of correct type instead of
3362 boolean_true_node or boolean_false_node for simplifying
3363 pointer comparisons to zero.
3365 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3367 PR tree-optimization/80021
3368 * tree.c (verify_type_variant): Make error call in verify_variant_match
3369 translatable and remove final full stop.
3371 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3373 * config/nds32/constants.md (unspec_volatile_element): Add
3374 UNSPEC_VOLATILE_EH_RETURN.
3375 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3376 nds32_output_stack_pop): Support dwarf exception handling process.
3377 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3378 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3379 exception handling process.
3380 (nds32_compute_stack_frame): Likewise.
3381 (nds32_return_addr_rtx): Likewise.
3382 (nds32_initial_elimination_offset): Likewise.
3383 (nds32_expand_prologue): Likewise.
3384 (nds32_expand_epilogue): Likewise.
3385 (nds32_dynamic_chain_address): New function.
3386 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3388 (DYNAMIC_CHAIN_ADDRESS): Define.
3389 (EH_RETURN_DATA_REGNO): Define.
3390 (EH_RETURN_STACKADJ_RTX): Define.
3391 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3392 patterns for dwarf exception handling.
3394 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3396 * config/nds32/nds32.h: Clean up obsolete macros.
3398 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3400 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3401 Add enum values for particular instructions.
3402 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3403 particular intrinsic functions.
3404 * config/nds32/nds32-intrinsic.md: Likewise.
3405 * config/nds32/nds32_intrinsic.h: Likewise.
3406 * config/nds32/nds32.h (nds32_builtins): Likewise.
3407 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3408 (btst, ave): New patterns for particular instructions.
3410 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3412 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3413 Add enum values for atomic load/store and memory sync.
3414 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3416 * config/nds32/nds32-intrinsic.md: Likewise.
3417 * config/nds32/nds32_intrinsic.h: Likewise.
3418 * config/nds32/nds32.h (nds32_builtins): Likewise.
3420 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3422 PR tree-optimization/85257
3423 * fold-const.c (native_encode_vector): If not all elts could fit
3424 and off is -1, return 0 rather than offset.
3425 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3426 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3427 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3428 adjust buffer in native_interpret_expr call.
3430 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3432 * config/nds32/constants.md (unspec_volatile_element): Add cache
3433 control enum values.
3434 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3435 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3436 * config/nds32/nds32.c (nds32_cctl_names): New.
3437 (nds32_print_operand): Handle cache control register names.
3438 * config/nds32/nds32.h (nds32_builtins): New enum values.
3439 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3441 * config/nds32/nds32.md (type): Add mmu.
3442 * config/nds32/pipelines.md (simple_insn): Add mmu.
3444 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3446 * config/nds32/nds32.md (type): Remove call.
3447 * config/nds32/pipelines.md (simple_insn): Likewise.
3449 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3451 * config/nds32/constants.md (unspec_volatile_element): Add
3452 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3453 UNSPEC_VOLATILE_FMFCFG.
3454 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3455 description for fmfcfg and fmfcsr.
3456 (bdesc_1arg): Add fmtcsr.
3457 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3458 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3459 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3460 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3461 unspec_fmfcfg): New patterns.
3462 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3463 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3464 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3465 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3466 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3467 __nds32__fmfcfg): Define.
3469 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3471 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3472 intrinsic register names.
3473 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3474 intrinsic register enum values and macros.
3476 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3478 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3479 for load/store addressing form.
3480 (nds32_print_operand_address): Likewise.
3482 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3485 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3486 based on LABEL_REF. Remove useless assertion.
3487 (pic_address_needs_scratch): Fix formatting.
3488 (sparc_legitimize_pic_address): Minor tweaks.
3489 (sparc_delegitimize_address): Adjust assertion accordingly.
3490 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3491 into symbolic_operand.
3492 (movsi_high_pic_label_ref): Likewise.
3493 (movsi_lo_sum_pic_label_ref): Likewise.
3494 (movdi_pic_label_ref): Likewise.
3495 (movdi_high_pic_label_ref): Likewise.
3496 (movdi_lo_sum_pic_label_ref): Likewise.
3498 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3500 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3501 custom LIB_SPEC setup.
3503 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3504 Kito Cheng <kito.cheng@gmail.com>
3506 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3507 * config/riscv/freebsd.h: New.
3509 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3511 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3512 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3515 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3516 Kito Cheng <kito.cheng@gmail.com>
3518 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3519 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3520 * config/nds32/nds32-protos.h (nds32_output_call,
3521 nds32_output_return): Declare.
3522 * config/nds32/nds32.md: Refine all the call and return patterns.
3524 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3527 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3528 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3530 PR rtl-optimization/84872
3531 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3532 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3535 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3537 * expr.c (copy_blkmode_to_reg): Revert 254862.
3538 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3540 2018-04-06 Richard Biener <rguenther@suse.de>
3543 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3544 after seeing a component reference with an adjacent field. Treat
3545 refs to arrays at struct end of external decls similar to
3546 refs to unconstrained commons.
3548 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3551 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3552 look through SAVE_EXPRs with non-side-effects argument. Adjust
3554 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3557 2018-04-06 Richard Biener <rguenther@suse.de>
3560 * alias.c (find_base_term): New wrapper around find_base_term
3561 unwinding CSELIB_VAL_PTR changes.
3562 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3565 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3567 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3569 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3570 constant definitions.
3571 ("nop"): lr 0,0 -> nopr r0
3572 ("nop_lr0", "nop_lr1"): New insn definitions.
3574 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3576 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3577 NDS32_V3PUSH_AVAILABLE_P macro.
3579 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3580 Chung-Ju Wu <jasonwucj@gmail.com>
3582 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3583 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3584 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3585 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3586 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3587 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3588 * config/nds32/constraints.md: New constraints and checking for hard
3589 float configuration.
3590 * config/nds32/iterators.md: New mode iterator and attribute for hard
3591 float configuration.
3592 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3594 * config/nds32/nds32-fpu.md: New file.
3595 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3596 deal with hard float code generation.
3597 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3599 (abi_type, float_reg_number): New enum type.
3600 * config/nds32/nds32-predicates.c: New predicates for hard float.
3601 * config/nds32/nds32-protos.h: Declare functions for hard float.
3602 * config/nds32/nds32.c: Implementation for hard float configuration.
3603 * config/nds32/nds32.h: Definitions for hard float configuration.
3604 * config/nds32/nds32.md: Include hard float machine description and
3605 modify patterns for hard float configuration.
3606 * config/nds32/nds32.opt: New options for hard float configuration.
3607 * config/nds32/predicates.md: New predicates for hard float
3610 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3612 * common/config/nds32/nds32-common.c
3613 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3615 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3618 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3619 CONSTRUCTOR_ELT (ctor, ...)->value.
3621 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3624 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3626 2018-04-05 Tom de Vries <tom@codesourcery.com>
3629 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3632 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3633 Kito Cheng <kito.cheng@gmail.com>
3635 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3636 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3637 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3638 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3640 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3641 Kito Cheng <kito.cheng@gmail.com>
3643 * config/nds32/constraints.md (Ufe): New memory constraint.
3644 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3645 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3646 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3648 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3649 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3651 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3653 * config/nds32/nds32.md: Use optimize_size in the condition for
3654 alu-shift instructions.
3656 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3658 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3660 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3662 * config/nds32/nds32.md (negsi2): Refine pattern.
3664 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3665 Chung-Ju Wu <jasonwucj@gmail.com>
3667 * config/nds32/iterators.md (shift_rotate): New code iterator.
3668 (shift): New code attribute.
3669 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3670 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3671 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3672 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3673 bit-wise operations.
3674 (andsi3, *andsi3): Ditto.
3675 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3676 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3677 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3678 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3679 nds32_ior_operand, nds32_xor_operand): New predicates.
3681 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3683 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3684 (addsi3, subsi3): ... this.
3686 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3688 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3690 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3692 * config/nds32/nds32.md: Adjust indention.
3694 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3696 * config/nds32/nds32.md (feature): New attribute.
3698 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3700 * config/nds32/nds32.md (subtype): New attribute.
3702 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3705 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3706 expansion to perform a bitwise AND of the argument followed by a
3707 boolean negation of the result.
3709 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3711 PR rtl-optimization/84878
3712 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3713 the basic block. Assert the use reference is not artificial and that
3714 it has an associated insn.
3716 2018-04-04 Michael Matz <matz@suse.de>
3718 * builtins.c (compute_objsize): Pass correct operand
3719 to array_at_struct_end_p.
3721 2018-04-04 Richard Biener <rguenther@suse.de>
3724 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3725 from contexts for DINFO_LEVEL_TERSE and below.
3727 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3729 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3730 resiter_operand condition.
3731 * config/nds32/nds32.md (*move<mode>): Ditto.
3733 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3734 Monk Chiang <sh.chiang04@gmail.com>
3736 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3738 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3740 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3742 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3743 Kito Cheng <kito.cheng@gmail.com>
3745 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3746 nds32_cond_code_str, output_cond_branch,
3747 output_cond_branch_compare_zero, nds32_expand_cbranch,
3748 nds32_expand_cstore, nds32_expand_movcc,
3749 nds32_output_cbranchsi4_equality_zero,
3750 nds32_output_cbranchsi4_equality_reg,
3751 nds32_output_cbranchsi4_equality_reg_or_const_int,
3752 nds32_output_cbranchsi4_greater_less_zero: New functions.
3753 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3754 nds32_expand_cstore, nds32_expand_movcc,
3755 nds32_output_cbranchsi4_equality_zero,
3756 nds32_output_cbranchsi4_equality_reg,
3757 nds32_output_cbranchsi4_equality_reg_or_const_int,
3758 nds32_output_cbranchsi4_greater_less_zero): Declare.
3759 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3760 nds32_rimm11s_operand): New predicates.
3761 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3762 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3765 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3767 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3768 * config/nds32/nds32.md: Ditto.
3769 * config/nds32/pipelines.md: Ditto.
3771 2018-04-04 Richard Biener <rguenther@suse.de>
3773 PR tree-optimization/85168
3774 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3775 propagating abnormals.
3777 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3779 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3781 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3782 Kito Cheng <kito.cheng@gmail.com>
3784 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3785 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3786 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3787 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3788 * config/nds32/nds32.md (sibcall_internal): New.
3789 (sibcall_register): Remove.
3790 (sibcall_immediate): Remove.
3791 (sibcall_value_internal): New.
3792 (sibcall_value_register): Remove.
3793 (sibcall_value_immediate): Remove.
3794 * config/nds32/predicates.md (nds32_general_register_operand): New.
3795 (nds32_call_address_operand): New.
3797 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3799 PR rtl-optimization/85167
3800 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3801 bb_defs if *split_p, instead preinitialize it to NULL.
3803 PR tree-optimization/85156
3804 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3805 evaluating the argument multiple times.
3807 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3809 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3811 (_mm_cvtpd_ps): Likewise.
3812 (_mm_cvttpd_epi32): Likewise.
3813 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3814 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3815 vector, pixel, and bool following altivec.h include.
3817 2018-04-03 Martin Sebor <msebor@redhat.com>
3819 * doc/extend.texi (Common Function Attributes): Clarify.
3820 (const attribute): Likewise.
3821 (pure attribute): Likewise.
3823 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3826 * config/i386/i386.c (ix86_expand_vector_set): Use
3827 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3829 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3831 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3832 instructions when changing rounding bits to preserve precision bits
3833 in the x87 control word.
3835 2018-04-03 Martin Liska <mliska@suse.cz>
3837 PR tree-optimization/82491
3838 * rtl.h (strip_offset_and_add): Replace += suboffset with
3839 poly_uint64 () + suboffset.
3841 2018-03-29 Martin Liska <mliska@suse.cz>
3842 Martin Jambor <mjambor@suse.cz>
3845 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3846 param_type is not an integral or pointer type.
3848 2018-04-03 Richard Biener <rguenther@suse.de>
3850 * sese.h (recompute_all_dominators): Remove.
3852 2018-04-02 Martin Sebor <msebor@redhat.com>
3854 * doc/invoke.texi (-Wrestrict): Fix typos.
3856 2018-04-02 Jim Wilson <jimw@sifive.com>
3858 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3859 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3860 (<optab>di3, <optab>si3_extend): Likewise.
3861 (<optab>si3_mask, <optab>si3_mask_1): New.
3862 (<optab>di3_mask, <optab>di3_mask_1): New.
3863 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3864 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3865 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3867 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3869 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3872 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3874 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3875 (nds32_canonicalize_comparison): New function.
3877 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3878 Kito Cheng <kito.cheng@gmail.com>
3879 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3881 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3882 * config/nds32/constants.md (unspec_volatile_element): Add
3883 UNSPEC_VOLATILE_RELAX_GROUP.
3884 * config/nds32/nds32-relax-opt.c: New file.
3885 * config/nds32/nds32-predicates.c
3886 (nds32_symbol_load_store_p): New function.
3887 * config/nds32/nds32-protos.h
3888 (nds32_symbol_load_store_p): Declare function.
3889 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3890 * config/nds32/nds32.c
3891 (nds32_register_pass): New function to register pass.
3892 (nds32_register_passes): New function to register passes.
3893 * config/nds32/nds32.md (relax_group): New pattern.
3894 * config/nds32/nds32.opt (mrelax-hint): New option.
3895 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3897 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3899 * config/nds32/t-nds32: Modify files dependency.
3901 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3903 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3904 (PROFILE_HOOK): Define its implementation.
3906 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3908 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3909 type and 32-bit size.
3911 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3914 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3915 (V_128_256): New mode iterator.
3916 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3917 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3918 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3920 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3921 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3923 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3926 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3927 NaN inputs correctly.
3929 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3932 * config/rs6000/vsx.md (??r): New mode attribute.
3933 (*vsx_mov<mode>_64bit): Use it.
3934 (*vsx_mov<mode>_32bit): Likewise.
3936 2018-03-30 Martin Sebor <msebor@redhat.com>
3938 PR tree-optimization/84818
3939 * builtins.c (check_access): Use warning_n.
3941 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3944 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3946 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3949 2018-03-30 Julia Koval <julia.koval@intel.com>
3952 * x86-tune.def (movx, partial_reg_dependency): Enable for
3955 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3958 * lra-constraints.c (process_alt_operands): Move setting
3959 this_alternative_matches below.
3961 2018-03-29 Martin Liska <mliska@suse.cz>
3964 * doc/invoke.texi: Document how LTO works with debug info.
3965 Describe auto-load support of binutils. Mention 'x86-64'
3966 as valid option value of -march option.
3968 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3970 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3973 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3974 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3975 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3978 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3981 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3982 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3983 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3984 for RS6000_BTM_POWERPC64.
3985 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3986 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3987 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3992 2018-03-28 Carl Love <cel@us.ibm.com>
3995 2017-09-27 Carl Love <cel@us.ibm.com>
3997 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3998 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3999 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4002 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4004 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4005 instead of __vector bool.
4006 (_mm_max_pu8): Likewise.
4007 (_mm_min_pi16): Likewise.
4009 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4012 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4013 (DIVWEUO): Likewise.
4015 (DIVDEUO): Likewise.
4016 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4017 DIVWEUO and DIVDEUO.
4018 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4019 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4020 (div_extend): Likewise.
4021 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4023 (__builtin_divweuo): Likewise.
4024 (__builtin_divdeo): Likewise.
4025 (__builtin_divdeuo): Likewise.
4027 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4030 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4031 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4033 PR tree-optimization/82004
4034 * gimple-match-head.c (optimize_pow_to_exp): New function.
4035 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4036 Don't fold to exp if optimize_pow_to_exp is false.
4038 2018-03-28 Martin Liska <mliska@suse.cz>
4041 * calls.c (initialize_argument_information): Fix trailing space.
4042 * common.opt: Fix typo and provide better explanation for
4043 -fsanitize-coverage option.
4044 * config/i386/i386.opt: Fix typo.
4046 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4047 Martin Liska <mliska@suse.cz>
4050 * gimplify.c (asan_poison_variable): Don't do the check for
4051 gimplify_omp_ctxp here.
4052 (gimplify_decl_expr): Do it here.
4053 (gimplify_target_expr): Likewise.
4055 2018-03-28 Martin Liska <mliska@suse.cz>
4058 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4059 chkp_type_bounds_count if MPX is not enabled.
4061 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4063 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4065 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4068 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4069 function to create the function decl for complex long double
4070 multiply and divide for -mabi=ieeelongdouble.
4071 (init_float128_ieee): Call it.
4073 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4076 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4077 -fcf-protection=branch -mibt.
4078 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4080 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4083 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4085 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4088 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4089 extern array declarations.
4091 2018-03-27 Richard Biener <rguenther@suse.de>
4094 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4095 explicit single_use checks.
4097 2018-03-27 Richard Biener <rguenther@suse.de>
4099 PR tree-optimization/85082
4100 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4103 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4105 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4106 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4107 Turn on fasynchronous-unwind-tables and funwind-tables.
4109 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4112 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4113 (*bmi_blsr_<mode>_ccz): Ditto.
4115 2018-03-26 Tom de Vries <tom@codesourcery.com>
4117 PR tree-optimization/85063
4118 * omp-general.c (offloading_function_p): New function. Factor out
4120 * omp-offload.c (pass_omp_target_link::gate): ... here.
4121 * omp-general.h (offloading_function_p): Declare.
4122 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4123 with attribute omp declare target for offloading functions.
4125 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4127 PR tree-optimization/84005
4128 * tree-data-ref.h (get_base_for_alignment): Declare.
4129 * tree-data-ref.c (get_base_for_alignment_1): New function.
4130 (get_base_for_alignment): Likewise.
4131 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4132 get_base_for_alignment to find a suitable base object, instead
4133 of always using drb->base_address.
4135 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4138 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4139 known size by default.
4141 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4144 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4145 and non BLKmode operands.
4147 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4150 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4151 Clean up attributes.
4153 2018-03-23 Richard Biener <rguenther@suse.de>
4156 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4157 we are going to emit early debug for LTO.
4159 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4162 * function.c (match_asm_constraints_1): Don't optimize if input
4163 doesn't satisfy general_operand predicate for output's mode.
4166 * alias.c (write_dependence_p): Don't require for x_canonicalized
4167 non-VOIDmode if x has VOIDmode.
4170 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4171 just don't try to optimize it rather than assert it never happens.
4173 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4175 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4176 macro expansions for definition of ST_INTERNAL_<mode> and
4177 LD_INTERNAL_<mode> builtins.
4178 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4180 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4182 (altivec_expand_st_builtin): Likewise.
4183 (altivec_expand_builtin): Remove calls to deleted functions.
4184 (rs6000_address_for_altivec): Delete this function.
4185 * config/rs6000/vector.md: Remove expands for
4186 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4188 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4191 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4192 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4193 re-computing once computed.
4194 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4195 (arm_init_machine_status): Initialize
4196 machine->static_chain_stack_bytes.
4198 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4201 * doc/extend.texi: Add four new prototypes for vec_ld.
4202 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4203 definitions for more logical presentation.
4204 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4205 entries for V1TI variants of __builtin_altivec_ld builtin.
4206 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4207 handling of V1TI variant of LVX icode pattern.
4208 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4209 (rs6000_gimple_fold_builtin): Likewise.
4210 (altivec_init_builtins): Add code to define
4211 __builtin_altivec_lvx_v1ti function.
4213 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4216 * function.c (match_asm_constraints_1): Don't do the optimization
4217 if input isn't a REG, SUBREG, MEM or constant.
4219 2018-03-22 Tom de Vries <tom@codesourcery.com>
4221 PR tree-optimization/84956
4222 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4223 bb_has_abnormal_pred.
4225 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4228 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4229 DECL_INITIAL (decl) to decl at the end.
4230 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4233 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4235 * doc/extend.texi (__builtin_tgmath): Document when complex
4236 integer types are treated as _Complex _Float64.
4238 2018-03-21 Tom de Vries <tom@codesourcery.com>
4240 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4242 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4244 PR tree-optimization/84960
4245 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4246 if it is ENTRY block, move them into single succ of ENTRY in that case.
4248 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4250 PR tree-optimization/84811
4251 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4252 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4253 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4254 (known_subrange_p): Update accordingly. Cast each value involved
4255 in the size comparison, rather than casting the result of the
4258 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4260 PR tree-optimization/84982
4261 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4262 by flipping the least significant bit rather than all bits from
4263 bitpos to bitpos + bitsize - 1.
4265 2018-03-21 Nathan Sidwell <nathan@acm.org>
4267 * doc/extend.texi (Deprecated Features): Remove mention of
4268 long-deleted deprecations.
4270 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4273 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4274 * configure: Regenerate.
4276 2018-03-21 Tom de Vries <tom@codesourcery.com>
4278 PR tree-optimization/83126
4279 * tree-parloops.c (num_phis): New function.
4280 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4282 2018-03-21 Nathan Sidwell <nathan@acm.org>
4284 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4285 mention anon-struct/union members and trailing attributes.
4287 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4289 PR tree-optimization/84969
4290 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4291 builtin memset partitions if they set different rhs values.
4293 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4295 PR rtl-optimization/84989
4296 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4297 VEC_DUPLICATE with scalar result mode.
4299 2018-03-21 Martin Liska <mliska@suse.cz>
4302 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4303 not intended return statement.
4305 2018-03-21 Martin Liska <mliska@suse.cz>
4308 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4309 (chkp_find_bound_slots_1): Limit number of iterations.
4311 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4314 * Minor grammar fixes for x86 options.
4316 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4319 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4320 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4322 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4325 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4326 (altivec_lvx_<mode>_1op): Likewise.
4327 (altivec_stvx_<mode>_2op): Likewise.
4328 (altivec_stvx_<mode>_1op): Likewise.
4329 (altivec_lvx_<VM2:mode>): New define_expand.
4330 (altivec_stvx_<VM2:mode>): Likewise.
4331 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4332 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4333 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4334 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4335 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4336 (rs6000_gen_lvx): Likewise.
4337 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4338 (altivec_expand_stv_builtin): Likewise.
4339 (altivec_expand_builtin): Likewise.
4340 * config/rs6000/vector.md: Likewise.
4342 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4345 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4348 2018-03-20 Richard Biener <rguenther@suse.de>
4351 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4352 sign-conversions as zero, fall back to standard scalar_stmt
4355 2018-03-20 Martin Liska <mliska@suse.cz>
4358 * predict.c (rebuild_frequencies): Handle case when we have
4359 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4361 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4364 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4365 flag_section_anchors.
4366 * varasm.c (use_blocks_for_decl_p): Remove hack for
4367 dw2_force_const_mem.
4370 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4372 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4373 be created, use lowpart_subreg of operands[0] rather than operands[0]
4375 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4376 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4377 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4378 and n constraint instead of aarch64_shift_imm_di and Usd.
4379 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4380 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4382 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4385 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4386 to favor GPR over NEON registers.
4387 (<shift>di3_neon): Likewise.
4389 2018-03-20 Tom de Vries <tom@codesourcery.com>
4392 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4393 (nvptx_process_pars): Emit bar.sync asap and alap.
4395 2018-03-20 Tom de Vries <tom@codesourcery.com>
4398 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4399 seen_label if seen_label is already set.
4401 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4404 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4405 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4406 Use 1U instead of 1. Formatting fixes.
4409 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4410 instead of TREE_TYPE (s1) for the return value.
4412 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4414 PR tree-optimization/84946
4415 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4416 bitsize + bitsize in poly_uint64 rather than poly_int64.
4419 * dwarf2asm.c: Include fold-const.c.
4420 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4421 of decl rather than decl itself.
4423 PR rtl-optimization/84643
4424 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4426 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4429 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4430 calling assemble_variable.
4432 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4435 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4436 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4438 2018-03-19 Jim Wilson <jimw@sifive.com>
4441 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4442 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4443 (riscv_first_stack_step): Likewise.
4444 (riscv_option_override): Use STACK_BOUNDARY instead of
4446 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4448 (BIGGEST_ALIGNMENT): Set to 128.
4449 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4450 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4453 2018-03-19 Richard Biener <rguenther@suse.de>
4455 PR tree-optimization/84933
4456 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4457 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4459 2018-03-19 Richard Biener <rguenther@suse.de>
4461 PR tree-optimization/84859
4462 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4463 (cond_if_else_store_replacement): Perform sinking operation on
4464 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4465 Generalize what a BB with a single eligible store is.
4467 2018-03-19 Richard Biener <rguenther@suse.de>
4469 PR tree-optimization/84929
4470 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4471 chrec_is_positive against non-chrec arg.
4473 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4476 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4478 2018-03-18 Martin Liska <mliska@suse.cz>
4480 PR rtl-optimization/84635
4481 * regrename.c (build_def_use): Use matches_mode only when
4484 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4486 PR tree-optimization/84913
4487 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4488 vectorize chains of COND_EXPRs.
4490 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4492 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4494 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4496 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4498 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4500 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4502 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4503 Kito Cheng <kito.cheng@gmail.com>
4505 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4506 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4507 (nds32_adjust_reg_alloc_order): New function.
4508 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4510 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4512 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4513 nds32_print_operand, nds32_print_operand_address): Use
4514 HOST_WIDE_INT_PRINT_DEC instead.
4516 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4518 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4520 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4523 * config/i386/i386.c (initial_ix86_tune_features,
4524 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4526 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4527 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4528 rather than 1u << ix86_tune. Formatting fix.
4529 (ix86_option_override_internal): Change ix86_arch_mask from
4530 unsigned int to unsigned HOST_WIDE_INT, initialize to
4531 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4532 (ix86_function_specific_restore): Likewise.
4534 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4537 * postreload.c (reload_combine_recognize_pattern): Perform
4538 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4539 truncate_int_for_mode the result for the destination's mode.
4542 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4543 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4545 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4548 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4551 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4552 regno_allocno_class_array and sorted_pseudos.
4553 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4554 insns where regno is used.
4556 2018-03-16 Martin Liska <mliska@suse.cz>
4559 * multiple_target.c (create_dispatcher_calls): Redirect
4560 reference in the symbol table.
4562 2018-03-16 Martin Liska <mliska@suse.cz>
4565 * multiple_target.c (create_dispatcher_calls): Redirect also
4568 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4572 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4573 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4574 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4576 2018-03-16 Julia Koval <julia.koval@intel.com>
4578 * doc/invoke.texi (Skylake Server): Add CLWB.
4579 Cannonlake): Remove CLWB.
4581 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4583 PR tree-optimization/84841
4584 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4586 (FLOAT_ONE_CONST_TYPE): Define.
4587 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4588 (sort_by_operand_rank): Put entries with higher constant_type last
4589 rather than first to match comments.
4591 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4593 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4596 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4599 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4600 check and use address of target always.
4602 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4605 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4606 (indirect_thunk_bnd_needed): Likewise.
4607 (indirect_thunks_used): Likewise.
4608 (indirect_thunks_bnd_used): Likewise.
4609 (indirect_return_needed): New.
4610 (indirect_return_bnd_needed): Likewise.
4611 (output_indirect_thunk_function): Add a bool argument for
4613 (output_indirect_thunk_function): Don't generate alias for
4614 function return thunk.
4615 (ix86_code_end): Call output_indirect_thunk_function to generate
4616 function return thunks.
4617 (ix86_output_function_return): Set indirect_return_bnd_needed
4618 and indirect_return_needed instead of indirect_thunk_bnd_needed
4619 and indirect_thunk_needed.
4621 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4623 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4624 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4625 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4627 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4628 Paul Hua <paul.hua.gm@gmail.com>
4631 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4633 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4635 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4636 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4639 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4642 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4643 instead of GET_MODE_SIZE when comparing Units.
4645 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4648 * varasm.c (hash_section): Return an unchangeble hash value
4649 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4650 Return !aarch64_can_use_per_function_literal_pools_p ().
4652 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4655 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4656 rather than address of cmode as last argument to prepare_cmp_insn.
4658 2018-03-15 Julia Koval <julia.koval@intel.com>
4660 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4661 F_AVX512VNNI, F_AVX512BITALG): New.
4663 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4666 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4667 insn for floating-point loads and stores.
4669 2018-03-14 Carl Love <cel@us.ibm.com>
4671 * config/rs6000/rs6000-c.c: Add macro definitions for
4672 ALTIVEC_BUILTIN_VEC_PERMXOR.
4673 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4674 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4675 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4676 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4678 * config/doc/extend.texi: Add prototypes for vec_permxor.
4680 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4683 * diagnostic-show-locus.c (class layout_point): Convert m_line
4684 from int to linenum_type.
4685 (line_span::comparator): Use linenum "compare" function when
4686 comparing line numbers.
4687 (test_line_span): New function.
4688 (layout_range::contains_point): Convert param "row" from int to
4690 (layout_range::intersects_line_p): Likewise.
4691 (layout::will_show_line_p): Likewise.
4692 (layout::print_source_line): Likewise.
4693 (layout::should_print_annotation_line_p): Likewise.
4694 (layout::print_annotation_line): Likewise.
4695 (layout::print_leading_fixits): Likewise.
4696 (layout::annotation_line_showed_range_p): Likewise.
4697 (struct line_corrections): Likewise for field m_row.
4698 (line_corrections::line_corrections): Likewise for param "row".
4699 (layout::print_trailing_fixits): Likewise.
4700 (layout::get_state_at_point): Likewise.
4701 (layout::get_x_bound_for_row): Likewise.
4702 (layout::print_line): Likewise.
4703 (diagnostic_show_locus): Likewise for locals "last_line" and
4705 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4706 * input.c (selftest::test_linenum_comparisons): New function.
4707 (selftest::input_c_tests): Call it.
4708 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4709 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4710 * selftest.h (ASSERT_GT): New macro.
4711 (ASSERT_GT_AT): New macro.
4712 (ASSERT_LT): New macro.
4713 (ASSERT_LT_AT): New macro.
4715 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4717 PR rtl-optimization/84780
4718 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4720 2018-03-14 Martin Liska <mliska@suse.cz>
4722 * tree.c (record_node_allocation_statistics): Use
4723 get_stats_node_kind.
4724 (get_stats_node_kind): New function extracted from
4725 record_node_allocation_statistics.
4726 (free_node): Use get_stats_node_kind.
4728 2018-03-14 Richard Biener <rguenther@suse.de>
4730 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4731 that the value-set of ANTIC_IN doesn't grow.
4734 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4736 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4737 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4739 2018-03-14 Julia Koval <julia.koval@intel.com>
4741 * config.gcc (icelake-client, icelake-server): New.
4743 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4744 (initial_ix86_arch_features): Ditto.
4745 (PTA_SKYLAKE): Add SGX.
4746 (PTA_ICELAKE): Remove.
4747 (PTA_ICELAKE_CLIENT): New.
4748 (PTA_ICELAKE_SERVER): New.
4749 (ix86_option_override_internal): Split up icelake on icelake client and
4751 (get_builtin_code_for_version): Ditto.
4752 (fold_builtin_cpu): Ditto.
4753 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4754 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4755 * config/i386/i386.h (processor_type): Ditto.
4756 * doc/invoke.texi: Ditto.
4758 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4761 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4762 INTEGER_CST offset, add it together with bitpos / 8 and
4763 sign extend based on POINTER_SIZE.
4767 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4770 * config/i386/constraints.md (Yc): New register constraint.
4771 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4772 Use Yc constraint for alternative 2 of operand 0. Remove
4773 preferred_for_speed attribute.
4775 2018-03-14 Richard Biener <rguenther@suse.de>
4777 PR tree-optimization/84830
4778 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4779 with the old one to avoid oscillations.
4781 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4784 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4786 (assign_by_spills): Return a flag of reload assignment failure.
4787 Do not process the reload assignment failures. Do not spill other
4788 reload pseudos if they has the same reg class. Update n if
4790 (lra_assign): Add a return arg. Set up from the result of
4791 assign_by_spills call.
4792 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4793 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4794 usage_insns if it is not NULL.
4795 (spill_hard_reg_in_range): New function.
4796 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4797 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4798 function prototypes.
4799 (lra_assign): Change prototype.
4800 * lra.c (lra): Add code to deal with fails by splitting hard reg
4803 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4805 * config/riscv/riscv.opt (mrelax): New option.
4806 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4807 "norelax" when riscv_mrelax is disabled.
4808 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4810 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4813 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4814 reassociation for int modes.
4816 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4818 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4819 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4821 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4822 * config/aarch64/aarch64-sve.md
4823 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4824 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4825 (*extend<mode><Vwide>2): Rename to...
4826 (aarch64_sve_extend<mode><Vwide>2): ...this.
4827 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4828 renaming the old pattern to...
4829 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4831 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4832 define_expand, renaming the old pattern to...
4833 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4834 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4835 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4836 account when deciding which SVE instruction the optab should use.
4837 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4839 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4841 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4842 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4843 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4844 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4845 (tlsdesc_small_<mode>): Turn a define_expand and use
4846 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4847 (tlsdesc_small_advsimd_<mode>): ...this.
4848 (tlsdesc_small_sve_<mode>): New pattern.
4850 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4852 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4853 (UNSPEC_UMUL_HIGHPART): New constants.
4854 (MUL_HIGHPART): New int iteraor.
4855 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4856 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4858 (*<su>mul<mode>3_highpart): New define_insn.
4860 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4863 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4866 2018-03-13 Martin Liska <mliska@suse.cz>
4869 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4871 (sem_item_optimizer::~sem_item_optimizer): Release it.
4872 (sem_item_optimizer::merge_classes): Register variable aliases.
4873 (sem_item_optimizer::fixup_pt_set): New function.
4874 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4875 * ipa-icf.h: Declare new variables and functions.
4877 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4880 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4881 integer_pow2p@2 and test integer_pow2p in condition.
4882 (A < 0 ? C : 0): Similarly for @1.
4885 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4886 characters starting at p contain '\0' character, don't look beyond
4890 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4891 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4894 * reg-stack.c (change_stack): Change update_end var from int to
4895 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4896 also call set_block_for_insn on the newly added insns and rescan.
4899 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4900 on the last operand.
4903 * tree.c (stabilize_reference_1): Return save_expr (e) for
4904 STATEMENT_LIST even if it doesn't have side-effects.
4906 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4908 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4910 2018-03-12 Renlin Li <renlin.li@arm.com>
4912 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4913 aarch64_output_scalar_simd_mov_immediate.
4915 2018-03-12 Martin Sebor <msebor@redhat.com>
4917 PR tree-optimization/83456
4918 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4919 for perfectly overlapping calls to memcpy.
4920 (gimple_fold_builtin_memory_chk): Same.
4921 (gimple_fold_builtin_strcpy): Handle no-warning.
4922 (gimple_fold_builtin_stxcpy_chk): Same.
4923 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4925 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4927 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4928 parameter. Use it for SFmode.
4929 (rs6000_function_arg_advance_1): Adjust.
4930 (rs6000_function_arg): Adjust.
4931 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4933 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4935 PR rtl-optimization/84169
4936 PR rtl-optimization/84780
4937 * combine.c (can_combine_p): Check for a 2-insn combination whether
4938 the destination register is used between the two insns, too.
4940 2018-03-12 Richard Biener <rguenther@suse.de>
4942 PR tree-optimization/84803
4943 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4944 for refs DR analysis didn't process.
4946 2018-03-12 Richard Biener <rguenther@suse.de>
4948 PR tree-optimization/84777
4949 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4950 force-vectorize loops ignore whether we are optimizing for size.
4952 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4954 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4955 (TARGET_MD_ASM_ADJUST): Define.
4957 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4958 Kito Cheng <kito.cheng@gmail.com>
4959 Chung-Ju Wu <jasonwucj@gmail.com>
4961 * config/nds32/nds32.c (nds32_compute_stack_frame,
4962 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4963 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4964 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4965 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4966 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4967 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4968 * config/nds32/nds32.md (prologue, epilogue): Use macro
4969 NDS32_V3PUSH_AVAILABLE_P to do checking.
4971 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4974 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4975 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4976 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4977 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4978 addition of most attributes on !orig_type_die or the attribute not
4979 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4981 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4982 Chung-Ju Wu <jasonwucj@gmail.com>
4984 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4986 * config/nds32/nds32.opt (mvh): New option.
4988 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4989 Chung-Ju Wu <jasonwucj@gmail.com>
4991 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4993 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4994 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4997 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4998 Chung-Ju Wu <jasonwucj@gmail.com>
5000 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5002 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5003 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5005 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5006 Kito Cheng <kito.cheng@gmail.com>
5007 Chung-Ju Wu <jasonwucj@gmail.com>
5009 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5010 UNSPEC_FFMISM and UNSPEC_FLMISM.
5011 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5012 for ffb, ffmism and flmism.
5013 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5014 (unspec_ffmism): Ditto.
5015 (unspec_flmism): Ditto.
5016 (nds32_expand_builtin_impl): Check if string extension is available.
5017 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5018 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5020 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5023 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5026 * lra-assigns.c (assign_by_spills): Return a flag of reload
5027 assignment failure. Do not process the reload assignment
5028 failures. Do not spill other reload pseudos if they has the same
5030 (lra_assign): Add a return arg. Set up from the result of
5031 assign_by_spills call.
5032 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5033 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5034 usage_insns if it is not NULL.
5035 (spill_hard_reg_in_range): New function.
5036 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5037 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5038 function prototypes.
5039 (lra_assign): Change prototype.
5040 * lra.c (lra): Add code to deal with fails by splitting hard reg
5043 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5046 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5048 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5051 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5052 (dw_val_node): Add val_symbolic_view.
5053 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5054 (symview_upper_bound): New.
5055 (new_line_info_table): Initialize symviews_since_reset.
5056 (dwarf2out_source_line): Count symviews_since_reset and set
5057 symview_upper_bound.
5058 (dw_val_equal_p): Handle symview.
5059 (add_AT_symview): New.
5060 (print_dw_val): Handle symview.
5061 (attr_checksum, attr_checksum_ordered): Likewise.
5062 (same_dw_val_p, size_of_die): Likewise.
5063 (value_format, output_die): Likewise.
5064 (add_high_low_attributes): Use add_AT_symview for entry_view.
5065 (dwarf2out_finish): Reset symview_upper_bound, clear
5068 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5071 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5072 Add strict argument and use it.
5073 (rs6000_split_multireg_move): Update for new strict argument.
5074 (mem_operand_gpr): Disallow all non-offsettable addresses.
5075 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5077 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5080 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5081 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5082 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5085 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5086 decl, use remap_type if we want to use the type.
5088 2018-03-09 Martin Sebor <msebor@redhat.com>
5090 PR tree-optimization/84526
5091 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5093 (builtin_access::generic_overlap): Be prepared to handle non-array
5096 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5098 PR rtl-optimization/84682
5099 * lra-constraints.c (process_address_1): Check is_address flag
5100 for address constraints.
5101 (process_alt_operands): Likewise.
5102 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5103 preprocess_constraints.
5104 * recog.h (preprocess_constraints): Add oploc parameter.
5106 * recog.c (preprocess_constraints): Test address_operand for
5107 CT_ADDRESS constraints.
5109 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5112 * lra-assigns.c (assign_by_spills): Return a flag of reload
5113 assignment failure. Do not process the reload assignment
5114 failures. Do not spill other reload pseudos if they has the same
5116 (lra_assign): Add a return arg. Set up from the result of
5117 assign_by_spills call.
5118 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5119 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5120 usage_insns if it is not NULL.
5121 (spill_hard_reg_in_range): New function.
5122 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5123 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5124 function prototypes.
5125 (lra_assign): Change prototype.
5126 * lra.c (lra): Add code to deal with fails by splitting hard reg
5129 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5132 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5133 Accept complain bool parameter. Only emit errors if it is true.
5134 (arm_parse_cpu_option_name): Likewise.
5135 (arm_target_thumb_only): Adjust callers of the above.
5136 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5137 prototype to take a default true bool parameter.
5138 (arm_parse_arch_option_name): Likewise.
5140 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5141 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5145 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5146 * configure: Regenerate.
5147 * configure.ac ("linker --version-script option"): New.
5148 ("linker soname option"): New.
5150 2018-03-09 Richard Biener <rguenther@suse.de>
5152 PR tree-optimization/84775
5153 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5154 immediate uses of predicate stmts and mark them modified.
5157 PR tree-optimization/84178
5158 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5160 (version_loop_for_if_conversion): Delay update_ssa call.
5161 (tree_if_conversion): Delay update_ssa until after predicate
5164 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5167 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5168 when the function accesses prior frames.
5170 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5173 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5174 gen_llsym, otherwise call maybe_gen_llsym.
5177 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5178 has ',' character inside of it.
5180 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5183 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5184 as clobbering CC_REGNUM.
5186 2018-03-08 Richard Biener <rguenther@suse.de>
5189 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5190 (follow_copies_to_constant): Do not follow SSA names registered
5193 2018-03-08 Richard Biener <rguenther@suse.de>
5195 PR tree-optimization/84178
5196 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5198 (version_loop_for_if_conversion): Delay update_ssa call.
5199 (tree_if_conversion): Delay update_ssa until after predicate
5202 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5204 PR tree-optimization/84178
5205 * tree-if-conv.c (release_bb_predicate): Remove the
5206 the assertion that the stmts have NULL use_ops.
5207 Discard the statements, asserting that they haven't
5208 yet been added to a BB.
5210 2018-03-08 Richard Biener <rguenther@suse.de>
5212 PR tree-optimization/84746
5213 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5214 (phi_translate): Pass in destination ANTIC_OUT set.
5215 (phi_translate_1): Likewise. For a simplified result lookup
5216 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5217 (phi_translate_set): Adjust.
5218 (do_pre_regular_insertion): Likewise.
5219 (do_pre_partial_partial_insertion): Likewise.
5221 2018-03-08 Martin Liska <mliska@suse.cz>
5223 PR gcov-profile/84735
5224 * doc/gcov.texi: Document usage of profile files.
5225 * gcov-io.h: Document changes in the format.
5227 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5231 * dwarf2out.c (struct dw_line_info_table): Update comments for
5233 (FORCE_RESET_NEXT_VIEW): New.
5234 (FORCE_RESETTING_VIEW_P): New.
5235 (RESETTING_VIEW_P): Check for -1 too.
5236 (ZERO_VIEW_P): Likewise.
5237 (new_line_info_table): Force-reset next view.
5238 (dwarf2out_begin_function): Likewise.
5239 (dwarf2out_source_line): Simplify zero_view_p initialization.
5240 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5241 view directly. Omit view when omitting .loc at line 0.
5243 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5245 PR tree-optimization/84740
5246 * tree-switch-conversion.c (process_switch): Call build_constructors
5247 only if info.phi_count is non-zero.
5249 PR tree-optimization/84739
5250 * tree-tailcall.c (find_tail_calls): Check call arguments against
5251 DECL_ARGUMENTS (current_function_decl) rather than
5252 DECL_ARGUMENTS (func) when checking for tail recursion.
5254 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5256 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5257 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5258 Volker Reichelt's entry and add entries for people that perform
5259 GCC fuzzy testing and report numerous bugs.
5261 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5264 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5265 readonly data in sdata, if that is disabled.
5266 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5267 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5268 -mreadonly-in-sdata option.
5270 2018-03-07 Martin Sebor <msebor@redhat.com>
5272 PR tree-optimization/84468
5273 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5274 basic block when looking for nul assignment.
5276 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5279 * except.h (output_function_exception_table): Adjust prototype.
5280 * except.c (output_function_exception_table): Remove FNNAME parameter
5281 and add SECTION parameter. Ouput one part of the table at a time.
5282 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5283 the first part of the exception table and emit unwind directives.
5284 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5285 (i386_pe_seh_cold_init): Likewise.
5286 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5287 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5288 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5289 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5290 * config/i386/winnt.c: Include except.h.
5291 (struct seh_frame_state): Add reg_offset, after_prologue and
5292 in_cold_section fields.
5293 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5294 (i386_pe_seh_cold_init): New function.
5295 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5296 to seh->in_cold_section.
5297 (seh_emit_push): Record the offset of the push.
5298 (seh_emit_save): Record the offet of the save.
5299 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5300 Test seh->after_prologue to disregard the epilogue.
5301 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5302 (i386_pe_end_cold_function): New function.
5304 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5307 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5308 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5311 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5313 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5314 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5317 * multiple_target.c: Include tree-inline.h and intl.h.
5318 (expand_target_clones): Diagnose and fail if node->definition and
5319 !tree_versionable_function_p (node->decl).
5321 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5323 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5325 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5326 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5327 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5329 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5332 * combine.c (try_combine): Use reg_or_subregno instead of handling
5333 just paradoxical SUBREGs and REGs.
5335 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5337 * config/arc/arc.c (arc_finalize_pic): Remove function.
5338 (arc_must_save_register): We use single base PIC register, remove
5339 checks to save/restore the PIC register.
5340 (arc_expand_prologue): Likewise.
5341 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5343 (arc_verify_short): Likewise.
5344 (arc_attr_type): Likewise.
5345 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5346 (walk_stores): Likewise.
5347 (arc_address_cost): Make it static.
5348 (arc_verify_short): Likewise.
5349 (branch_dest): Likewise.
5350 (arc_attr_type): Likewise.
5351 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5352 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5353 (arc_final_prescan_insn): Remove inserting the nops due to
5354 hardware hazards. It is done in reorg step.
5355 (insn_length_variant_t): Remove.
5356 (insn_length_parameters_t): Likewise.
5357 (arc_insn_length_parameters): Likewise.
5358 (arc_get_insn_variants): Likewise.
5359 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5361 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5364 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5367 PR tree-optimization/84687
5368 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5370 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5372 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5374 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5375 Rename to ppc_speculation_barrier.
5376 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5377 __builtin_ppc_speculation_barrier.
5379 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5382 * combine.c (combine_simplify_rtx): Don't try to simplify if
5383 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5386 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5388 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5389 to 32 bytes when compiling for POWER9.
5391 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5394 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5395 regparm >= 3 with no arg reg available also for calls with
5396 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5399 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5401 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5403 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5406 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5408 2018-03-05 Richard Biener <rguenther@suse.de>
5410 PR tree-optimization/84486
5411 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5412 When inserting a __builtin_assume_aligned call set the LHS
5413 SSA name alignment info accordingly.
5415 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5417 PR tree-optimization/84114
5418 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5419 Avoid reassociation of FLOAT_MODE addition.
5421 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5423 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5424 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5425 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5426 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5427 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5428 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5429 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5431 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5432 __builtin_ia32_wbinvd): New builtins.
5433 (SPECIAL_ARGS2): New.
5434 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5435 (SPECIAL_ARGS2): New.
5436 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5437 (ix86_valid_target_attribute_inner_p): Ditto.
5438 (ix86_init_mmx_sse_builtins): Add special_args2.
5439 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5440 TARGET_WBNOINVD_P): New.
5441 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5442 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5443 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5444 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5445 * config/i386/pconfigintrin.h: New file.
5446 * config/i386/wbnoinvdintrin.h: Ditto.
5447 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5448 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5450 2018-03-05 Richard Biener <rguenther@suse.de>
5452 PR tree-optimization/84670
5453 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5455 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5456 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5457 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5458 doesn't grow if all successors (recursively) were visited at least
5461 2018-03-05 Richard Biener <rguenther@suse.de>
5463 PR tree-optimization/84650
5464 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5465 if executed in the loop pipeline.
5467 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5469 * doc/configfiles.texi (Configuration Files): Move info about
5470 conditionalizing $target-protos.h to...
5471 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5472 differs from $target-protos.h.
5474 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5475 Chung-Ju Wu <jasonwucj@gmail.com>
5477 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5478 * config/nds32/nds32-multiple.md (setmemsi): Define.
5479 * config/nds32/nds32-memory-manipulation.c
5480 (nds32_gen_dup_4_byte_to_word_value): New.
5481 (emit_setmem_word_loop): New.
5482 (emit_setmem_byte_loop): New.
5483 (nds32_expand_setmem_loop): New.
5484 (nds32_expand_setmem_loop_v3m): New.
5485 (nds32_expand_setmem_unroll): New.
5486 (nds32_expand_setmem): New.
5488 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5489 Chung-Ju Wu <jasonwucj@gmail.com>
5491 * config/nds32/nds32-memory-manipulation.c
5492 (nds32_emit_load_store): New.
5493 (nds32_emit_post_inc_load_store): New.
5494 (nds32_emit_mem_move): New.
5495 (nds32_emit_mem_move_block): New.
5496 (nds32_expand_movmemsi_loop_unknown_size): New.
5497 (nds32_expand_movmemsi_loop_known_size): New.
5498 (nds32_expand_movmemsi_loop): New.
5499 (nds32_expand_movmemsi_unroll): New.
5500 (nds32_expand_movmemqi): Rename ...
5501 (nds32_expand_movmemsi): ... to this.
5502 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5503 (movmemsi): ... to this.
5504 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5505 (nds32_expand_movmemsi): ... to this.
5507 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5508 Monk Chiang <sh.chiang04@gmail.com>
5509 Chung-Ju Wu <jasonwucj@gmail.com>
5511 * config/nds32/nds32-protos.h
5512 (nds32_expand_load_multiple): New arguments.
5513 (nds32_expand_store_multiple): Ditto.
5514 (nds32_valid_multiple_load_store): Rename ...
5515 (nds32_valid_multiple_load_store_p): ... to this.
5516 * config/nds32/nds32-memory-manipulation.c
5517 (nds32_expand_load_multiple): Refine implementation.
5518 (nds32_expand_store_multiple): Ditto.
5519 * config/nds32/nds32-multiple.md
5520 (load_multiple): Update nds32_expand_load_multiple interface.
5521 (store_multiple): Update nds32_expand_store_multiple interface.
5522 * config/nds32/nds32-predicates.c
5523 (nds32_valid_multiple_load_store): Rename ...
5524 (nds32_valid_multiple_load_store_p): ... to this and refine
5526 * config/nds32/predicates.md
5527 (nds32_load_multiple_and_update_address_operation): New predicate.
5528 (nds32_store_multiple_and_update_address_operation): New predicate.
5530 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5531 Chung-Ju Wu <jasonwucj@gmail.com>
5533 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5534 (combo): New attribute.
5535 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5537 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5539 * config/nds32/nds32.opt: Change -mcmodel= default value.
5541 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5542 Monk Chiang <sh.chiang04@gmail.com>
5543 Chung-Ju Wu <jasonwucj@gmail.com>
5545 * config/nds32/constants.md (unspec_element): New enum.
5546 * config/nds32/constraints.md (Umw): New constraint.
5547 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5548 * config/nds32/nds32-intrinsic.md: Likewise.
5549 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5550 (nds32_valid_smw_lwm_base_p): New.
5551 (nds32_output_smw_single_word): New.
5552 (nds32_output_lmw_single_word): New.
5553 (nds32_expand_unaligned_load): New.
5554 (nds32_expand_unaligned_store): New.
5555 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5556 (nds32_output_smw_single_word): Declare.
5557 (nds32_output_lmw_single_word): Declare.
5558 (nds32_expand_unaligned_load): Declare.
5559 (nds32_expand_unaligned_store): Declare.
5560 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5561 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5562 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5563 NDS32_BUILTIN_UASTORE_DW.
5564 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5567 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5568 Kito Cheng <kito.cheng@gmail.com>
5569 Chung-Ju Wu <jasonwucj@gmail.com>
5571 * config/nds32/nds32-intrinsic.c
5572 (nds32_expand_builtin_null_ftype_reg): Delete.
5573 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5574 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5575 (nds32_read_argument): New.
5576 (nds32_legitimize_target): Ditto.
5577 (nds32_legitimize_argument): Ditto.
5578 (nds32_check_constant_argument): Ditto.
5579 (nds32_expand_unop_builtin): Ditto.
5580 (nds32_expand_unopimm_builtin): Ditto.
5581 (nds32_expand_binop_builtin): Ditto.
5582 (nds32_builtin_decl_impl): Ditto.
5583 (builtin_description): Ditto.
5584 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5585 (nds32_init_builtins_impl): Ditto.
5586 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5587 (nds32_builtin_decl): New.
5588 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5589 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5591 2018-03-02 Jeff Law <law@redhat.com>
5593 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5594 (redundant_insn, fill_simple_delay_slots): Likewise.
5595 (fill_slots_from_thread): Likewise.
5596 * resource.c (mark_referenced_resources): Likewise.
5597 (mark_set_resources, find_dead_or_set_registers): Likewise.
5599 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5601 * substring-locations.h (format_warning_va): Formatting fix for
5603 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5605 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5606 * substring-locations.c: Include intl.h.
5607 (format_warning_va): Turned into small wrapper around
5608 format_warning_n_va, renamed to ...
5609 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5610 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5612 (format_warning_at_substring_n): New function.
5613 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5614 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5615 format_warning_at_substring with just a shorter name instead of
5616 const function pointer.
5617 (fmtwarn_n): New function.
5618 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5619 appropriate, get rid of all the fmtstr temporaries, move conditionals
5620 with G_() wrapped string literals directly into fmtwarn arguments,
5621 cast dir.len to (int), formatting fixes.
5623 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5625 * doc/invoke.texi: Remove "Cilk Plus" references.
5627 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5628 Richard Biener <rguenther@suse.de>
5631 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5632 for error or warning attributes if CALL_FROM_THUNK_P is set.
5635 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5638 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5639 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5642 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5643 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5645 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5646 instead of -1U in last predictors element's probability member.
5648 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5651 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5652 arguments if they are comparable.
5654 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5656 PR tree-optimization/84634
5657 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5658 masks and masked_loop_p with a single loop_masks, making sure it's
5659 null for bb vectorization.
5661 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5663 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5664 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5665 loop->force_vectorize to check whether there is no alias.
5667 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5670 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5672 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5674 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5675 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5676 instead of a loop around prev_real_insn.
5677 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5681 * config/i386/i386.c (ix86_print_operand): Use conditional
5682 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5685 2018-03-02 Richard Biener <rguenther@suse.de>
5687 PR tree-optimization/84427
5688 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5689 (bitmap_set_subtract_values): Rewrite to handle multiple
5692 (prune_clobbered_mems): Likewise.
5693 (phi_translate): Take edge instead of pred/phiblock.
5694 (phi_translate_1): Likewise.
5695 (phi_translate_set): Likewise. Insert all translated
5696 exprs for a value into the set, keeping possibly multiple
5697 expressions per value.
5698 (compute_antic_aux): Adjust for phi_translate changes.
5699 When intersecting union the expressions and prune those
5700 not in the final value set, keeping possibly multiple
5701 expressions per value. Do not use value-insertion
5702 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5703 all expressions. Add verification that the value-sets
5704 only shrink during iteration.
5705 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5706 (do_pre_regular_insertion): Likewise.
5707 (do_pre_partial_partial_insertion): Likewise.
5709 2018-03-02 Richard Biener <rguenther@suse.de>
5712 * config/darwin.c (saved_debug_info_level): New static global.
5713 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5714 (darwin_asm_lto_end): Restore debug info generation settings.
5716 2018-03-01 Martin Liska <mliska@suse.cz>
5719 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5722 2018-03-01 Richard Biener <rguenther@suse.de>
5725 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5726 type annotation with LTO when debug was disabled at compile-time.
5728 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5730 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5732 (mips_final_postscan_insn): Likewise.
5734 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5736 PR rtl-optimization/84528
5737 * alias.c (init_alias_target): Add commentary.
5738 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5739 a unique base value if the frame pointer is not eliminated
5740 to the stack pointer.
5742 2018-03-01 Tom de Vries <tom@codesourcery.com>
5744 PR rtl-optimization/83327
5745 * lra-int.h (hard_regs_spilled_into): Declare.
5746 * lra.c (hard_regs_spilled_into): Define.
5747 (init_reg_info): Init hard_regs_spilled_into.
5748 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5749 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5750 (process_bb_lives): Handle hard_regs_spilled_into.
5751 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5752 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5754 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5756 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5757 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5758 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5759 * config/rs6000/aix72.h: New file.
5761 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5763 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5764 instead of warning_at with conditional singular and plural messages
5768 * stor-layout.c (update_alignment_for_field): For
5769 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5770 && !DECL_PACKED (field), do the alignment update, just use
5771 only desired_align instead of MAX (type_align, desired_align)
5773 (place_field): Don't do known_align < desired_align handling
5774 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5775 is non-NULL, instead do it after rli->prev_field handling and
5776 only if not within a bitfield word. For DECL_PACKED (field)
5777 use type_align of BITS_PER_UNIT.
5779 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5781 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5782 superfluous parentheses and trailing spaces.
5784 2018-02-28 Richard Biener <rguenther@suse.de>
5786 PR tree-optimization/84584
5787 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5788 SCoPs with fake exit edge.
5790 2018-02-28 Martin Liska <mliska@suse.cz>
5793 * timevar.c (timer::print): Fix format to properly print 100%
5796 2018-02-28 Richard Biener <rguenther@suse.de>
5799 * genmatch.c (capture_info::walk_match): Do not mark
5800 captured expressions without operands as expr_p given
5801 they act more like predicates and should be subject to
5802 "lost tail" side-effect preserving.
5804 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5806 PR rtl-optimization/81611
5807 * auto-inc-dec.c (attempt_change): Move dead note from
5808 mem_insn if it's the next use of regno
5809 (find_address): Take address use of reg holding
5810 non-incremented value. Add parm to limit search to the named
5812 (merge_in_block): Attempt to use a mem insn that is the next
5813 use of the original regno.
5815 2018-02-27 Martin Sebor <msebor@redhat.com>
5818 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5819 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5821 2018-02-27 Martin Sebor <msebor@redhat.com>
5823 PR translation/84207
5824 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5825 n argument to unsigned HOST_WIDE_INT.
5826 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5827 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5828 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5829 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5831 2018-02-27 Richard Biener <rguenther@suse.de>
5833 PR tree-optimization/84512
5834 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5835 Do not use the estimate returned from record_stmt_cost for
5836 the scalar iteration cost but sum properly using add_stmt_cost.
5838 2018-02-27 Richard Biener <rguenther@suse.de>
5840 PR tree-optimization/84466
5841 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5842 Adjust last change to less strictly validate use operands.
5844 2018-02-27 Martin Liska <mliska@suse.cz>
5846 PR gcov-profile/84548
5847 * gcov.c (process_file): Allow partial overlap and consider it
5848 also as group functions.
5849 (output_lines): Properly calculate range of lines for a group.
5851 2018-02-27 Martin Liska <mliska@suse.cz>
5853 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5854 'ggc' suffixes. Change first column width.
5855 (timer::print): Fix formatting of the column.
5857 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5859 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5860 preserve inline entry blocks for the sake of debug inline
5861 entry point markers alone.
5862 (remove_unused_locals): Suggest in comments a better place to
5863 force the preservation of inline entry blocks that are
5864 otherwise unused, but do not preserve them.
5866 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5868 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5870 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5873 * config/i386/constraints.md (Bs): Replace
5874 ix86_indirect_branch_register with
5875 TARGET_INDIRECT_BRANCH_REGISTER.
5877 * config/i386/i386.md (indirect_jump): Likewise.
5878 (tablejump): Likewise.
5879 (*sibcall_memory): Likewise.
5880 (*sibcall_value_memory): Likewise.
5881 Peepholes of indirect call and jump via memory: Likewise.
5882 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5883 (*sibcall_value_GOT_32): Likewise.
5884 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5885 (GOT_memory_operand): Likewise.
5886 (call_insn_operand): Likewise.
5887 (sibcall_insn_operand): Likewise.
5888 (GOT32_symbol_operand): Likewise.
5889 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5891 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5893 PR rtl-optimization/83496
5894 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5895 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5896 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5897 redundant insn, if any.
5898 (relax_delay_slots): Likewise.
5899 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5901 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5903 PR tree-optimization/83965
5904 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5905 that grouped statements are part of a reduction chain. Return
5906 true if the statement is not marked as a reduction itself but
5908 (vect_recog_dot_prod_pattern): Don't check whether the statement
5909 is part of a group here.
5910 (vect_recog_sad_pattern): Likewise.
5911 (vect_recog_widen_sum_pattern): Likewise.
5913 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5916 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5918 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5921 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5923 (ix86_output_indirect_function_return): New prototype.
5924 (ix86_split_simple_return_pop_internal): Likewise.
5925 * config/i386/i386.c (indirect_return_via_cx): New.
5926 (indirect_return_via_cx_bnd): Likewise.
5927 (indirect_thunk_name): Handle return va CX_REG.
5928 (output_indirect_thunk_function): Create alias for
5929 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5930 (ix86_output_indirect_jmp): Remove the bool argument.
5931 (ix86_output_indirect_function_return): New function.
5932 (ix86_split_simple_return_pop_internal): Likewise.
5933 * config/i386/i386.md (*indirect_jump): Don't pass false
5934 to ix86_output_indirect_jmp.
5935 (*tablejump_1): Likewise.
5936 (simple_return_pop_internal): Change it to define_insn_and_split.
5937 Call ix86_split_simple_return_pop_internal to split it for
5939 (simple_return_indirect_internal): Call
5940 ix86_output_indirect_function_return instead of
5941 ix86_output_indirect_jmp.
5943 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5946 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5947 memset and value initialization afterwards.
5949 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5951 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5953 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5956 * common/config/aarch64/aarch64-common.c
5957 (aarch_option_optimization_table[]): Switch
5958 off fomit-frame-pointer
5960 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5961 Chung-Ju Wu <jasonwucj@gmail.com>
5963 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5965 (store_multiple): Ditto.
5967 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5969 * config.gcc: Add --with-cpu support for nds32 target.
5970 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5971 * config/nds32/nds32.opt: Add -mcpu= option.
5973 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5975 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5976 isel=yes): Warn for these deprecated options.
5978 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5980 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5981 ISA_2_5_MASKS_EMBEDDED.
5983 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5985 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5986 p->max as pointers rather than using iterative_hash_expr.
5988 2018-02-23 Carl Love <cel@us.ibm.com>
5990 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5991 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5993 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5994 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5995 P8V_BUILTIN_VEC_VUNSIGNED2.
5997 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6000 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6001 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6003 * lra-constraints.c (curr_insn_transform): Set up
6004 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6006 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6007 * lra-eliminations.c (spill_pseudos): Ditto.
6008 (process_insn_for_elimination): Ditto.
6009 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6010 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6011 LRA_NON_CLOBBERED_ALT.
6013 2018-02-22 Martin Sebor <msebor@redhat.com>
6015 PR tree-optimization/84480
6016 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6017 to maybe_diag_stxncpy_trunc. Call it.
6018 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6019 from gimple_fold_builtin_strcpy. Print inlining stack.
6020 (handle_builtin_stxncpy): Print inlining stack.
6021 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6023 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6026 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6027 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6028 and -fcheck-pointer-bounds are used together.
6029 (indirect_thunk_prefix): New enum.
6030 (indirect_thunk_need_prefix): New function.
6031 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6032 "_nt" instead of "_bnd" for NOTRACK prefix.
6033 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6034 (output_indirect_thunk_function): Likewise.
6036 (ix86_code_end): Update output_indirect_thunk_function calls.
6037 (ix86_output_indirect_branch_via_reg): Replace
6038 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6039 (ix86_output_indirect_branch_via_push): Likewise.
6040 (ix86_output_function_return): Likewise.
6041 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6042 incompatible with -fcf-protection=branch and
6043 -fcheck-pointer-bounds.
6045 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6048 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6049 Change gcc_assert call to output_operand_lossage.
6051 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6053 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6055 2018-02-22 DJ Delorie <dj@redhat.com>
6056 Sebastian Perta <sebastian.perta@renesas.com>
6057 Oleg Endo <olegendo@gcc.gnu.org>
6059 * config/rx/rx.c (rx_rtx_costs): New function.
6060 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6062 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6064 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6066 2018-02-22 Martin Liska <mliska@suse.cz>
6069 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6070 Add "native" as a possible value.
6072 2018-02-22 Martin Liska <mliska@suse.cz>
6075 * config/i386/i386.c (ix86_option_override_internal):
6076 Add "native" as a possible value for -march and -mtune.
6078 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6081 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6082 to all type variants.
6084 PR tree-optimization/84503
6085 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6086 width as info->bitpos + info->bitsize - start.
6087 (merged_store_group::merge_overlapping): Simplify width computation.
6088 (check_no_overlap): New function.
6089 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6090 start + width and last_order of the group, fail if check_no_overlap
6092 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6093 to group if check_no_overlap fails.
6095 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6097 * config/rs6000/altivec.md: Delete contraint arguments to
6098 define_expand, define_split, and define_peephole2, and in
6099 define_insn_and_split if always unused.
6100 * config/rs6000/darwin.md: Ditto.
6101 * config/rs6000/dfp.md: Ditto.
6102 * config/rs6000/rs6000.md: Ditto.
6103 * config/rs6000/sync.md: Ditto.
6104 * config/rs6000/vector.md: Ditto.
6105 * config/rs6000/vsx.md: Ditto.
6107 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6109 * config/rs6000/altivec.md: Write output control strings as braced
6110 blocks instead of double-quoted strings.
6111 * config/rs6000/darwin.md: Ditto.
6112 * config/rs6000/rs6000.md: Ditto.
6113 * config/rs6000/vector.md: Ditto.
6114 * config/rs6000/vsx.md: Ditto.
6116 2018-02-21 Jason Merrill <jason@redhat.com>
6118 PR c++/84314 - ICE with templates and fastcall attribute.
6119 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6121 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6123 * ipa-cp.c (determine_versionability): Fix comment typos.
6125 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6128 * ipa-cp.c (determine_versionability): Do not version functions caling
6131 2018-02-21 Martin Liska <mliska@suse.cz>
6134 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6135 Add "native" as a possible value.
6136 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6137 the macro when native cpu detection is available.
6139 2018-02-21 Martin Liska <mliska@suse.cz>
6142 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6143 Add "native" as a possible value.
6144 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6145 when native cpu detection is available.
6147 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6148 Martin Sebor <msebor@redhat.com>
6150 PR tree-optimization/84478
6151 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6153 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6154 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6155 support which is conservatively correct, for 2 only stay conservative
6156 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6157 argument to the 2 argument get_range_strlen, adjust 6 arg
6158 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6160 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6161 (gimple_fold_builtin_strlen): Pass true as last argument to
6164 2018-02-20 Martin Sebor <msebor@redhat.com>
6167 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6168 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6169 (builtin_memref::builtin_memref): Factor out parts into
6170 set_base_and_offset and call it.
6172 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6175 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6176 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6177 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6178 search at the associated MODE_INT.
6180 2018-02-20 Jeff Law <law@redhat.com>
6183 PR tree-optimization/81592
6185 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6186 for range data rather than using global data.
6187 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6188 range data rather than using global data.
6189 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6190 pass it to children as needed.
6191 (struct directive::fmtresult): Similarly.
6192 (struct directive::set_width): Similarly.
6193 (struct directive::set_precision): Similarly.
6194 (format_integer, format_directive, parse_directive): Similarly.
6195 (format_none): Accept unnamed vr_values parameter.
6196 (format_percent, format_floating, format_character): Similarly.
6197 (format_string, format_plain): Similarly.
6198 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6199 the EVRP range analyzer for range data rather than using global data.
6200 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6201 gimple-ssa-evrp-analyze.h
6202 (class sprintf_dom_walker): Add after_dom_children member function.
6203 Add evrp_range_analyzer member.
6204 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6205 range analyzer as needed.
6206 (sprintf_dom_walker::after_dom_children): New member function.
6207 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6209 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6210 (evrp_range_analyzer::pop_to_marker): Likewise.
6212 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6214 PR tree-optimization/84419
6215 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6216 with the required type if its current type is compatible but
6219 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6222 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6223 after vectorization.
6225 2018-02-20 Martin Liska <mliska@suse.cz>
6228 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6229 possible values if we don't have a hint.
6231 2018-02-20 Martin Liska <mliska@suse.cz>
6235 * final.c (shorten_branches): Build align_tab array with one
6237 * opts.c (finish_options): Add alignment option limit check.
6238 (MAX_CODE_ALIGN): Likewise.
6239 (MAX_CODE_ALIGN_VALUE): Likewise.
6240 * doc/invoke.texi: Document maximum allowed option value for
6241 all -falign-* options.
6243 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6246 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6247 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6248 * var-tracking.c (emit_note_insn_var_location): Remove all references
6249 to NOTE_INSN_CALL_ARG_LOCATION.
6250 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6251 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6252 Use copy_rtx_if_shared.
6253 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6254 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6255 (dwarf2out_var_location): Remove handling of
6256 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6258 * final.c (final_scan_insn): Remove all references to
6259 NOTE_INSN_CALL_ARG_LOCATION.
6260 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6261 before dumping final insns.
6262 * except.c (emit_note_eh_region_end): Remove all references to
6263 NOTE_INSN_CALL_ARG_LOCATION.
6264 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6265 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6266 * config/arc/arc.c (hwloop_optimize): Likewise.
6267 * config/arm/arm.c (create_fix_barrier): Likewise.
6268 * config/s390/s390.c (s390_chunkify_start): Likewise.
6269 * config/sh/sh.c (find_barrier): Likewise.
6270 * config/i386/i386.c (rest_of_insert_endbranch,
6271 ix86_seh_fixup_eh_fallthru): Likewise.
6272 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6273 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6274 * config/frv/frv.c (frv_function_prologue): Likewise.
6275 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6277 (note_outside_basic_block_p): Remove all references to
6278 NOTE_INSN_CALL_ARG_LOCATION.
6279 * gengtype.c (adjust_field_rtx_def): Likewise.
6280 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6282 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6283 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6286 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6289 PR tree-optimization/84452
6290 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6291 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6294 2018-02-19 Martin Liska <mliska@suse.cz>
6297 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6299 2018-02-19 Martin Liska <mliska@suse.cz>
6300 Richard Sandiford <richard.sandiford@linaro.org>
6302 PR tree-optimization/82491
6303 * gimple-fold.c (get_base_constructor): Make earlier bail out
6306 2018-02-19 Carl Love <cel@us.ibm.com>
6308 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6309 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6311 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6312 P8V_BUILTIN_VEC_NEG.
6314 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6316 * config/rl78/rl78.md (movdf): New define expand.
6318 2018-02-19 Martin Liska <mliska@suse.cz>
6321 * doc/invoke.texi: Fix typo.
6322 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6324 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6326 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6327 handle rs6000_single_float and rs6000_double_float specially for
6330 2018-02-16 Jeff Law <law@redhat.com>
6332 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6335 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6338 * ipa-inline.c (inline_small_functions): Fix a typo.
6340 2018-02-16 Nathan Sidwell <nathan@acm.org>
6342 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6344 2018-02-16 Carl Love <cel@us.ibm.com>
6346 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6347 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6348 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6349 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6350 expansion to P8V_BUILTIN_VEC_FLOAT2.
6352 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6354 PR rtl-optimization/70023
6355 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6356 src_regno into account.
6358 2018-02-16 Carl Love <cel@us.ibm.com>
6360 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6361 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6362 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6363 * config/rs6000/rs6000.c: Remove case statements for
6364 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6365 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6366 and P9V_BUILTIN_VEC_VINSERT4B.
6367 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6368 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6369 * config/rs6000/vsx.md:
6370 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6373 2018-02-16 Carl Love <cel@us.ibm.com>
6375 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6377 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6379 * config/rs6000/rs6000-c.c: Add the definitions for
6380 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6381 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6382 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6383 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6384 definition for insert4b and define insn *insert3b_internal.
6385 * doc/extend.texi: Add documentation for vec_extract4b.
6387 2018-02-16 Nathan Sidwell <nathan@acm.org>
6389 * doc/extend.texi (Backwards Compatibility): Mention friend
6390 injection. Note for-scope is deprecated.
6391 * doc/invoke.texi (-ffriend-injection): Deprecate.
6393 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6395 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6396 that moved to I2, also allow destinations that are a paradoxical
6397 subreg (instead of a normal reg).
6399 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6402 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6405 2018-02-16 Richard Biener <rguenther@suse.de>
6407 PR tree-optimization/84037
6408 PR tree-optimization/84016
6410 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6411 Adjust vec_construct for the fact we need additional higher latency
6412 128bit inserts for AVX256 and AVX512 vector builds.
6413 (ix86_add_stmt_cost): Scale vector construction cost for
6416 2018-02-16 Richard Biener <rguenther@suse.de>
6418 PR tree-optimization/84417
6419 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6420 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6421 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6423 2018-02-16 Martin Liska <mliska@suse.cz>
6426 * internal-fn.def (ASAN_CHECK): Set proper flags.
6427 (ASAN_MARK): Likewise.
6429 2018-02-16 Julia Koval <julia.koval@intel.com>
6431 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6432 from PTA_CANNONLAKE.
6434 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6437 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6438 Use ++iter rather than iter++ for std::list iterators.
6439 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6440 defer deleting them until all nodes in the forest are processed. Do
6441 free even leaf nodes. Change to_process into auto_vec.
6444 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6445 * vec.h (vec_default_construct): Use memset instead of placement new
6446 if BROKEN_VALUE_INITIALIZATION is defined.
6447 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6448 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6451 PR rtl-optimization/83723
6452 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6453 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6454 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6456 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6458 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6460 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6462 PR rtl-optimization/81443
6463 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6464 from inner REGs to paradoxical SUBREGs.
6466 2018-02-16 Richard Biener <rguenther@suse.de>
6468 PR tree-optimization/84399
6469 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6470 For operands we can analyze at their definition make sure we can
6471 analyze them at each use as well.
6473 2018-02-16 Richard Biener <rguenther@suse.de>
6475 PR tree-optimization/84190
6476 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6477 volatile accesses if the decl isn't volatile.
6479 2018-02-15 Jason Merrill <jason@redhat.com>
6481 PR c++/84314 - ICE with templates and fastcall attribute.
6482 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6483 TYPE_CANONICAL on an existing type.
6485 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6487 PR tree-optimization/84383
6488 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6489 dstoff nor call operand_equal_p if dstbase is NULL.
6491 PR tree-optimization/84334
6492 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6493 also a CONSTANT_CLASS_P, punt.
6495 2018-02-14 Jim Wilson <jimw@sifive.com>
6497 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6498 first SMALL_OPERAND check. New local min_second_step. Move assert
6499 to where locals are set. Add TARGET_RVC support.
6500 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6502 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6504 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6506 2018-02-14 Martin Sebor <msebor@redhat.com>
6508 PR tree-optimization/83698
6509 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6510 arrays constrain the offset range to their bounds.
6511 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6512 (builtin_access::overlap): Avoid setting the size of overlap if it's
6514 (maybe_diag_overlap): Also consider arrays when deciding what values
6515 of offsets to include in diagnostics.
6517 2018-02-14 Martin Sebor <msebor@redhat.com>
6520 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6521 that correspond to the kind of a declaration.
6523 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6526 * config/pa/pa.md: Load address of PIC label using the linkage table
6527 if the label is nonlocal.
6529 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6531 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6532 warning message if user requests -maltivec=be.
6533 * doc/invoke.texi: Document deprecation of -maltivec=be.
6535 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6538 * config/rs6000/rs6000-c.c: Update definitions for
6539 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6540 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6542 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6545 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6546 add _get_ssp intrinsics. Remove argument from
6547 __builtin_ia32_rdssp[d|q].
6548 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6549 * config/i386/i386-builtin.def: Remove argument from
6550 __builtin_ia32_rdssp[d|q].
6551 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6552 ix86_expand_special_args_builtin for _rdssp[d|q].
6553 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6554 Clear register before usage.
6555 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6556 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6558 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6560 PR tree-optimization/84357
6561 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6562 operand 1 of an ARRAY_REF too.
6564 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6567 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6568 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6570 (set_of_reg): New struct.
6571 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6572 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6573 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6575 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6576 Split into bitclr, bitset, bitinvert patterns if appropriate.
6577 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6578 use rx_fuse_in_memory_bitop.
6579 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6580 to named insn, correct maximum insn length.
6582 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6585 * machmode.def: Define a complex mode for PARTIAL_INT.
6586 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6588 * doc/rtl.texi: Document CSPImode.
6589 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6591 (msp430_hard_regno_nregs_with_padding): Likewise.
6593 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6596 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6598 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6600 PR rtl-optimization/84169
6601 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6602 we generated a parallel as new i3 and we split that to new i2 and i3
6603 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6604 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6605 those to i2, not i1. Partially rewrite this scan code.
6607 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6610 * stor-layout.c (place_field): For variable length fields, adjust
6611 offset_align afterwards not just based on the field's alignment,
6612 but also on the size.
6615 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6616 of exps and logs in the use_exp2 case.
6618 2018-02-13 Jeff Law <law@redhat.com>
6620 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6623 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6626 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6630 * common.opt (gas-loc-support, gas-locview-support): New.
6631 (ginline-points, ginternal-reset-location-views): New.
6632 * doc/invoke.texi: Document them. Use @itemx where intended.
6633 (gvariable-location-views): Adjust.
6634 * target.def (reset_location_view): New.
6635 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6636 (TARGET_RESET_LOCATION_VIEW): New.
6637 * doc/tm.texi: Rebuilt.
6638 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6639 (dwarf2out_default_as_locview_support): New.
6640 (output_asm_line_debug_info): Use option variables.
6641 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6642 (output_loc_list): Likewise.
6643 (add_high_low_attributes): Check option variables.
6644 Don't output entry view attribute in strict mode.
6645 (gen_inlined_subroutine_die): Check option variables.
6646 (dwarf2out_inline_entry): Likewise.
6647 (init_sections_and_labels): Likewise.
6648 (dwarf2out_early_finish): Likewise.
6649 (maybe_reset_location_view): New, from...
6650 (dwarf2out_var_location): ... here. Call it.
6651 * debug.h (dwarf2out_default_as_loc_support): Declare.
6652 (dwarf2out_default_as_locview_support): Declare.
6653 * hooks.c (hook_int_rtx_insn_0): New.
6654 * hooks.h (hook_int_rtx_insn_0): Declare.
6655 * toplev.c (process_options): Take -gas-loc-support and
6656 -gas-locview-support from dwarf2out. Enable
6657 -gvariable-location-views by default only with locview
6658 assembler support. Enable -ginternal-reset-location-views by
6659 default only if the target defines the corresponding hook.
6660 Enable -ginline-points by default if location views are
6661 enabled; force it disabled if statement frontiers are
6663 * tree-inline.c (expand_call_inline): Check option variables.
6664 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6666 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6668 PR tree-optimization/84321
6669 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6670 handling. Also check whether the anti-range contains any values
6671 that satisfy the mask; switch to a VR_RANGE if not.
6673 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6676 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6678 2018-02-13 Martin Jambor <mjambor@suse.cz>
6681 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6682 of call statements, also set location of a load to a temporary.
6684 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6686 * config/rl78/rl78.c (add_vector_labels): New function.
6687 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6688 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6689 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6690 which checks that no arguments are passed.
6691 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6692 * doc/extend.texi: Documentation for the new attribute.
6694 2018-02-13 Andreas Schwab <schwab@suse.de>
6696 * config/riscv/linux.h (CPP_SPEC): Define.
6698 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6701 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6702 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6703 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6704 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6705 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6706 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6707 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6708 temporarily for AES and PCLMUL builtins.
6710 PR tree-optimization/84339
6711 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6712 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6716 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6717 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6718 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6720 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6722 * omp-simd-clone.h: New file.
6723 * omp-simd-clone.c: Include omp-simd-clone.h.
6724 (expand_simd_clones): No longer static.
6725 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6726 cgraph.h and omp-simd-clone.h.
6727 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6728 (vect_recog_widen_shift_pattern): Formatting fix.
6729 (vect_pattern_recog_1): Don't check optab for calls.
6732 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6733 operands[2] into a REG before using gen_lowpart on it.
6735 2018-02-12 Jeff Law <law@redhat.com>
6738 * config/sh/sh.c (find_barrier): Consider a sibling call
6741 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6742 successfully back substituting a reg.
6744 2018-02-12 Richard Biener <rguenther@suse.de>
6746 PR tree-optimization/84037
6747 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6748 parameter, move visited init to caller.
6749 (vect_slp_analyze_operations): Separate cost from validity
6750 check, initialize visited once for all instances.
6751 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6753 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6755 (vect_model_promotion_demotion_cost): Likewise.
6756 (vectorizable_bswap): Guard cost modeling with !slp_node
6757 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6759 (vectorizable_call): Likewise.
6760 (vectorizable_conversion): Likewise.
6761 (vectorizable_assignment): Likewise.
6762 (vectorizable_shift): Likewise.
6763 (vectorizable_operation): Likewise.
6764 (vectorizable_store): Likewise.
6765 (vectorizable_load): Likewise.
6766 (vectorizable_condition): Likewise.
6767 (vectorizable_comparison): Likewise.
6769 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6772 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6773 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6774 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6776 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6779 * params.def (inline-min-speedup): Increase from 8 to 15.
6780 (max-inline-insns-auto): Decrease from 40 to 30.
6781 * ipa-split.c (consider_split): Add some buffer for function to
6782 be considered inlining candidate.
6783 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6786 2018-02-12 Richard Biener <rguenther@suse.de>
6788 PR tree-optimization/84037
6789 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6790 matched stmts if we cannot swap the non-matched ones.
6792 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6794 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6795 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6796 _mm_maskz_scalef_round_ss): New intrinsics.
6797 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6798 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6799 __builtin_ia32_scalefss_round): Remove.
6800 (__builtin_ia32_scalefsd_mask_round,
6801 __builtin_ia32_scalefss_mask_round): New intrinsics.
6802 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6803 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6804 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6805 "<round_constraint>")): Changed to ...
6806 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6807 "<round_scalar_constraint>")): ... this.
6808 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6809 %0, %1, %2<round_op3>}"): Changed to ...
6810 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6811 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6812 %2<round_scalar_mask_op3>}"): ... this.
6813 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6815 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6817 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6818 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6819 (_mm_maskz_sqrt_round_ss): New intrinsics.
6820 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6821 (__builtin_ia32_sqrtsd_mask_round)
6822 (__builtin_ia32_sqrtss_mask_round): New builtins.
6823 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6824 (__builtin_ia32_sqrtss_round): Remove.
6825 (__builtin_ia32_sqrtsd_mask_round)
6826 (__builtin_ia32_sqrtss_mask_round): New builtins.
6827 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6828 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6829 ((match_operand:VF_128 1 "vector_operand"
6830 "xBm,<round_constraint>")): Changed to ...
6831 ((match_operand:VF_128 1 "vector_operand"
6832 "xBm,<round_scalar_constraint>")): ... this.
6833 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6834 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6835 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6836 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6837 %<iptr>1<round_scalar_mask_op3>}): ... this.
6838 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6839 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6841 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6844 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6845 Cast vec_cmpeq result to correct type.
6846 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6847 Cast vec_cmpgt result to correct type.
6849 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6851 * final.c (final_scan_insn_1): Renamed from...
6852 (final_scan_insn): ... this. New wrapper, to recover
6853 seen from the outermost call in recursive ones.
6854 * config/sparc/sparc.c (output_return): Drop seen from call.
6855 (output_sibcall): Likewise.
6856 * config/visium/visium.c (output_branch): Likewise.
6858 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6860 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6863 2018-02-10 Alan Modra <amodra@gmail.com>
6866 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6867 Specify LR as an input.
6869 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6872 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6873 remove_member_access_dummy_vars): New functions.
6874 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6875 lower_omp_1, execute_lower_omp): Use them.
6877 PR rtl-optimization/84308
6878 * shrink-wrap.c (spread_components): Release todo vector.
6880 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6882 PR rtl-optimization/57193
6883 * ira-color.c (struct allocno_color_data): Add member
6884 conflict_allocno_hard_prefs.
6885 (update_conflict_allocno_hard_prefs): New.
6886 (bucket_allocno_compare_func): Add a preference based on
6887 conflict_allocno_hard_prefs.
6888 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6889 (color_allocnos): Remove a dead code. Initiate
6890 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6892 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6895 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6896 constraint from =wa to wa. Avoid a subreg on the output operand,
6897 instead use a pseudo and subreg it in a move.
6898 (p9_xxbrd_<mode>): Changed to ...
6899 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6900 (p9_xxbrd_v2df): New expander.
6901 (p9_xxbrw_<mode>): Changed to ...
6902 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6903 (p9_xxbrw_v4sf): New expander.
6905 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6907 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6909 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6912 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6913 multiply in 32-bit mode.
6914 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6915 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6918 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6920 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6921 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6922 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6923 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6925 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6928 * dwarf2out.c (is_trivial_indirect_ref): New function.
6929 (dwarf2out_late_global_decl): Do not generate a location
6930 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6931 and that are not defined in the current unit.
6933 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6935 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6936 instead of a libcall for UNORDERED.
6938 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6941 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6942 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6944 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6947 * config/s390/s390.c (s390_set_current_function): Invoke
6948 s390_indirect_branch_settings also if fndecl didn't change.
6950 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6952 * config/rs6000/rs6000.md (blockage): Set length to zero.
6954 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6956 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6958 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6961 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6962 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6966 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6967 PARALLEL incoming that failed vt_get_decl_and_offset check.
6970 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6971 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6973 (get_variable_section): For decls in named .bss* sections pass true as
6974 second argument to bss_initializer_p.
6976 2018-02-09 Marek Polacek <polacek@redhat.com>
6977 Jakub Jelinek <jakub@redhat.com>
6980 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6981 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6982 Sync some changes from cxx_fold_indirect_ref.
6984 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6986 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6988 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6989 (BLOCK_INLINE_ENTRY_LABEL): New.
6990 (dwarf2out_var_location): Disregard inline entry markers.
6991 (inline_entry_data): New struct.
6992 (inline_entry_data_hasher): New hashtable type.
6993 (inline_entry_data_hasher::hash): New.
6994 (inline_entry_data_hasher::equal): New.
6995 (inline_entry_data_table): New variable.
6996 (add_high_low_attributes): Add DW_AT_entry_pc and
6997 DW_AT_GNU_entry_view attributes if a pending entry is found
6998 in inline_entry_data_table. Add old entry_pc attribute only
6999 if debug nonbinding markers are disabled.
7000 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7001 markers are enabled.
7002 (block_within_block_p, dwarf2out_inline_entry): New.
7003 (dwarf2out_finish): Check that no entries remained in
7004 inline_entry_data_table.
7005 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7006 (final_scan_insn, notice_source_line): Likewise.
7007 (rest_of_clean_state): Skip inline entry markers.
7008 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7010 * gimple.c (gimple_build_debug_inline_entry): New.
7011 * gimple.h (enum gimple_debug_subcode): Add
7012 GIMPLE_DEBUG_INLINE_ENTRY.
7013 (gimple_build_debug_inline_entry): Declare.
7014 (gimple_debug_inline_entry_p): New.
7015 (gimple_debug_nonbind_marker_p): Adjust.
7016 * insn-notes.def (INLINE_ENTRY): New.
7017 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7018 inline entry marker notes.
7019 (print_insn): Likewise.
7020 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7021 (INSN_DEBUG_MARKER_KIND): Likewise.
7022 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7023 * tree-inline.c (expand_call_inline): Build and insert
7024 debug_inline_entry stmt.
7025 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7026 inline entry blocks early, if nonbind markers are enabled.
7027 (dump_scope_block): Dump fragment info.
7028 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7029 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7030 (gimple_build_debug_inline_entry): New.
7031 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7032 Enable/disable inline entry points too.
7033 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7034 (DEBUG_INSN): Describe inline entry markers.
7036 * common.opt (gvariable-location-views): New.
7037 (gvariable-location-views=incompat5): New.
7038 * config.in: Rebuilt.
7039 * configure: Rebuilt.
7040 * configure.ac: Test assembler for view support.
7041 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7042 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7043 * dwarf2out.c (var_loc_view): New typedef.
7044 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7045 (dwarf2out_locviews_in_attribute): New.
7046 (dwarf2out_locviews_in_loclist): New.
7047 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7048 (enum dw_line_info_opcode): Add LI_adv_address.
7049 (struct dw_line_info_table): Add view.
7050 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7051 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7052 (zero_view_p): New variable.
7053 (ZERO_VIEW_P): New macro.
7054 (output_asm_line_debug_info): New.
7055 (struct var_loc_node): Add view.
7056 (add_AT_view_list, AT_loc_list): New.
7057 (add_var_loc_to_decl): Add view param. Test it against last.
7058 (new_loc_list): Add view params. Record them.
7059 (AT_loc_list_ptr): Handle loc and view lists.
7060 (view_list_to_loc_list_val_node): New.
7061 (print_dw_val): Handle dw_val_class_view_list.
7062 (size_of_die): Likewise.
7063 (value_format): Likewise.
7064 (loc_list_has_views): New.
7065 (gen_llsym): Set vl_symbol too.
7066 (maybe_gen_llsym, skip_loc_list_entry): New.
7067 (dwarf2out_maybe_output_loclist_view_pair): New.
7068 (output_loc_list): Output view list or entries too.
7069 (output_view_list_offset): New.
7070 (output_die): Handle dw_val_class_view_list.
7071 (output_dwarf_version): New.
7072 (output_compilation_unit_header): Use it.
7073 (output_skeleton_debug_sections): Likewise.
7074 (output_rnglists, output_line_info): Likewise.
7075 (output_pubnames, output_aranges): Update version comments.
7076 (output_one_line_info_table): Output view numbers in asm comments.
7077 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7078 Call maybe_gen_llsym.
7079 (loc_list_from_tree_1): Adjust.
7080 (add_AT_location_description): Create view list attribute if
7081 needed, check it's absent otherwise.
7082 (convert_cfa_to_fb_loc_list): Adjust.
7083 (maybe_emit_file): Call output_asm_line_debug_info for test.
7084 (dwarf2out_var_location): Reset views as needed. Precompute
7085 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7086 attribute. Set view.
7087 (new_line_info_table): Reset next view.
7088 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7089 (dwarf2out_source_line): Likewise. Output view resets and labels to
7090 the assembler, or select appropriate line info opcodes.
7091 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7092 (optimize_string_length): Catch it. Adjust.
7093 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7094 dw_val_class_view_list, and remove it if no longer needed.
7095 (hash_loc_list): Hash view numbers.
7096 (loc_list_hasher::equal): Compare them.
7097 (optimize_location_lists): Check whether a view list symbol is
7098 needed, and whether the locview attribute is present, and
7099 whether they match. Remove the locview attribute if no longer
7101 (index_location_lists): Call skip_loc_list_entry for test.
7102 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7103 Use output_dwarf_version.
7104 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7105 (struct dw_val_node): Add val_view_list.
7106 * final.c (SEEN_NEXT_VIEW): New.
7107 (set_next_view_needed): New.
7108 (clear_next_view_needed): New.
7109 (maybe_output_next_view): New.
7110 (final_start_function): Rename to...
7111 (final_start_function_1): ... this. Take pointer to FIRST,
7112 add SEEN parameter. Emit param bindings in the initial view.
7113 (final_start_function): Reintroduce SEEN-less interface.
7114 (final): Rename to...
7115 (final_1): ... this. Take SEEN parameter. Output final pending
7116 next view at the end.
7117 (final): Reintroduce seen-less interface.
7118 (final_scan_insn): Output pending next view before switching
7119 sections or ending a block. Mark the next view as needed when
7120 outputting variable locations. Notify debug backend of section
7121 changes, and of location view changes.
7122 (rest_of_handle_final): Adjust.
7123 * toplev.c (process_options): Autodetect value for debug variable
7124 location views option. Warn on incompat5 without -gdwarf-5.
7125 * doc/invoke.texi (gvariable-location-views): New.
7126 (gvariable-location-views=incompat5): New.
7127 (gno-variable-location-views): New.
7129 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7131 PR tree-optimization/84136
7132 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7133 that the result of find_edge is non-NULL.
7135 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7138 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7139 storing integer register in SImode. Fix cost of 256 and 512
7140 byte aligned SSE register store.
7142 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7144 * config/i386/i386.c (ix86_multiplication_cost): Fix
7145 multiplication cost for TARGET_AVX512DQ.
7147 2018-02-08 Marek Polacek <polacek@redhat.com>
7149 PR tree-optimization/84238
7150 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7153 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7155 PR tree-optimization/84265
7156 * tree-vect-stmts.c (vectorizable_store): Don't treat
7157 VMAT_CONTIGUOUS accesses as grouped.
7158 (vectorizable_load): Likewise.
7160 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7162 PR tree-optimization/81635
7163 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7164 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7165 (test_round_for_mask): New functions.
7166 (wide_int_cc_tests): Call test_round_for_mask.
7167 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7168 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7169 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7170 range returned by get_range_info.
7172 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7175 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7176 * symtab.c: Include builtins.h
7177 (symtab_node::output_to_lto_symbol_table_p): Move here
7178 from lto-streamer-out.c:output_symbol_p.
7179 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7180 (output_symbol_p): Move all logic to symtab.c
7181 (produce_symtab): Update.
7183 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7185 * config/s390/s390-opts.h (enum indirect_branch): Define.
7186 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7187 (s390_indirect_branch_via_thunk)
7188 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7189 (enum s390_indirect_branch_type): Define.
7190 * config/s390/s390.c (struct s390_frame_layout, struct
7191 machine_function): Remove.
7192 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7193 (indirect_branch_table_label_no, indirect_branch_table_name):
7195 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7196 (enum s390_indirect_branch_option): Define.
7197 (s390_return_addr_from_memory): New function.
7198 (s390_handle_string_attribute): New function.
7199 (s390_attribute_table): Add new attribute handler.
7200 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7201 (s390_indirect_branch_via_thunk): New function.
7202 (s390_indirect_branch_via_inline_thunk): New function.
7203 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7204 sibling call optimization for non z10 compiles.
7205 (s390_emit_call): Force indirect branch target to be a single
7206 register. Add r1 clobber for non-z10 compiles.
7207 (s390_emit_epilogue): Emit return jump via return_use expander.
7208 (s390_reorg): Handle JUMP_INSNs as execute targets.
7209 (s390_option_override_internal): Perform validity checks for the
7210 new command line options.
7211 (s390_indirect_branch_attrvalue): New function.
7212 (s390_indirect_branch_settings): New function.
7213 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7214 (s390_output_indirect_thunk_function): New function.
7215 (s390_code_end): Implement target hook.
7216 (s390_case_values_threshold): Implement target hook.
7217 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7219 * config/s390/s390.h (struct s390_frame_layout)
7220 (struct machine_function): Move here from s390.c.
7221 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7222 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7223 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7224 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7225 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7226 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7227 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7228 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7229 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7230 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7231 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7232 (mnemonic attribute): Add values which aren't recognized
7234 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7235 pattern for branch conversion. Fix mnemonic attribute.
7236 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7237 indirect branch via thunk if requested.
7238 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7239 ("*indirect_jump"): Disable for branch conversion using out of
7241 ("indirect_jump_via_thunk<mode>_z10")
7242 ("indirect_jump_via_thunk<mode>")
7243 ("indirect_jump_via_inlinethunk<mode>_z10")
7244 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7245 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7246 ("casesi_jump_via_inlinethunk<mode>_z10")
7247 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7248 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7249 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7250 ("*indirect2_jump"): Disable for branch conversion.
7251 ("casesi_jump"): Turn into expander and expand patterns for branch
7253 ("return_use"): New expander.
7254 ("*return"): Emit return via thunk and rename it to ...
7255 ("*return<mode>"): ... this one.
7256 * config/s390/s390.opt: Add new options and and enum for the
7259 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7261 * lra-constraints.c (match_reload): Unconditionally use
7262 gen_lowpart_SUBREG, rather than selecting between that
7263 and equivalent gen_rtx_SUBREG code.
7265 2018-02-08 Richard Biener <rguenther@suse.de>
7267 PR tree-optimization/84233
7268 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7269 changed flag instead of boguously re-using phi_inserted.
7271 2018-02-08 Martin Jambor <mjambor@suse.cz>
7273 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7274 static local variables.
7276 2018-02-08 Richard Biener <rguenther@suse.de>
7278 PR tree-optimization/84278
7279 * tree-vect-stmts.c (vectorizable_store): When looking for
7280 smaller vector types to perform grouped strided loads/stores
7281 make sure the mode is supported by the target.
7282 (vectorizable_load): Likewise.
7284 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7286 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7287 Increase LDP/STP opportunities by adding adjacent callee-saves.
7289 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7291 PR rtl-optimization/84068
7292 PR rtl-optimization/83459
7293 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7295 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7297 PR tree-optimization/84224
7298 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7299 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7302 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7305 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7306 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7307 count, remove one USE.
7309 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7311 * doc/install.texi (Configuration): Document the
7312 --with-long-double-format={ibm,ieee} PowerPC configuration
7316 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7317 Convert from define_expand to be define_insn_and_split. Rework
7318 float/double/_Float128 conversions to QI/HI/SImode to work with
7319 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7320 conversions to QI/HImode types did a store and then a load to
7321 truncate the value. For conversions to VSX registers, don't split
7322 the insn, instead emit the code directly. Use the code iterator
7323 any_fix to combine signed and unsigned conversions.
7324 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7325 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7326 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7327 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7328 (fix_<mode>di2_hw): Likewise.
7329 (fixuns_<mode>di2_hw): Likewise.
7330 (fix_<mode>si2_hw): Likewise.
7331 (fixuns_<mode>si2_hw): Likewise.
7332 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7333 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7334 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7335 fix<uns>_trunc<SFDF:mode>si2_p8.
7336 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7338 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7339 (fix<uns>_<mode>_mem): Likewise.
7340 (fctiw<u>z_<mode>_mem): Likewise.
7341 (fix<uns>_<mode>_mem): Likewise.
7342 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7343 the register allocator from doing a direct move to the GPRs to do
7344 a store, and instead use the ISA 3.0 store byte/half-word from
7345 vector register instruction. For IEEE 128-bit floating point,
7346 also optimize stores of 32-bit ints.
7347 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7349 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7351 * genextract.c (push_pathstr_operand): New function to support
7353 (walk_rtx): Call push_pathstr_operand.
7354 (print_path): Support [a-zA-Z].
7356 2018-02-07 Richard Biener <rguenther@suse.de>
7358 PR tree-optimization/84037
7359 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7360 (cse_and_gimplify_to_preheader): Declare.
7361 (vect_get_place_in_interleaving_chain): Likewise.
7362 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7364 (_loop_vec_info::~_loop_vec_info): Delete it.
7365 (cse_and_gimplify_to_preheader): New function.
7366 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7367 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7368 (vectorizable_load): Likewise. For grouped stores always base
7369 the IV on the first element.
7370 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7371 condition before gimplifying.
7373 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7375 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7376 *DIV_EXPR and *MOD_EXPR.
7378 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7381 * config/i386/i386.c (ix86_option_override_internal): Mask out
7382 the CF_SET bit when checking -fcf-protection.
7384 2018-02-07 Tom de Vries <tom@codesourcery.com>
7387 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7390 2018-02-07 Richard Biener <rguenther@suse.de>
7392 PR tree-optimization/84204
7393 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7396 PR tree-optimization/84205
7397 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7398 special-case isl_ast_op_zdiv_r.
7400 PR tree-optimization/84223
7401 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7402 Only add conditions from within the region.
7403 (gather_bbs::after_dom_children): Adjust.
7405 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7408 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7409 * config/avr/avr.md: Only post-reload split REG-REG moves if
7410 either register is GENERAL_REG_P.
7412 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7414 PR tree-optimization/84235
7415 * tree-ssa-scopedtables.c
7416 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7417 if the subtraction is performed in floating point type where NaNs are
7418 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7419 build 1. Formatting fix.
7421 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7424 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7425 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7426 and skip it regardless of bb boundaries. Use CALL_P macro,
7427 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7428 unnecessarily, formatting fix.
7430 2018-02-06 Michael Collison <michael.collison@arm.com>
7432 * config/arm/thumb2.md:
7433 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7434 (*thumb_mov_notscc): Ditto.
7436 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7439 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7440 unsigned_fix, not "s".
7442 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7444 * configure.ac (gcc_fn_eh_frame_ro): New function.
7445 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7446 correct .eh_frame permissions.
7447 * configure: Regenerate.
7449 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7451 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7454 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7456 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7457 Display warning message for -mno-speculate-indirect-jumps.
7459 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7461 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7463 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7465 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7467 PR tree-optimization/84225
7468 * tree-eh.c (find_trapping_overflow): Only call
7469 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7471 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7474 * config/i386/i386.c: Reimplement the check of possible options
7475 -mibt/-mshstk conbination. Change error messages.
7476 * doc/invoke.texi: Fix a typo: remove extra '='.
7478 2018-02-06 Marek Polacek <polacek@redhat.com>
7480 PR tree-optimization/84228
7481 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7483 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7486 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7487 emitted arch directives.
7488 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7489 __ARM_FEATURE_COPROC before changing architectures.
7491 2018-02-06 Richard Biener <rguenther@suse.de>
7493 * config/i386/i386.c (print_reg): Fix typo.
7494 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7496 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7498 * configure: Regenerate.
7500 2018-02-05 Martin Sebor <msebor@redhat.com>
7502 PR tree-optimization/83369
7503 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7506 2018-02-05 Martin Liska <mliska@suse.cz>
7508 * doc/invoke.texi: Cherry-pick upstream r323995.
7510 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7512 * ira.c (ira_init_register_move_cost): Adjust comment.
7514 2018-02-05 Martin Liska <mliska@suse.cz>
7516 PR gcov-profile/84137
7517 * doc/gcov.texi: Fix typo in documentation.
7519 2018-02-05 Martin Liska <mliska@suse.cz>
7521 PR gcov-profile/83879
7522 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7523 using dlopen functionality.
7525 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7527 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7528 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7529 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7530 _mm_maskz_range_round_ss): New intrinsics.
7531 (__builtin_ia32_rangesd128_round)
7532 (__builtin_ia32_rangess128_round): Remove.
7533 (__builtin_ia32_rangesd128_mask_round,
7534 __builtin_ia32_rangess128_mask_round): New builtins.
7535 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7536 __builtin_ia32_rangess128_round): Remove.
7537 (__builtin_ia32_rangesd128_mask_round,
7538 __builtin_ia32_rangess128_mask_round): New builtins.
7539 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7540 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7541 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7542 "<round_saeonly_constraint>")): Changed to ...
7543 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7544 "<round_saeonly_scalar_constraint>")): ... this.
7545 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7546 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7547 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7548 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7549 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7551 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7553 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7555 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7556 Remove all values except native, 8540 and 8548.
7558 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7560 * config/i386/i386.c (ix86_output_function_return): Pass
7561 INVALID_REGNUM, instead of -1, as invalid register number to
7562 indirect_thunk_name and output_indirect_thunk.
7564 2018-02-02 Julia Koval <julia.koval@intel.com>
7566 * config.gcc: Add -march=icelake.
7567 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7568 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7569 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7570 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7571 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7572 (processor_target_table): Add icelake.
7573 (ix86_option_override_internal): Handle new PTAs.
7574 (get_builtin_code_for_version): Handle icelake.
7575 (M_INTEL_COREI7_ICELAKE): New.
7576 (fold_builtin_cpu): Handle icelake.
7577 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7578 * doc/invoke.texi: Add -march=icelake.
7580 2018-02-02 Julia Koval <julia.koval@intel.com>
7582 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7583 to wide_int_bitmask.
7584 * wide-int-bitmask.h: New.
7586 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7589 * config/i386/i386.md: Replace Pmode with word_mode in
7590 builtin_setjmp_setup and builtin_longjmp to support x32.
7592 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7596 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7598 (rs6000_supported_cpu_names): New static variable.
7599 (linux_cpu_translation_table): Likewise.
7600 (elf_platform) <cpu>: Define new static variable and use it.
7601 Translate kernel AT_PLATFORM name to canonical name if needed.
7602 Error if platform name is unknown.
7604 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7607 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7609 2018-02-01 Jeff Law <law@redhat.com>
7612 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7613 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7614 the scratch if RELEASE_VIA_POP is false.
7615 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7616 If we have to save a temporary register, decrement SIZE appropriately.
7617 Pass new arguments to release_scratch_register_on_entry.
7618 (ix86_adjust_stack_and_probe): Likewise.
7619 (ix86_emit_probe_stack_range): Pass new arguments to
7620 release_scratch_register_on_entry.
7622 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7624 PR rtl-optimization/84157
7625 * combine.c (change_zero_ext): Use REG_P predicate in
7626 front of HARD_REGISTER_P predicate.
7628 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7630 * config/avr/avr.c (avr_option_override): Move disabling of
7631 -fdelete-null-pointer-checks to...
7632 * common/config/avr/avr-common.c (avr_option_optimization_table):
7635 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7637 PR tree-optimization/81635
7638 * tree-data-ref.c (split_constant_offset_1): For types that
7639 wrap on overflow, try to use range info to prove that wrapping
7642 2018-02-01 Renlin Li <renlin.li@arm.com>
7645 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7647 (aarch64_register_move_cost): Likewise.
7648 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7650 (REG_CLASS_NAMES): Likewise.
7651 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7652 TAILCALL_ADDR_REGS. Remove IP registers.
7653 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7655 2018-02-01 Richard Biener <rguenther@suse.de>
7657 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7658 for specifying RPO order and allow NULL for that.
7659 * domwalk.c (dom_walker::dom_walker): Likewise.
7660 (dom_walker::walk): Handle NULL RPO order.
7661 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7663 (rewrite_update_dom_walker): Likewise.
7664 (mark_def_dom_walker): Likewise.
7666 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7668 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7669 (aarch64_maybe_expand_sve_subreg_move): Declare.
7670 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7671 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7673 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7674 that are semantically a reverse operation.
7675 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7676 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7677 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7679 (aarch64_can_change_mode_class): For big-endian, forbid changes
7680 between two SVE modes if they have different element sizes.
7682 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7684 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7685 the TImode handling for big-endian targets.
7687 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7689 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7690 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7692 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7693 Remove BSWAP handing for big-endian targets and use the form of
7694 LD1RQ appropariate for the mode.
7696 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7698 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7699 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7702 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7705 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7706 check for operands that need to go through aarch64_sve_reload_be.
7708 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7710 PR tree-optimization/81661
7711 PR tree-optimization/84117
7712 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7713 * tree-eh.c: Include gimplify.h.
7714 (find_trapping_overflow, replace_trapping_overflow,
7715 rewrite_to_non_trapping_overflow): New functions.
7716 * tree-vect-loop.c: Include tree-eh.h.
7717 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7718 * tree-data-ref.c: Include tree-eh.h.
7719 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7721 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7723 PR rtl-optimization/84123
7724 * combine.c (change_zero_ext): Check if hard register satisfies
7725 can_change_dest_mode before calling gen_lowpart_SUBREG.
7727 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7730 * ira.c (ira_init_register_move_cost): Remove assert.
7732 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7734 PR rtl-optimization/84071
7735 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7736 * doc/tm.texi: Regenerate.
7738 2018-01-31 Richard Biener <rguenther@suse.de>
7740 PR tree-optimization/84132
7741 * tree-data-ref.c (analyze_miv_subscript): Properly
7742 check whether evolution_function_is_affine_multivariate_p
7743 before calling gcd_of_steps_may_divide_p.
7745 2018-01-31 Julia Koval <julia.koval@intel.com>
7748 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7749 * config/i386/i386.md (rdpid_rex64) New.
7750 (rdpid): Make 32bit only.
7752 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7755 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7756 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7758 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7761 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7763 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7765 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7767 PR rtl-optimization/84071
7768 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7769 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7771 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7773 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7774 (arc_attribute_table): Add 'aux' attribute.
7775 (arc_in_small_data_p): Consider aux like variables.
7776 (arc_is_aux_reg_p): New function.
7777 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7778 (arc_get_aux_arg): New function.
7779 (prepare_move_operands): Handle aux-register access.
7780 (arc_handle_aux_attribute): New function.
7781 * doc/extend.texi (ARC Variable attributes): Add subsection.
7783 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7785 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7786 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7787 (arc_attribute_table): Add 'uncached' attribute.
7788 (arc_print_operand): Print '.di' flag for uncached memory
7790 (arc_in_small_data_p): Do not consider for small data the uncached
7792 (arc_is_uncached_mem_p): New function.
7793 * config/arc/predicates.md (compact_store_memory_operand): Check
7794 for uncached memory accesses.
7795 (nonvol_nonimm_operand): Likewise.
7796 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7798 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7801 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7802 falign-loops=): Add Optimization flag.
7804 2018-01-30 Jeff Law <law@redhat.com>
7807 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7808 INT_REGISTERS_SAVED. Check it prior to calling
7809 get_scratch_register_on_entry.
7810 (ix86_adjust_stack_and_probe): Similarly.
7811 (ix86_emit_probe_stack_range): Similarly.
7812 (ix86_expand_prologue): Corresponding changes.
7814 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7817 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7818 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7820 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7823 * lra-constraints.c (curr_insn_transform): Process AND in the
7826 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7828 PR rtl-optimization/83986
7829 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7830 dependence against last_pending_memory_flush in addition to
7833 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7835 PR tree-optimization/81611
7836 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7839 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7842 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7845 2018-01-30 Richard Biener <rguenther@suse.de>
7846 Jakub Jelinek <jakub@redhat.com>
7848 PR tree-optimization/84111
7849 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7850 inner loops added during recursion, as they don't have up-to-date
7853 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7856 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7857 (can_inline_edge_by_limits_p): ... here.
7858 (can_early_inline_edge_p, check_callers,
7859 update_caller_keys, update_callee_keys, recursive_inlining,
7860 add_new_edges_to_heap, speculation_useful_p,
7861 inline_small_functions,
7862 inline_small_functions, flatten_function,
7863 inline_to_all_callers_1): Update.
7865 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7867 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7870 2018-01-30 Richard Biener <rguenther@suse.de>
7872 PR tree-optimization/83008
7873 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7874 invariant and constant vector uses in stmts when they need
7877 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7880 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7881 * configure: Regenerate.
7883 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7885 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7887 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7888 Use gen_rtx_REG rather than gen_lowpart.
7890 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7892 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7893 rather than 0 when creating partial subregs.
7895 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7897 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7900 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7903 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7904 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7905 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7906 flags. This restores the settings used before the 2017-07-24.
7907 Turning off pre increment/decrement/modify allows IVOPTS to
7908 optimize DF/SF loops where the index is an int.
7910 2018-01-29 Richard Biener <rguenther@suse.de>
7911 Kelvin Nilsen <kelvin@gcc.gnu.org>
7914 * tree-vect-stmts.c (vectorizable_call): Don't call
7915 targetm.vectorize_builtin_md_vectorized_function if callee is
7918 2018-01-22 Carl Love <cel@us.ibm.com>
7920 * doc/extend.tex: Fix typo in second arg in
7921 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7923 2018-01-29 Richard Biener <rguenther@suse.de>
7925 PR tree-optimization/84086
7926 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7927 (flush_ssaname_freelist): When SSA names were released reset
7928 the SCEV hash table.
7930 2018-01-29 Richard Biener <rguenther@suse.de>
7932 PR tree-optimization/84057
7933 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7934 removed paths when removing edges.
7936 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7938 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7939 -mfunction-return=@var{choice}.
7941 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7944 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7946 (layout::print_source_line): Likewise.
7947 (test_get_line_width_without_trailing_whitespace): Add test cases.
7949 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7952 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7955 2018-01-26 Jim Wilson <jimw@sifive.com>
7957 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7959 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7962 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7964 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7965 and CMP + SUB-immediate -> SUBS.
7967 2018-01-26 Martin Sebor <msebor@redhat.com>
7969 PR tree-optimization/83896
7970 * tree-ssa-strlen.c (get_string_len): Rename...
7971 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7972 Avoid assuming length is constant.
7973 (handle_char_store): Use HOST_WIDE_INT for string length.
7975 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7978 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7979 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7981 2018-01-26 Richard Biener <rguenther@suse.de>
7983 PR rtl-optimization/84003
7984 * dse.c (record_store): Only record redundant stores when
7985 the earlier store aliases at least all accesses the later one does.
7987 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7989 PR rtl-optimization/83985
7990 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7991 REG_CFA_RESTORE insns.
7992 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7993 REG_CFA_RESTORE insns here.
7996 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7997 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7999 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8001 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8002 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8003 (arc_init): Likewise.
8004 (arc_override_options): Likewise.
8005 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8007 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8009 * config/arc/arc.h (TARGET_DBNZ): Define.
8010 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8011 properly set the tune attribute.
8012 (dbnz): Use TARGET_DBNZ guard.
8013 * config/arc/arc.opt (mtune): Add core3 option.
8015 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8017 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8018 recognize new pic like addresses.
8019 (arc_delegitimize_address): Clean up.
8021 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8023 * config/arc/arc-arches.def: Option mrf16 valid for all
8025 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8026 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8027 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8028 * config/arc/arc-tables.opt: Regenerate.
8029 * config/arc/arc.c (arc_conditional_register_usage): Handle
8030 reduced register file case.
8031 (arc_file_start): Set must have build attributes.
8032 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8034 * config/arc/arc.opt (mrf16): Add new option.
8035 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8036 * config/arc/genmultilib.awk: Handle new mrf16 option.
8037 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8038 * config/arc/t-multilib: Regenerate.
8039 * doc/invoke.texi (ARC Options): Document mrf16 option.
8041 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8043 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8044 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8045 (arc_attribute_table): Add 'secure_call' attribute.
8046 (arc_print_operand): Print secure call operand.
8047 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8049 (arc_is_secure_call_p): New function. * config/arc/arc.md
8050 (call_i): Add support for sjli instruction.
8051 (call_value_i): Likewise.
8052 * config/arc/constraints.md (Csc): New constraint.
8054 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8055 John Eric Martin <John.Martin@emmicro-us.com>
8057 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8058 * config/arc/arc.c (_arc_jli_section): New struct.
8059 (arc_jli_section): New type.
8060 (rc_jli_sections): New static variable.
8061 (arc_handle_jli_attribute): New function.
8062 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8063 (arc_file_end): New function.
8064 (TARGET_ASM_FILE_END): Define.
8065 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8066 (arc_add_jli_section): New function.
8067 (jli_call_scan): Likewise.
8068 (arc_reorg): Call jli_call_scan.
8069 (arc_output_addsi): Remove 'S' from printing asm operand.
8070 (arc_is_jli_call_p): New function.
8071 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8073 (movhi_insn): Likewise.
8074 (movsi_insn): Likewise.
8075 (movsi_set_cc_insn): Likewise.
8076 (loadqi_update): Likewise.
8077 (load_zeroextendqisi_update): Likewise.
8078 (load_signextendqisi_update): Likewise.
8079 (loadhi_update): Likewise.
8080 (load_zeroextendhisi_update): Likewise.
8081 (load_signextendhisi_update): Likewise.
8082 (loadsi_update): Likewise.
8083 (loadsf_update): Likewise.
8084 (movsicc_insn): Likewise.
8085 (bset_insn): Likewise.
8086 (bxor_insn): Likewise.
8087 (bclr_insn): Likewise.
8088 (bmsk_insn): Likewise.
8089 (bicsi3_insn): Likewise.
8090 (cmpsi_cc_c_insn): Likewise.
8091 (movsi_ne): Likewise.
8092 (movsi_cond_exec): Likewise.
8093 (clrsbsi2): Likewise.
8104 (call_i): Remove 'S' asm letter, add jli instruction.
8105 (call_value_i): Likewise.
8106 * config/arc/arc.op (mjli-always): New option.
8107 * config/arc/constraints.md (Cji): New constraint.
8108 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8110 (subsf3_fpx): Likewise.
8111 (mulsf3_fpx): Likewise.
8112 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8114 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8116 * doc/invoke.texi (ARC): Document mjli-always option.
8118 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8120 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8121 avoid addition with 0 and use incw and decw where possible.
8123 2018-01-26 Richard Biener <rguenther@suse.de>
8125 PR tree-optimization/81082
8126 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8127 association if it requires casting to unsigned.
8128 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8129 from fold_plusminus_mult_expr to catch important cases late when
8130 range info is available.
8132 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8134 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8135 * configure.ac (hidden_linkonce): New test.
8136 * configure: Regenerate.
8137 * config.in: Regenerate.
8139 2018-01-26 Julia Koval <julia.koval@intel.com>
8141 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8142 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8143 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8144 _mm_mask_bitshuffle_epi64_mask): Fix type.
8145 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8146 USI_FTYPE_V4DI_V4DI_USI): Remove.
8147 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8148 __builtin_ia32_vpshufbitqmb256_mask,
8149 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8150 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8151 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8153 2018-01-26 Alan Modra <amodra@gmail.com>
8156 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8157 UNSPEC_VBPERMQ. Sort other unspecs.
8159 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8161 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8163 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8166 * predict.c (drop_profile): Do not push/pop cfun; update also
8168 (handle_missing_profiles): Fix logic looking for zero profiles.
8170 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8173 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8174 on functions with #pragma omp declare simd or functions with simd
8176 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8177 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8178 Remove trailing \n from warning_at calls.
8180 2018-01-25 Tom de Vries <tom@codesourcery.com>
8183 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8184 for neutered workers.
8186 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8189 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8190 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8192 2018-01-24 Jeff Law <law@redhat.com>
8195 * i386.c (get_probe_interval): Move to earlier point.
8196 (ix86_compute_frame_layout): If -fstack-clash-protection and
8197 the frame is larger than the probe interval, then use pushes
8198 to save registers rather than reg->mem moves.
8199 (ix86_expand_prologue): Remove conditional for int_registers_saved
8202 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8205 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8206 min/max for never referenced object.
8208 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8211 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8213 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8214 attributes from DECL_ATTRIBUTES (decl) without affecting
8215 DECL_ATTRIBUTES (current_function_decl).
8216 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8217 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8219 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8221 PR tree-optimization/83979
8222 * fold-const.c (fold_comparison): Use constant_boolean_node
8223 instead of boolean_{true,false}_node.
8225 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8227 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8230 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8232 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8233 Simplify the clause that sets the length attribute.
8234 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8235 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8236 clause that sets the length attribute.
8237 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8239 2018-01-24 Tom de Vries <tom@codesourcery.com>
8242 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8243 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8244 Add strict parameter.
8245 (prevent_branch_around_nothing): Insert dummy insn between branch to
8246 label and label with no ptx insn inbetween.
8247 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8249 2018-01-24 Tom de Vries <tom@codesourcery.com>
8252 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8253 for neutered threads in warp.
8254 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8256 2018-01-24 Richard Biener <rguenther@suse.de>
8258 PR tree-optimization/83176
8259 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8262 2018-01-24 Richard Biener <rguenther@suse.de>
8264 PR tree-optimization/82819
8265 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8266 code generating pluses that are no-ops in the target precision.
8268 2018-01-24 Richard Biener <rguenther@suse.de>
8271 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8273 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8275 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8276 to merge probabilities.
8277 * predict.c (probably_never_executed): Also mark as cold functions
8278 with global 0 profile and guessed local profile.
8279 * profile-count.c (profile_probability::combine_with_count): New
8281 * profile-count.h (profile_probability::operator*,
8282 profile_probability::operator*=, profile_probability::operator/,
8283 profile_probability::operator/=): Reduce precision to adjusted
8284 and set value to guessed on contradictory divisions.
8285 (profile_probability::combine_with_freq): Remove.
8286 (profile_probability::combine_wiht_count): Declare.
8287 (profile_count::force_nonzero):: Set to adjusted.
8288 (profile_count::probability_in):: Set quality to adjusted.
8289 * tree-ssa-tail-merge.c (replace_block_by): Use
8292 2018-01-23 Andrew Waterman <andrew@sifive.com>
8293 Jim Wilson <jimw@sifive.com>
8295 * config/riscv/riscv.c (riscv_stack_boundary): New.
8296 (riscv_option_override): Set riscv_stack_boundary. Handle
8297 riscv_preferred_stack_boundary_arg.
8298 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8299 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8300 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8301 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8302 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8303 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8305 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8308 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8309 of struct ix86_frame.
8310 (ix86_expand_epilogue): Likewise. Add a local variable for
8311 the reg_save_offset field in struct ix86_frame.
8313 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8315 PR tree-optimization/82604
8316 * tree-loop-distribution.c (enum partition_kind): New enum item
8317 PKIND_PARTIAL_MEMSET.
8318 (partition_builtin_p): Support above new enum item.
8319 (generate_code_for_partition): Ditto.
8320 (compute_access_range): Differentiate cases that equality can be
8321 proven at all loops, the innermost loops or no loops.
8322 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8323 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8324 (finalize_partitions, distribute_loop): Don't fuse partition of
8325 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8326 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8329 2018-01-23 Martin Liska <mliska@suse.cz>
8331 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8332 order to ignore the predictor.
8333 (PRED_POLYMORPHIC_CALL): Likewise.
8334 (PRED_RECURSIVE_CALL): Likewise.
8336 2018-01-23 Martin Liska <mliska@suse.cz>
8338 * tree-profile.c (tree_profiling): Print function header to
8339 aware reader which function we are working on.
8340 * value-prof.c (gimple_find_values_to_profile): Do not print
8341 not interesting value histograms.
8343 2018-01-23 Martin Liska <mliska@suse.cz>
8345 * profile-count.h (enum profile_quality): Add
8346 profile_uninitialized as the first value. Do not number values
8347 as they are zero based.
8348 (profile_count::verify): Update sanity check.
8349 (profile_probability::verify): Likewise.
8351 2018-01-23 Nathan Sidwell <nathan@acm.org>
8353 * doc/invoke.texi (ffor-scope): Deprecate.
8355 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8357 PR tree-optimization/83510
8358 * domwalk.c (set_all_edges_as_executable): New function.
8359 (dom_walker::dom_walker): Convert bool param
8360 "skip_unreachable_blocks" to enum reachability. Move setup of
8361 edge flags to set_all_edges_as_executable and only do it when
8362 reachability is REACHABLE_BLOCKS.
8363 * domwalk.h (enum dom_walker::reachability): New enum.
8364 (dom_walker::dom_walker): Convert bool param
8365 "skip_unreachable_blocks" to enum reachability.
8366 (set_all_edges_as_executable): New decl.
8367 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8368 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8370 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8371 but converting true to REACHABLE_BLOCKS.
8372 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8374 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8375 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8376 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8378 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8379 if check_all_array_refs will be called.
8381 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8383 * tree.c (selftest::test_location_wrappers): Add more test
8386 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8388 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8389 (selftest::test_bit_in_range): Likewise.
8391 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8394 * doc/sourcebuild.texi (vect_float): Say that the selector
8395 only describes the situation when -funsafe-math-optimizations is on.
8396 (vect_float_strict): Document.
8398 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8400 PR tree-optimization/83965
8401 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8402 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8403 instead of checking only for a reduction.
8404 (vect_recog_widen_sum_pattern): Likewise.
8406 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8408 * predict.c (probably_never_executed): Only use precise profile info.
8409 (compute_function_frequency): Skip after inlining hack since we now
8410 have quality checking.
8412 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8414 * profile-count.h (profile_probability::very_unlikely,
8415 profile_probability::unlikely, profile_probability::even): Set
8416 precision to guessed.
8418 2018-01-23 Richard Biener <rguenther@suse.de>
8420 PR tree-optimization/83963
8421 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8422 Properly terminate dominator walk when crossing the exit edge not
8423 when visiting its source block.
8425 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8428 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8429 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8431 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8433 PR tree-optimization/83957
8434 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8435 semicolon after for body surrounded by braces.
8437 PR tree-optimization/83081
8438 * profile-count.h (profile_probability::split): New method.
8439 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8440 Use profile_probability::split.
8441 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8442 when splitting a single conditional jump into 2.
8444 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8446 PR tree-optimization/69452
8447 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8450 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8452 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8453 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8454 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8456 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8458 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8459 * config/rl78/rl78.md (movdi): New define_expand.
8460 * config/rl78/rl78.c (rl78_split_movdi): New function.
8462 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8465 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8467 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8468 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8469 128-bit to produce an UNSPEC move to get the double word with the
8470 signbit and then a shift directly to do signbit.
8471 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8472 implementation with a new version that just does either a direct
8473 move or a regular move. Move memory interface to separate insns.
8474 Move insns so they are next to the expander.
8475 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8476 with signbit move. Split big and little endian case.
8477 (signbit<mode>2_dm_mem_le): Likewise.
8478 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8479 (signbit<mode>2_dm2): Likewise.
8481 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8483 * config/rl78/rl78.md (anddi3): New define_expand.
8485 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8487 * config/rl78/rl78.md (umindi3): New define_expand.
8489 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8491 * config/rl78/rl78.md (smindi3): New define_expand.
8493 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8495 * config/rl78/rl78.md (smaxdi3): New define_expand.
8497 2018-01-22 Carl Love <cel@us.ibm.com>
8499 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8500 LVX_V1TI): Add macro expansion.
8501 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8502 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8503 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8504 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8505 Change check to determine if the instruction is a byte reversing
8506 entry. Fix typo in comment.
8507 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8508 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8509 Add def_builtin calls for new builtins.
8510 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8511 Add define_insn expansion.
8513 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8515 * config/rl78/rl78.md (umaxdi3): New define_expand.
8517 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8519 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8520 for non-QImode registers.
8522 2018-01-22 Richard Biener <rguenther@suse.de>
8524 PR tree-optimization/83963
8525 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8526 including the loop exit block.
8527 (scop_detection::merge_sese): Likewise.
8528 (scop_detection::add_scop): Do it here instead.
8530 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8532 * doc/sourcebuild.texi (arm_softfloat): Document.
8534 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8537 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8538 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8539 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8541 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8542 David Edelsohn <dje.gcc@gmail.com>
8545 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8546 Change "crset eq" to "crset 2".
8547 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8548 (*call_indirect_aix<mode>_nospec): Likewise.
8549 (*call_value_indirect_aix<mode>_nospec): Likewise.
8550 (*call_indirect_elfv2<mode>_nospec): Likewise.
8551 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8552 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8553 change assembly output from . to $.
8554 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8555 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8556 (*tablejump<mode>_internal1_nospec): Likewise.
8558 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8561 * config/sh/sh_optimize_sett_clrt.cc:
8562 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8564 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8566 PR tree-optimization/83940
8567 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8568 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8569 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8570 use it to pass back the definition type.
8571 (vect_check_store_rhs): Likewise rhs_dt_out.
8572 (vect_build_gather_load_calls): Add a mask_dt argument and use
8573 it instead of a call to vect_is_simple_use.
8574 (vectorizable_store): Update calls to vect_check_load_store_mask
8575 and vect_check_store_rhs. Use the dt returned by the latter instead
8576 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8577 instead of calls to vect_is_simple_use. Pass the scalar rather
8578 than the vector operand to vect_is_simple_use when handling
8579 second and subsequent copies of an rhs value.
8580 (vectorizable_load): Update calls to vect_check_load_store_mask
8581 and vect_build_gather_load_calls. Use the cached mask_dt and
8582 gs_info.offset_dt instead of calls to vect_is_simple_use.
8584 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8587 * tree-emutls.c: Include gimplify.h.
8588 (lower_emutls_2): New function.
8589 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8590 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8591 it before further processing.
8594 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8595 UINTVAL (trueop1) instead of INTVAL (op1).
8597 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8601 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8602 INCOMING_FRAME_SP_OFFSET if not defined.
8603 (scan_trace): Add ENTRY argument. If true and
8604 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8605 emit a note to adjust the CFA offset.
8606 (create_cfi_notes): Adjust scan_trace callers.
8607 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8608 INCOMING_FRAME_SP_OFFSET in the CIE.
8609 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8610 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8612 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8613 * doc/tm.texi: Regenerated.
8615 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8617 PR rtl-optimization/83147
8618 * lra-constraints.c (remove_inheritance_pseudos): Use
8619 lra_substitute_pseudo_within_insn.
8621 2018-01-19 Tom de Vries <tom@codesourcery.com>
8622 Cesar Philippidis <cesar@codesourcery.com>
8625 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8627 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8630 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8631 spaces for function labels.
8633 2018-01-19 Martin Liska <mliska@suse.cz>
8635 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8636 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8637 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8638 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8639 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8640 (PRED_CONST_RETURN): Change from 69 to 65.
8641 (PRED_NULL_RETURN): Change from 91 to 71.
8642 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8643 (PRED_LOOP_GUARD): Change from 66 to 73.
8645 2018-01-19 Martin Liska <mliska@suse.cz>
8647 * predict.c (predict_insn_def): Add new assert.
8648 (struct branch_predictor): Change type to signed integer.
8649 (test_prediction_value_range): Amend test to cover
8651 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8652 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8653 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8654 (PRED_LOOP_IV_COMPARE): Likewise.
8655 * predict.h (PROB_UNINITIALIZED): Define new constant.
8657 2018-01-19 Martin Liska <mliska@suse.cz>
8659 * predict.c (dump_prediction): Add new format for
8660 analyze_brprob.py script which is enabled with -details
8662 * profile-count.h (precise_p): New function.
8664 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8666 PR tree-optimization/83922
8667 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8668 there are no statements that need masking.
8669 (vect_active_double_reduction_p): New function.
8670 (vect_analyze_loop_operations): Use it when handling phis that
8671 are not in the loop header.
8673 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8675 PR tree-optimization/83914
8676 * tree-vect-loop.c (vectorizable_induction): Don't convert
8677 init_expr or apply the peeling adjustment for inductions
8678 that are nested within the vectorized loop.
8680 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8682 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8685 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8689 * function.h (gimplify_parameters): Add gimple_seq * argument.
8690 * function.c: Include gimple.h and options.h.
8691 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8692 for the added local temporaries if needed.
8693 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8694 if there are any parameter cleanups, wrap whole body into a
8695 try/finally with the cleanups.
8697 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8700 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8701 Use GET_MODE_CLASS for scalar floating point.
8703 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8707 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8708 Fix call of call_cgraph_insertion_hooks.
8710 2018-01-18 Martin Sebor <msebor@redhat.com>
8712 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8714 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8717 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8720 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8723 * common.opt: (-ffile-prefix-map): New option.
8724 * opts.c (common_handle_option): Defer it.
8725 * opts-global.c (handle_common_deferred_options): Handle it.
8726 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8727 * file-prefix-map.h: New file.
8728 (remap_debug_filename, add_debug_prefix_map): ...here.
8729 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8730 * final.c (debug_prefix_map, add_debug_prefix_map
8731 remap_debug_filename): Move to...
8732 * file-prefix-map.c: New file.
8733 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8734 generalize, get rid of alloca(), use strrchr() instead of strchr().
8735 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8736 Implement in terms of add_prefix_map().
8737 (remap_macro_filename, remap_debug_filename): Implement in term of
8739 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8740 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8741 * dbxout.c: Include file-prefix-map.h.
8742 * varasm.c: Likewise.
8743 * vmsdbgout.c: Likewise.
8744 * xcoffout.c: Likewise.
8745 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8746 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8747 * doc/invoke.texi (-ffile-prefix-map): Document.
8748 (-fdebug-prefix-map): Update description.
8750 2018-01-18 Martin Liska <mliska@suse.cz>
8752 * config/i386/i386.c (indirect_thunk_name): Document that also
8754 (output_indirect_thunk): Document why both instructions
8755 (pause and lfence) are generated.
8757 2018-01-18 Richard Biener <rguenther@suse.de>
8759 PR tree-optimization/83887
8760 * graphite-scop-detection.c
8761 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8762 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8763 (scop_detection::merge_sese): Re-implement with a flood-fill
8764 algorithm that properly finds a SESE region if it exists.
8766 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8769 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8770 pointer_diff optimizations use view_convert instead of convert.
8772 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8774 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8775 Generate different code for -mno-speculate-indirect-jumps.
8776 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8777 (*call_indirect_aix<mode>): Disable for
8778 -mno-speculate-indirect-jumps.
8779 (*call_indirect_aix<mode>_nospec): New define_insn.
8780 (*call_value_indirect_aix<mode>): Disable for
8781 -mno-speculate-indirect-jumps.
8782 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8783 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8784 -mno-speculate-indirect-jumps.
8785 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8787 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8789 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8790 long double type, set the flags for noting the default long double
8791 type, even if we don't pass or return a long double type.
8793 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8796 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8799 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8801 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8802 support for merge[hl].
8803 (fold_mergehl_helper): New helper function.
8804 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8805 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8806 (altivec_vmrglw_direct): Add xxmrglw insn.
8808 2018-01-17 Andrew Waterman <andrew@sifive.com>
8810 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8811 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8813 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8816 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8817 call the lto_location_cache before reading the
8818 DECL_SOURCE_LOCATION of the types.
8820 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8821 Richard Sandiford <richard.sandiford@linaro.org>
8823 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8824 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8825 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8826 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8827 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8829 * config/aarch64/constraints.md (aarch64_movti_operand):
8831 * config/aarch64/predicates.md (Uti): Add new constraint.
8833 2018-01-17 Carl Love <cel@us.ibm.com>
8835 * config/rs6000/vsx.md (define_expand xl_len_r,
8836 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8837 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8839 (define_expand, define_insn): Move the shift left from the
8840 define_insn to the define_expand for lxvl and stxvl instructions.
8841 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8842 and XL_LEN_R definitions to PURE.
8844 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8846 * config/i386/i386.c (indirect_thunk_name): Declare regno
8847 as unsigned int. Compare regno with INVALID_REGNUM.
8848 (output_indirect_thunk): Ditto.
8849 (output_indirect_thunk_function): Ditto.
8850 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8851 in the call to output_indirect_thunk_function.
8853 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8856 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8857 rather than the size of inner_type to determine the stack slot size
8858 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8860 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8863 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8866 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8868 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8869 endian Linux systems to optionally enable multilibs for selecting
8870 the long double type if the user configured an explicit type.
8871 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8872 have no long double multilibs if not defined.
8873 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8874 warn if the user used -mabi={ieee,ibm}longdouble and we built
8875 multilibs for long double.
8876 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8877 appropriate multilib option.
8878 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8880 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8881 for building long double multilibs.
8882 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8884 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8886 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8889 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8891 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8894 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8897 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8900 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8902 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8903 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8905 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8907 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8908 different rtl trees depending on TARGET_64BIT.
8909 (rs6000_gen_lvx): Likewise.
8911 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8913 * config/visium/visium.md (nop): Tweak comment.
8914 (hazard_nop): Likewise.
8916 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8918 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8919 -mspeculate-indirect-jumps.
8920 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8921 for -mno-speculate-indirect-jumps.
8922 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8923 (*call_value_indirect_elfv2<mode>): Disable for
8924 -mno-speculate-indirect-jumps.
8925 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8926 (indirect_jump): Emit different RTL for
8927 -mno-speculate-indirect-jumps.
8928 (*indirect_jump<mode>): Disable for
8929 -mno-speculate-indirect-jumps.
8930 (*indirect_jump<mode>_nospec): New define_insn.
8931 (tablejump): Emit different RTL for
8932 -mno-speculate-indirect-jumps.
8933 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8934 (tablejumpsi_nospec): New define_expand.
8935 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8936 (tablejumpdi_nospec): New define_expand.
8937 (*tablejump<mode>_internal1): Disable for
8938 -mno-speculate-indirect-jumps.
8939 (*tablejump<mode>_internal1_nospec): New define_insn.
8940 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8943 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8945 * caller-save.c (insert_save): Drop unnecessary parameter. All
8948 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8949 Richard Biener <rguenth@suse.de>
8952 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8953 return early, inline manually is_gimple_sizepos. Make sure if we
8954 call gimplify_expr we don't end up with a gimple constant.
8955 * tree.c (variably_modified_type_p): Don't return true for
8956 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8957 * gimplify.h (is_gimple_sizepos): Remove.
8959 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8961 PR tree-optimization/83857
8962 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8963 vectorizable_live_operation for pure SLP statements.
8964 (vectorizable_live_operation): Handle PHIs.
8966 2018-01-16 Richard Biener <rguenther@suse.de>
8968 PR tree-optimization/83867
8969 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8970 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8972 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8975 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8976 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8977 If off is not INTEGER_CST, issue a may not be aligned warning
8978 rather than isn't aligned. Use isn%'t rather than isn't.
8979 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8981 <case MULT_EXPR>: Improve the case when bottom and one of the
8982 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8983 operand, in that case check if the other operand is multiple of
8984 bottom divided by the INTEGER_CST operand.
8986 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8989 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8990 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8991 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8992 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8993 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8994 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8995 (pa_function_arg_size): New function.
8997 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8999 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9000 in a separate statement.
9002 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9004 PR tree-optimization/83847
9005 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9006 group gathers and scatters.
9008 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9010 PR rtl-optimization/86620
9011 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9013 PR rtl-optimization/83213
9014 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9015 to last if both are JUMP_INSNs.
9017 PR tree-optimization/83843
9018 * gimple-ssa-store-merging.c
9019 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9020 store_immediate_info for bswap/nop orig_stores.
9022 2018-01-15 Andrew Waterman <andrew@sifive.com>
9024 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9026 <UDIV>: Increase cost if !TARGET_DIV.
9028 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9030 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9031 (define_attr "cr_logical_3op"): New.
9032 (cceq_ior_compare): Adjust.
9033 (cceq_ior_compare_complement): Adjust.
9034 (*cceq_rev_compare): Adjust.
9035 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9036 (is_cracked_insn): Adjust.
9037 (insn_must_be_first_in_group): Adjust.
9038 * config/rs6000/40x.md: Adjust.
9039 * config/rs6000/440.md: Adjust.
9040 * config/rs6000/476.md: Adjust.
9041 * config/rs6000/601.md: Adjust.
9042 * config/rs6000/603.md: Adjust.
9043 * config/rs6000/6xx.md: Adjust.
9044 * config/rs6000/7450.md: Adjust.
9045 * config/rs6000/7xx.md: Adjust.
9046 * config/rs6000/8540.md: Adjust.
9047 * config/rs6000/cell.md: Adjust.
9048 * config/rs6000/e300c2c3.md: Adjust.
9049 * config/rs6000/e500mc.md: Adjust.
9050 * config/rs6000/e500mc64.md: Adjust.
9051 * config/rs6000/e5500.md: Adjust.
9052 * config/rs6000/e6500.md: Adjust.
9053 * config/rs6000/mpc.md: Adjust.
9054 * config/rs6000/power4.md: Adjust.
9055 * config/rs6000/power5.md: Adjust.
9056 * config/rs6000/power6.md: Adjust.
9057 * config/rs6000/power7.md: Adjust.
9058 * config/rs6000/power8.md: Adjust.
9059 * config/rs6000/power9.md: Adjust.
9060 * config/rs6000/rs64.md: Adjust.
9061 * config/rs6000/titan.md: Adjust.
9063 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9065 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9066 ix86_indirect_branch_register logic.
9068 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9070 * config/i386/constraints.md (Bs): Update
9071 ix86_indirect_branch_register check. Don't check
9072 ix86_indirect_branch_register with GOT_memory_operand.
9074 * config/i386/predicates.md (GOT_memory_operand): Don't check
9075 ix86_indirect_branch_register here.
9076 (GOT32_symbol_operand): Likewise.
9078 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9080 * config/i386/predicates.md (constant_call_address_operand):
9081 Rewrite ix86_indirect_branch_register logic.
9082 (sibcall_insn_operand): Likewise.
9084 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9086 * config/i386/constraints.md (Bs): Replace
9087 ix86_indirect_branch_thunk_register with
9088 ix86_indirect_branch_register.
9090 * config/i386/i386.md (indirect_jump): Likewise.
9091 (tablejump): Likewise.
9092 (*sibcall_memory): Likewise.
9093 (*sibcall_value_memory): Likewise.
9094 Peepholes of indirect call and jump via memory: Likewise.
9095 * config/i386/i386.opt: Likewise.
9096 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9097 (GOT_memory_operand): Likewise.
9098 (call_insn_operand): Likewise.
9099 (sibcall_insn_operand): Likewise.
9100 (GOT32_symbol_operand): Likewise.
9102 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9105 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9106 type rather than type addr's type points to.
9107 (expand_omp_atomic_mutex): Likewise.
9108 (expand_omp_atomic): Likewise.
9110 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9113 * config/i386/i386.c (output_indirect_thunk_function): Use
9114 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9115 for __x86_return_thunk.
9117 2018-01-15 Richard Biener <rguenther@suse.de>
9120 * expmed.c (extract_bit_field_1): Fix typo.
9122 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9125 * config/arm/iterators.md (VF): New mode iterator.
9126 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9127 Remove integer-related logic from pattern.
9128 (neon_vabd<mode>_3): Likewise.
9130 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9133 * common.opt (fstrict-overflow): No longer an alias.
9134 (fwrapv-pointer): New option.
9135 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9136 also for pointer types based on flag_wrapv_pointer.
9137 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9138 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9139 opts->x_flag_wrapv got set.
9140 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9141 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9142 POINTER_TYPE_OVERFLOW_UNDEFINED.
9143 * match.pd: Likewise in address comparison pattern.
9144 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9146 2018-01-15 Richard Biener <rguenther@suse.de>
9149 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9150 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9151 Reset type names to their identifier if their TYPE_DECL doesn't
9152 have linkage (and thus is used for ODR and devirt).
9153 (save_debug_info_for_decl): Remove.
9154 (save_debug_info_for_type): Likewise.
9155 (add_tree_to_fld_list): Adjust.
9156 * tree-pretty-print.c (dump_generic_node): Make dumping of
9157 type names more robust.
9159 2018-01-15 Richard Biener <rguenther@suse.de>
9161 * BASE-VER: Bump to 8.0.1.
9163 2018-01-14 Martin Sebor <msebor@redhat.com>
9166 * builtins.c (check_access): Avoid warning when the no-warning bit
9169 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9171 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9172 * ira-color (allocno_hard_regs_compare): Likewise.
9174 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9177 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9178 Use .pushsection/.popsection.
9180 2018-01-14 Martin Sebor <msebor@redhat.com>
9183 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9185 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9187 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9188 entry from extra_headers.
9189 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9190 extra_headers, make the list bitwise identical to the i?86-*-* one.
9192 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9194 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9195 -mcmodel=large with -mindirect-branch=thunk,
9196 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9197 -mfunction-return=thunk-extern.
9198 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9199 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9200 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9202 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9204 * config/i386/i386.c (print_reg): Print the name of the full
9205 integer register without '%'.
9206 (ix86_print_operand): Handle 'V'.
9207 * doc/extend.texi: Document 'V' modifier.
9209 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9211 * config/i386/constraints.md (Bs): Disallow memory operand for
9212 -mindirect-branch-register.
9214 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9215 (GOT_memory_operand): Likewise.
9216 (call_insn_operand): Likewise.
9217 (sibcall_insn_operand): Likewise.
9218 (GOT32_symbol_operand): Likewise.
9219 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9220 for -mindirect-branch-register.
9221 (tablejump): Likewise.
9222 (*sibcall_memory): Likewise.
9223 (*sibcall_value_memory): Likewise.
9224 Disallow peepholes of indirect call and jump via memory for
9225 -mindirect-branch-register.
9226 (*call_pop): Replace m with Bw.
9227 (*call_value_pop): Likewise.
9228 (*sibcall_pop_memory): Replace m with Bs.
9229 * config/i386/i386.opt (mindirect-branch-register): New option.
9230 * doc/invoke.texi: Document -mindirect-branch-register option.
9232 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9234 * config/i386/i386-protos.h (ix86_output_function_return): New.
9235 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9236 set function_return_type.
9237 (indirect_thunk_name): Add ret_p to indicate thunk for function
9239 (output_indirect_thunk_function): Pass false to
9240 indirect_thunk_name.
9241 (ix86_output_indirect_branch_via_reg): Likewise.
9242 (ix86_output_indirect_branch_via_push): Likewise.
9243 (output_indirect_thunk_function): Create alias for function
9244 return thunk if regno < 0.
9245 (ix86_output_function_return): New function.
9246 (ix86_handle_fndecl_attribute): Handle function_return.
9247 (ix86_attribute_table): Add function_return.
9248 * config/i386/i386.h (machine_function): Add
9249 function_return_type.
9250 * config/i386/i386.md (simple_return_internal): Use
9251 ix86_output_function_return.
9252 (simple_return_internal_long): Likewise.
9253 * config/i386/i386.opt (mfunction-return=): New option.
9254 (indirect_branch): Mention -mfunction-return=.
9255 * doc/extend.texi: Document function_return function attribute.
9256 * doc/invoke.texi: Document -mfunction-return= option.
9258 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9260 * config/i386/i386-opts.h (indirect_branch): New.
9261 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9262 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9263 with local indirect jump when converting indirect call and jump.
9264 (ix86_set_indirect_branch_type): New.
9265 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9266 (indirectlabelno): New.
9267 (indirect_thunk_needed): Likewise.
9268 (indirect_thunk_bnd_needed): Likewise.
9269 (indirect_thunks_used): Likewise.
9270 (indirect_thunks_bnd_used): Likewise.
9271 (INDIRECT_LABEL): Likewise.
9272 (indirect_thunk_name): Likewise.
9273 (output_indirect_thunk): Likewise.
9274 (output_indirect_thunk_function): Likewise.
9275 (ix86_output_indirect_branch_via_reg): Likewise.
9276 (ix86_output_indirect_branch_via_push): Likewise.
9277 (ix86_output_indirect_branch): Likewise.
9278 (ix86_output_indirect_jmp): Likewise.
9279 (ix86_code_end): Call output_indirect_thunk_function if needed.
9280 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9282 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9283 (ix86_attribute_table): Add indirect_branch.
9284 * config/i386/i386.h (machine_function): Add indirect_branch_type
9285 and has_local_indirect_jump.
9286 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9288 (tablejump): Likewise.
9289 (*indirect_jump): Use ix86_output_indirect_jmp.
9290 (*tablejump_1): Likewise.
9291 (simple_return_indirect_internal): Likewise.
9292 * config/i386/i386.opt (mindirect-branch=): New option.
9293 (indirect_branch): New.
9296 (thunk-inline): Likewise.
9297 (thunk-extern): Likewise.
9298 * doc/extend.texi: Document indirect_branch function attribute.
9299 * doc/invoke.texi: Document -mindirect-branch= option.
9301 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9304 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9306 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9308 * ipa-inline.c (want_inline_small_function_p): Return false if
9309 inlining has already failed with CIF_FINAL_ERROR.
9310 (update_caller_keys): Call want_inline_small_function_p before
9312 (update_callee_keys): Likewise.
9314 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9316 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9318 (rs6000_quadword_masked_address_p): Likewise.
9319 (quad_aligned_load_p): Likewise.
9320 (quad_aligned_store_p): Likewise.
9321 (const_load_sequence_p): Add comment to describe the outer-most loop.
9322 (mimic_memory_attributes_and_flags): New function.
9323 (rs6000_gen_stvx): Likewise.
9324 (replace_swapped_aligned_store): Likewise.
9325 (rs6000_gen_lvx): Likewise.
9326 (replace_swapped_aligned_load): Likewise.
9327 (replace_swapped_load_constant): Capitalize argument name in
9328 comment describing this function.
9329 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9330 and stores that access quad-word aligned addresses and replace
9331 with stvx or lvx instructions when appropriate.
9332 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9333 New function prototype.
9334 (rs6000_quadword_masked_address_p): Likewise.
9335 (rs6000_gen_lvx): Likewise.
9336 (rs6000_gen_stvx): Likewise.
9337 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9338 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9339 when memory address is aligned.
9340 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9341 this split to select lvx instruction when memory address is aligned.
9342 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9343 instruction when memory address is aligned.
9344 (*vsx_le_perm_load_v16qi): Likewise.
9345 (four unnamed splitters): Modify to select the stvx instruction
9346 when memory is aligned.
9348 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9350 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9351 which appears in the queue multiple times.
9353 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9354 Alan Hayward <alan.hayward@arm.com>
9355 David Sherwood <david.sherwood@arm.com>
9357 * tree-vectorizer.h (vec_lower_bound): New structure.
9358 (_loop_vec_info): Add check_nonzero and lower_bounds.
9359 (LOOP_VINFO_CHECK_NONZERO): New macro.
9360 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9361 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9362 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9363 fields. Make seg_len the distance travelled, not including the
9365 (dr_direction_indicator): Declare.
9366 (dr_zero_step_indicator): Likewise.
9367 (dr_known_forward_stride_p): Likewise.
9368 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9370 (runtime_alias_check_p): Allow runtime alias checks with
9372 (operator ==): Compare access_size and align.
9373 (prune_runtime_alias_test_list): Rework for new distinction between
9374 the access_size and seg_len.
9375 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9377 (get_segment_min_max): New function.
9378 (create_intersect_range_checks): Use it.
9379 (dr_step_indicator): New function.
9380 (dr_direction_indicator): Likewise.
9381 (dr_zero_step_indicator): Likewise.
9382 (dr_known_forward_stride_p): Likewise.
9383 * tree-loop-distribution.c (data_ref_segment_size): Return
9384 DR_STEP * (niters - 1).
9385 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9387 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9388 (vect_preserves_scalar_order_p): New function, split out from...
9389 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9390 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9391 (vect_vfa_access_size): New function.
9392 (vect_vfa_align): Likewise.
9393 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9394 (dump_lower_bound): New function.
9395 (vect_check_lower_bound): Likewise.
9396 (vect_small_gap_p): Likewise.
9397 (vectorizable_with_step_bound_p): Likewise.
9398 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9399 depencies if the vectorization factor is 1. Convert the checks
9400 for nonzero steps into checks on the bounds of DR_STEP. Try using
9401 a bunds check for variable steps if the minimum required step is
9402 relatively small. Update calls to the dr_with_seg_len
9403 constructor and to vect_compile_time_alias.
9404 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9406 (vect_loop_versioning): Call it.
9407 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9409 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9411 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9412 Alan Hayward <alan.hayward@arm.com>
9413 David Sherwood <david.sherwood@arm.com>
9415 * doc/sourcebuild.texi (vect_scatter_store): Document.
9416 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9418 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9420 * genopinit.c (main): Add supports_vec_scatter_store and
9421 supports_vec_scatter_store_cached to target_optabs.
9422 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9423 IFN_MASK_SCATTER_STORE.
9424 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9426 * internal-fn.h (internal_store_fn_p): Declare.
9427 (internal_fn_stored_value_index): Likewise.
9428 * internal-fn.c (scatter_store_direct): New macro.
9429 (expand_scatter_store_optab_fn): New function.
9430 (direct_scatter_store_optab_supported_p): New macro.
9431 (internal_store_fn_p): New function.
9432 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9433 IFN_MASK_SCATTER_STORE.
9434 (internal_fn_mask_index): Likewise.
9435 (internal_fn_stored_value_index): New function.
9436 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9438 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9439 * optabs-query.c (supports_vec_scatter_store_p): New function.
9440 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9441 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9442 true for scatter stores.
9443 (vect_gather_scatter_fn_p): Handle scatter stores too.
9444 (vect_check_gather_scatter): Consider using scatter stores if
9445 supports_vec_scatter_store_p.
9446 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9448 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9449 internal_fn_stored_value_index.
9450 (check_load_store_masking): Handle scatter stores too.
9451 (vect_get_store_rhs): Make public.
9452 (vectorizable_call): Use internal_store_fn_p.
9453 (vectorizable_store): Handle scatter store internal functions.
9454 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9455 when deciding whether the end of the group has been reached.
9456 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9457 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9458 (mask_scatter_store<mode>): New insns.
9460 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9461 Alan Hayward <alan.hayward@arm.com>
9462 David Sherwood <david.sherwood@arm.com>
9464 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9465 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9466 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9468 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9469 Use vect_truncate_gather_scatter_offset if we can't treat the
9470 operation as a normal gather load or scatter store.
9471 (get_group_load_store_type): Take the gather_scatter_info
9472 as argument. Try using a gather load or scatter store for
9473 single-element groups.
9474 (get_load_store_type): Update calls to get_group_load_store_type
9475 and vect_use_strided_gather_scatters_p.
9477 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9478 Alan Hayward <alan.hayward@arm.com>
9479 David Sherwood <david.sherwood@arm.com>
9481 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9482 optional tree argument.
9483 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9485 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9486 but continue to use the current value as a fallback.
9487 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9488 to compare the updates.
9489 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9490 (get_load_store_type): Use it when handling a strided access.
9491 (vect_get_strided_load_store_ops): New function.
9492 (vect_get_data_ptr_increment): Likewise.
9493 (vectorizable_load): Handle strided gather loads. Always pass
9494 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9496 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9497 Alan Hayward <alan.hayward@arm.com>
9498 David Sherwood <david.sherwood@arm.com>
9500 * doc/md.texi (gather_load@var{m}): Document.
9501 (mask_gather_load@var{m}): Likewise.
9502 * genopinit.c (main): Add supports_vec_gather_load and
9503 supports_vec_gather_load_cached to target_optabs.
9504 * optabs-tree.c (init_tree_optimization_optabs): Use
9505 ggc_cleared_alloc to allocate target_optabs.
9506 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9507 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9509 * internal-fn.h (internal_load_fn_p): Declare.
9510 (internal_gather_scatter_fn_p): Likewise.
9511 (internal_fn_mask_index): Likewise.
9512 (internal_gather_scatter_fn_supported_p): Likewise.
9513 * internal-fn.c (gather_load_direct): New macro.
9514 (expand_gather_load_optab_fn): New function.
9515 (direct_gather_load_optab_supported_p): New macro.
9516 (direct_internal_fn_optab): New function.
9517 (internal_load_fn_p): Likewise.
9518 (internal_gather_scatter_fn_p): Likewise.
9519 (internal_fn_mask_index): Likewise.
9520 (internal_gather_scatter_fn_supported_p): Likewise.
9521 * optabs-query.c (supports_at_least_one_mode_p): New function.
9522 (supports_vec_gather_load_p): Likewise.
9523 * optabs-query.h (supports_vec_gather_load_p): Declare.
9524 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9525 and memory_type field.
9526 (NUM_PATTERNS): Bump to 15.
9527 * tree-vect-data-refs.c: Include internal-fn.h.
9528 (vect_gather_scatter_fn_p): New function.
9529 (vect_describe_gather_scatter_call): Likewise.
9530 (vect_check_gather_scatter): Try using internal functions for
9531 gather loads. Recognize existing calls to a gather load function.
9532 (vect_analyze_data_refs): Consider using gather loads if
9533 supports_vec_gather_load_p.
9534 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9535 (vect_get_gather_scatter_offset_type): Likewise.
9536 (vect_convert_mask_for_vectype): Likewise.
9537 (vect_add_conversion_to_patterm): Likewise.
9538 (vect_try_gather_scatter_pattern): Likewise.
9539 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9540 (vect_vect_recog_func_ptrs): Add it.
9541 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9542 internal_fn_mask_index and internal_gather_scatter_fn_p.
9543 (check_load_store_masking): Take the gather_scatter_info as an
9544 argument and handle gather loads.
9545 (vect_get_gather_scatter_ops): New function.
9546 (vectorizable_call): Check internal_load_fn_p.
9547 (vectorizable_load): Likewise. Handle gather load internal
9549 (vectorizable_store): Update call to check_load_store_masking.
9550 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9551 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9552 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9553 (aarch64_gather_scale_operand_d): New predicates.
9554 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9555 (mask_gather_load<mode>): New insns.
9557 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9558 Alan Hayward <alan.hayward@arm.com>
9559 David Sherwood <david.sherwood@arm.com>
9561 * optabs.def (fold_left_plus_optab): New optab.
9562 * doc/md.texi (fold_left_plus_@var{m}): Document.
9563 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9564 * internal-fn.c (fold_left_direct): Define.
9565 (expand_fold_left_optab_fn): Likewise.
9566 (direct_fold_left_optab_supported_p): Likewise.
9567 * fold-const-call.c (fold_const_fold_left): New function.
9568 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9569 * tree-parloops.c (valid_reduction_p): New function.
9570 (gather_scalar_reductions): Use it.
9571 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9572 (vect_finish_replace_stmt): Declare.
9573 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9574 (needs_fold_left_reduction_p): New function, split out from...
9575 (vect_is_simple_reduction): ...here. Accept reductions that
9576 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9577 (vect_force_simple_reduction): Also store the reduction type in
9578 the assignment's STMT_VINFO_REDUC_TYPE.
9579 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9580 (merge_with_identity): New function.
9581 (vect_expand_fold_left): Likewise.
9582 (vectorize_fold_left_reduction): Likewise.
9583 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9584 scalar phi in place for it. Check for target support and reject
9585 cases that would reassociate the operation. Defer the transform
9586 phase to vectorize_fold_left_reduction.
9587 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9588 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9589 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9591 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9593 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9594 call to ifc_temp_var.
9596 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9597 Alan Hayward <alan.hayward@arm.com>
9598 David Sherwood <david.sherwood@arm.com>
9600 * target.def (legitimize_address_displacement): Take the original
9601 offset as a poly_int.
9602 * targhooks.h (default_legitimize_address_displacement): Update
9604 * targhooks.c (default_legitimize_address_displacement): Likewise.
9605 * doc/tm.texi: Regenerate.
9606 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9607 as an argument, moving assert of ad->disp == ad->disp_term to...
9608 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9609 Try calling targetm.legitimize_address_displacement before expanding
9610 the address rather than afterwards, and adjust for the new interface.
9611 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9612 Match the new hook interface. Handle SVE addresses.
9613 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9616 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9618 * Makefile.in (OBJS): Add early-remat.o.
9619 * target.def (select_early_remat_modes): New hook.
9620 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9621 * doc/tm.texi: Regenerate.
9622 * targhooks.h (default_select_early_remat_modes): Declare.
9623 * targhooks.c (default_select_early_remat_modes): New function.
9624 * timevar.def (TV_EARLY_REMAT): New timevar.
9625 * passes.def (pass_early_remat): New pass.
9626 * tree-pass.h (make_pass_early_remat): Declare.
9627 * early-remat.c: New file.
9628 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9630 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9632 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9633 Alan Hayward <alan.hayward@arm.com>
9634 David Sherwood <david.sherwood@arm.com>
9636 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9637 vfm1 with a bound_epilog parameter.
9638 (vect_do_peeling): Update calls accordingly, and move the prologue
9639 call earlier in the function. Treat the base bound_epilog as 0 for
9640 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9641 this base when peeling for gaps.
9642 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9643 with fully-masked loops.
9644 (vect_estimate_min_profitable_iters): Handle the single peeled
9645 iteration in that case.
9647 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9648 Alan Hayward <alan.hayward@arm.com>
9649 David Sherwood <david.sherwood@arm.com>
9651 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9652 single-element interleaving even if the size is not a power of 2.
9653 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9654 accesses for single-element interleaving if the group size is
9657 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9658 Alan Hayward <alan.hayward@arm.com>
9659 David Sherwood <david.sherwood@arm.com>
9661 * doc/md.texi (fold_extract_last_@var{m}): Document.
9662 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9663 * optabs.def (fold_extract_last_optab): New optab.
9664 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9665 * internal-fn.c (fold_extract_direct): New macro.
9666 (expand_fold_extract_optab_fn): Likewise.
9667 (direct_fold_extract_optab_supported_p): Likewise.
9668 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9669 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9670 EXTRACT_LAST_REDUCTION.
9671 (get_initial_def_for_reduction): Do not create an initial vector
9672 for EXTRACT_LAST_REDUCTION reductions.
9673 (vectorizable_reduction): Leave the scalar phi in place for
9674 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9675 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9676 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9677 transform phase to vectorizable_condition.
9678 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9680 (vect_finish_stmt_generation): ...here.
9681 (vect_finish_replace_stmt): New function.
9682 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9683 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9685 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9687 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9688 Alan Hayward <alan.hayward@arm.com>
9689 David Sherwood <david.sherwood@arm.com>
9691 * doc/md.texi (extract_last_@var{m}): Document.
9692 * optabs.def (extract_last_optab): New optab.
9693 * internal-fn.def (EXTRACT_LAST): New internal function.
9694 * internal-fn.c (cond_unary_direct): New macro.
9695 (expand_cond_unary_optab_fn): Likewise.
9696 (direct_cond_unary_optab_supported_p): Likewise.
9697 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9698 loops using EXTRACT_LAST.
9699 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9700 (extract_last_<mode>): ...this optab.
9701 (vec_extract<mode><Vel>): Update accordingly.
9703 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9704 Alan Hayward <alan.hayward@arm.com>
9705 David Sherwood <david.sherwood@arm.com>
9707 * target.def (empty_mask_is_expensive): New hook.
9708 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9709 * doc/tm.texi: Regenerate.
9710 * targhooks.h (default_empty_mask_is_expensive): Declare.
9711 * targhooks.c (default_empty_mask_is_expensive): New function.
9712 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9713 if the target says that empty masks are expensive.
9714 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9716 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9718 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9719 Alan Hayward <alan.hayward@arm.com>
9720 David Sherwood <david.sherwood@arm.com>
9722 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9723 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9724 (vect_use_loop_mask_for_alignment_p): New function.
9725 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9726 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9727 niters_skip argument. Make sure that the first niters_skip elements
9728 of the first iteration are inactive.
9729 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9730 Update call to vect_set_loop_masks_directly.
9731 (get_misalign_in_elems): New function, split out from...
9732 (vect_gen_prolog_loop_niters): ...here.
9733 (vect_update_init_of_dr): Take a code argument that specifies whether
9734 the adjustment should be added or subtracted.
9735 (vect_update_init_of_drs): Likewise.
9736 (vect_prepare_for_masked_peels): New function.
9737 (vect_do_peeling): Skip prologue peeling if we're using a mask
9738 instead. Update call to vect_update_inits_of_drs.
9739 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9741 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9742 alignment. Do not include the number of peeled iterations in
9743 the minimum threshold in that case.
9744 (vectorizable_induction): Adjust the start value down by
9745 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9746 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9747 Take the number of skipped iterations into account when calculating
9749 * tree-vect-stmts.c (vect_gen_while_not): New function.
9751 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9752 Alan Hayward <alan.hayward@arm.com>
9753 David Sherwood <david.sherwood@arm.com>
9755 * doc/sourcebuild.texi (vect_fully_masked): Document.
9756 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9758 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9760 (vect_analyze_loop_2): ...here. Don't check the vectorization
9761 factor against the number of loop iterations if the loop is
9764 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9765 Alan Hayward <alan.hayward@arm.com>
9766 David Sherwood <david.sherwood@arm.com>
9768 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9769 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9770 (dump_groups): Update accordingly.
9771 (iv_use::mem_type): New member variable.
9772 (address_p): New function.
9773 (record_use): Add a mem_type argument and initialize the new
9775 (record_group_use): Add a mem_type argument. Use address_p.
9776 Remove obsolete null checks of base_object. Update call to record_use.
9777 (find_interesting_uses_op): Update call to record_group_use.
9778 (find_interesting_uses_cond): Likewise.
9779 (find_interesting_uses_address): Likewise.
9780 (get_mem_type_for_internal_fn): New function.
9781 (find_address_like_use): Likewise.
9782 (find_interesting_uses_stmt): Try find_address_like_use before
9783 calling find_interesting_uses_op.
9784 (addr_offset_valid_p): Use the iv mem_type field as the type
9785 of the addressed memory.
9786 (add_autoinc_candidates): Likewise.
9787 (get_address_cost): Likewise.
9788 (split_small_address_groups_p): Use address_p.
9789 (split_address_groups): Likewise.
9790 (add_iv_candidate_for_use): Likewise.
9791 (autoinc_possible_for_pair): Likewise.
9792 (rewrite_groups): Likewise.
9793 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9794 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9795 (get_alias_ptr_type_for_ptr_address): New function.
9796 (rewrite_use_address): Rewrite address uses in calls that were
9797 identified by find_address_like_use.
9799 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9800 Alan Hayward <alan.hayward@arm.com>
9801 David Sherwood <david.sherwood@arm.com>
9803 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9805 * gimple-expr.h (is_gimple_addressable: Likewise.
9806 * gimple-expr.c (is_gimple_address): Likewise.
9807 * internal-fn.c (expand_call_mem_ref): New function.
9808 (expand_mask_load_optab_fn): Use it.
9809 (expand_mask_store_optab_fn): Likewise.
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/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9816 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9817 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9818 (cond_umax@var{mode}): Document.
9819 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9820 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9821 (cond_umin_optab, cond_umax_optab): New optabs.
9822 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9823 (COND_IOR, COND_XOR): New internal functions.
9824 * internal-fn.h (get_conditional_internal_fn): Declare.
9825 * internal-fn.c (cond_binary_direct): New macro.
9826 (expand_cond_binary_optab_fn): Likewise.
9827 (direct_cond_binary_optab_supported_p): Likewise.
9828 (get_conditional_internal_fn): New function.
9829 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9830 Cope with reduction statements that are vectorized as calls rather
9832 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9833 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9834 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9835 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9836 (UNSPEC_COND_EOR): New unspecs.
9837 (optab): Add mappings for them.
9838 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9839 (sve_int_op, sve_fp_op): New int attributes.
9841 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9842 Alan Hayward <alan.hayward@arm.com>
9843 David Sherwood <david.sherwood@arm.com>
9845 * optabs.def (while_ult_optab): New optab.
9846 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9847 * internal-fn.def (WHILE_ULT): New internal function.
9848 * internal-fn.h (direct_internal_fn_supported_p): New override
9849 that takes two types as argument.
9850 * internal-fn.c (while_direct): New macro.
9851 (expand_while_optab_fn): New function.
9852 (convert_optab_supported_p): Likewise.
9853 (direct_while_optab_supported_p): New macro.
9854 * wide-int.h (wi::udiv_ceil): New function.
9855 * tree-vectorizer.h (rgroup_masks): New structure.
9856 (vec_loop_masks): New typedef.
9857 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9859 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9860 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9861 (vect_max_vf): New function.
9862 (slpeel_make_loop_iterate_ntimes): Delete.
9863 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9864 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9865 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9866 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9867 internal-fn.h, stor-layout.h and optabs-query.h.
9868 (vect_set_loop_mask): New function.
9869 (add_preheader_seq): Likewise.
9870 (add_header_seq): Likewise.
9871 (interleave_supported_p): Likewise.
9872 (vect_maybe_permute_loop_masks): Likewise.
9873 (vect_set_loop_masks_directly): Likewise.
9874 (vect_set_loop_condition_masked): Likewise.
9875 (vect_set_loop_condition_unmasked): New function, split out from
9876 slpeel_make_loop_iterate_ntimes.
9877 (slpeel_make_loop_iterate_ntimes): Rename to..
9878 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9879 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9880 (vect_do_peeling): Update call accordingly.
9881 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9883 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9884 mask_compare_type, can_fully_mask_p and fully_masked_p.
9885 (release_vec_loop_masks): New function.
9886 (_loop_vec_info): Use it to free the loop masks.
9887 (can_produce_all_loop_masks_p): New function.
9888 (vect_get_max_nscalars_per_iter): Likewise.
9889 (vect_verify_full_masking): Likewise.
9890 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9891 retries, and free the mask rgroups before retrying. Check loop-wide
9892 reasons for disallowing fully-masked loops. Make the final decision
9893 about whether use a fully-masked loop or not.
9894 (vect_estimate_min_profitable_iters): Do not assume that peeling
9895 for the number of iterations will be needed for fully-masked loops.
9896 (vectorizable_reduction): Disable fully-masked loops.
9897 (vectorizable_live_operation): Likewise.
9898 (vect_halve_mask_nunits): New function.
9899 (vect_double_mask_nunits): Likewise.
9900 (vect_record_loop_mask): Likewise.
9901 (vect_get_loop_mask): Likewise.
9902 (vect_transform_loop): Handle the case in which the final loop
9903 iteration might handle a partial vector. Call vect_set_loop_condition
9904 instead of slpeel_make_loop_iterate_ntimes.
9905 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9906 (check_load_store_masking): New function.
9907 (prepare_load_store_mask): Likewise.
9908 (vectorizable_store): Handle fully-masked loops.
9909 (vectorizable_load): Likewise.
9910 (supportable_widening_operation): Use vect_halve_mask_nunits for
9912 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9913 (vect_gen_while): New function.
9914 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9915 (aarch64_uqdec<mode>): New insn.
9917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9918 Alan Hayward <alan.hayward@arm.com>
9919 David Sherwood <david.sherwood@arm.com>
9921 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9922 (reduc_xor_scal_optab): New optabs.
9923 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9924 (reduc_xor_scal_@var{m}): Document.
9925 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9926 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9928 * fold-const-call.c (fold_const_call): Handle them.
9929 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9930 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9931 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9932 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9933 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9934 (UNSPEC_XORV): New unspecs.
9935 (optab): Add entries for them.
9936 (BITWISEV): New int iterator.
9937 (bit_reduc_op): New int attributes.
9939 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9940 Alan Hayward <alan.hayward@arm.com>
9941 David Sherwood <david.sherwood@arm.com>
9943 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9944 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9945 * optabs.def (vec_shl_insert_optab): New optab.
9946 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9947 (duplicate_and_interleave): Likewise.
9948 * tree-vect-loop.c: Include internal-fn.h.
9949 (neutral_op_for_slp_reduction): New function, split out from
9950 get_initial_defs_for_reduction.
9951 (get_initial_def_for_reduction): Handle option 2 for variable-length
9952 vectors by loading the neutral value into a vector and then shifting
9953 the initial value into element 0.
9954 (get_initial_defs_for_reduction): Replace the code argument with
9955 the neutral value calculated by neutral_op_for_slp_reduction.
9956 Use gimple_build_vector for constant-length vectors.
9957 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9958 but the first group_size elements have a neutral value.
9959 Use duplicate_and_interleave otherwise.
9960 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9961 Update call to get_initial_defs_for_reduction. Handle SLP
9962 reductions for variable-length vectors by creating one vector
9963 result for each scalar result, with the elements associated
9964 with other scalar results stubbed out with the neutral value.
9965 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9966 Require IFN_VEC_SHL_INSERT for double reductions on
9967 variable-length vectors, or SLP reductions that have
9968 a neutral value. Require can_duplicate_and_interleave_p
9969 support for variable-length unchained SLP reductions if there
9970 is no neutral value, such as for MIN/MAX reductions. Also require
9971 the number of vector elements to be a multiple of the number of
9972 SLP statements when doing variable-length unchained SLP reductions.
9973 Update call to vect_create_epilog_for_reduction.
9974 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9975 and remove initial values.
9976 (duplicate_and_interleave): Make public.
9977 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9978 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9980 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9981 Alan Hayward <alan.hayward@arm.com>
9982 David Sherwood <david.sherwood@arm.com>
9984 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9985 (can_duplicate_and_interleave_p): New function.
9986 (vect_get_and_check_slp_defs): Take the vector of statements
9987 rather than just the current one. Remove excess parentheses.
9988 Restriction rejectinon of vect_constant_def and vect_external_def
9989 for variable-length vectors to boolean types, or types for which
9990 can_duplicate_and_interleave_p is false.
9991 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9992 (duplicate_and_interleave): New function.
9993 (vect_get_constant_vectors): Use gimple_build_vector for
9994 constant-length vectors and suitable variable-length constant
9995 vectors. Use duplicate_and_interleave for other variable-length
9996 vectors. Don't defer the update when inserting new statements.
9998 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9999 Alan Hayward <alan.hayward@arm.com>
10000 David Sherwood <david.sherwood@arm.com>
10002 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10003 min_profitable_iters doesn't go negative.
10005 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10006 Alan Hayward <alan.hayward@arm.com>
10007 David Sherwood <david.sherwood@arm.com>
10009 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10010 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10011 * optabs.def (vec_mask_load_lanes_optab): New optab.
10012 (vec_mask_store_lanes_optab): Likewise.
10013 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10014 (MASK_STORE_LANES): Likewise.
10015 * internal-fn.c (mask_load_lanes_direct): New macro.
10016 (mask_store_lanes_direct): Likewise.
10017 (expand_mask_load_optab_fn): Handle masked operations.
10018 (expand_mask_load_lanes_optab_fn): New macro.
10019 (expand_mask_store_optab_fn): Handle masked operations.
10020 (expand_mask_store_lanes_optab_fn): New macro.
10021 (direct_mask_load_lanes_optab_supported_p): Likewise.
10022 (direct_mask_store_lanes_optab_supported_p): Likewise.
10023 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10025 (vect_load_lanes_supported): Likewise.
10026 * tree-vect-data-refs.c (strip_conversion): New function.
10027 (can_group_stmts_p): Likewise.
10028 (vect_analyze_data_ref_accesses): Use it instead of checking
10029 for a pair of assignments.
10030 (vect_store_lanes_supported): Take a masked_p parameter.
10031 (vect_load_lanes_supported): Likewise.
10032 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10033 vect_store_lanes_supported and vect_load_lanes_supported.
10034 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10035 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10036 parameter. Don't allow gaps for masked accesses.
10037 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10038 and vect_load_lanes_supported.
10039 (get_load_store_type): Take a masked_p parameter and update
10040 call to get_group_load_store_type.
10041 (vectorizable_store): Update call to get_load_store_type.
10042 Handle IFN_MASK_STORE_LANES.
10043 (vectorizable_load): Update call to get_load_store_type.
10044 Handle IFN_MASK_LOAD_LANES.
10046 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10047 Alan Hayward <alan.hayward@arm.com>
10048 David Sherwood <david.sherwood@arm.com>
10050 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10052 * config/aarch64/aarch64-protos.h
10053 (aarch64_sve_struct_memory_operand_p): Declare.
10054 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10055 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10056 (VPRED, vpred): Handle SVE structure modes.
10057 * config/aarch64/constraints.md (Utx): New constraint.
10058 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10059 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10060 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10061 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10062 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10063 structure modes. Split into pieces after RA.
10064 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10065 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10067 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10068 SVE structure modes.
10069 (aarch64_classify_address): Likewise.
10070 (sizetochar): Move earlier in file.
10071 (aarch64_print_operand): Handle SVE register lists.
10072 (aarch64_array_mode): New function.
10073 (aarch64_sve_struct_memory_operand_p): Likewise.
10074 (TARGET_ARRAY_MODE): Redefine.
10076 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10077 Alan Hayward <alan.hayward@arm.com>
10078 David Sherwood <david.sherwood@arm.com>
10080 * target.def (array_mode): New target hook.
10081 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10082 * doc/tm.texi: Regenerate.
10083 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10084 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10085 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10086 targetm.array_mode.
10087 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10090 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10091 Alan Hayward <alan.hayward@arm.com>
10092 David Sherwood <david.sherwood@arm.com>
10094 * fold-const.c (fold_binary_loc): Check the argument types
10095 rather than the result type when testing for a vector operation.
10097 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10099 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10100 * doc/tm.texi: Regenerate.
10102 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10103 Alan Hayward <alan.hayward@arm.com>
10104 David Sherwood <david.sherwood@arm.com>
10106 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10107 (sve): Document new AArch64 extension.
10108 * doc/md.texi (w): Extend the description of the AArch64
10109 constraint to include SVE vectors.
10110 (Upl, Upa): Document new AArch64 predicate constraints.
10111 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10113 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10114 (msve-vector-bits=): New option.
10115 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10116 SVE when these are disabled.
10117 (sve): New extension.
10118 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10119 modes. Adjust their number of units based on aarch64_sve_vg.
10120 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10121 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10122 aarch64_addr_query_type.
10123 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10124 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10125 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10126 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10127 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10128 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10129 (aarch64_simd_imm_zero_p): Delete.
10130 (aarch64_check_zero_based_sve_index_immediate): Declare.
10131 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10132 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10133 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10134 (aarch64_sve_float_mul_immediate_p): Likewise.
10135 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10136 rather than an rtx.
10137 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10138 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10139 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10140 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10141 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10142 (aarch64_regmode_natural_size): Likewise.
10143 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10144 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10146 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10147 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10148 for VG and the SVE predicate registers.
10149 (V_ALIASES): Add a "z"-prefixed alias.
10150 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10151 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10152 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10153 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10154 (REG_CLASS_NAMES): Add entries for them.
10155 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10156 and the predicate registers.
10157 (aarch64_sve_vg): Declare.
10158 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10159 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10160 (REGMODE_NATURAL_SIZE): Define.
10161 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10163 * config/aarch64/aarch64.c: Include cfgrtl.h.
10164 (simd_immediate_info): Add a constructor for series vectors,
10165 and an associated step field.
10166 (aarch64_sve_vg): New variable.
10167 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10168 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10169 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10170 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10171 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10172 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10173 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10174 (aarch64_get_mask_mode): New functions.
10175 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10176 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10177 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10178 predicate modes and predicate registers. Explicitly restrict
10179 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10180 to store a vector mode if it is recognized by
10181 aarch64_classify_vector_mode.
10182 (aarch64_regmode_natural_size): New function.
10183 (aarch64_hard_regno_caller_save_mode): Return the original mode
10185 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10186 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10187 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10188 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10190 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10191 does not overlap dest if the function is frame-related. Handle
10193 (aarch64_split_add_offset): New function.
10194 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10195 them aarch64_add_offset.
10196 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10197 and update call to aarch64_sub_sp.
10198 (aarch64_add_cfa_expression): New function.
10199 (aarch64_expand_prologue): Pass extra temporary registers to the
10200 functions above. Handle the case in which we need to emit new
10201 DW_CFA_expressions for registers that were originally saved
10202 relative to the stack pointer, but now have to be expressed
10203 relative to the frame pointer.
10204 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10206 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10207 IP0 and IP1 values for SVE frames.
10208 (aarch64_expand_vec_series): New function.
10209 (aarch64_expand_sve_widened_duplicate): Likewise.
10210 (aarch64_expand_sve_const_vector): Likewise.
10211 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10212 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10213 into the register, rather than emitting a SET directly.
10214 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10215 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10216 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10217 (offset_9bit_signed_scaled_p): New functions.
10218 (aarch64_replicate_bitmask_imm): New function.
10219 (aarch64_bitmask_imm): Use it.
10220 (aarch64_cannot_force_const_mem): Reject expressions involving
10221 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10222 (aarch64_classify_index): Handle SVE indices, by requiring
10223 a plain register index with a scale that matches the element size.
10224 (aarch64_classify_address): Handle SVE addresses. Assert that
10225 the mode of the address is VOIDmode or an integer mode.
10226 Update call to aarch64_classify_symbol.
10227 (aarch64_classify_symbolic_expression): Update call to
10228 aarch64_classify_symbol.
10229 (aarch64_const_vec_all_in_range_p): New function.
10230 (aarch64_print_vector_float_operand): Likewise.
10231 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10232 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10233 and the FP immediates 1.0 and 0.5.
10234 (aarch64_print_address_internal): Handle SVE addresses.
10235 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10236 (aarch64_regno_regclass): Handle predicate registers.
10237 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10239 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10240 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10241 (aarch64_convert_sve_vector_bits): New function.
10242 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10243 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10244 rather than an rtx.
10245 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10246 Handle SVE vector and predicate modes. Accept VL-based constants
10247 that need only one temporary register, and VL offsets that require
10248 no temporary registers.
10249 (aarch64_conditional_register_usage): Mark the predicate registers
10250 as fixed if SVE isn't available.
10251 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10252 Return true for SVE vector and predicate modes.
10253 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10254 rather than an unsigned int. Handle SVE modes.
10255 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10257 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10259 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10260 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10261 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10262 (aarch64_sve_float_mul_immediate_p): New functions.
10263 (aarch64_sve_valid_immediate): New function.
10264 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10265 Explicitly reject structure modes. Check for INDEX constants.
10266 Handle PTRUE and PFALSE constants.
10267 (aarch64_check_zero_based_sve_index_immediate): New function.
10268 (aarch64_simd_imm_zero_p): Delete.
10269 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10270 vector modes. Accept constants in the range of CNT[BHWD].
10271 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10272 ask for an Advanced SIMD mode.
10273 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10274 (aarch64_simd_vector_alignment): Handle SVE predicates.
10275 (aarch64_vectorize_preferred_vector_alignment): New function.
10276 (aarch64_simd_vector_alignment_reachable): Use it instead of
10278 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10279 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10281 (MAX_VECT_LEN): Delete.
10282 (expand_vec_perm_d): Add a vec_flags field.
10283 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10284 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10285 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10287 (aarch64_evpc_rev): Rename to...
10288 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10289 (aarch64_evpc_rev_global): New function.
10290 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10291 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10293 (aarch64_evpc_sve_tbl): New function.
10294 (aarch64_expand_vec_perm_const_1): Update after rename of
10295 aarch64_evpc_rev. Handle SVE permutes too, trying
10296 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10297 than aarch64_evpc_tbl.
10298 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10299 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10300 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10301 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10302 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10303 (aarch64_expand_sve_vcond): New functions.
10304 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10305 of aarch64_vector_mode_p.
10306 (aarch64_dwarf_poly_indeterminate_value): New function.
10307 (aarch64_compute_pressure_classes): Likewise.
10308 (aarch64_can_change_mode_class): Likewise.
10309 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10310 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10311 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10312 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10313 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10314 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10315 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10316 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10318 (Dn, Dl, Dr): Accept const as well as const_vector.
10319 (Dz): Likewise. Compare against CONST0_RTX.
10320 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10321 of "vector" where appropriate.
10322 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10323 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10324 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10325 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10326 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10327 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10328 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10329 (v_int_equiv): Extend to SVE modes.
10330 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10332 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10333 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10334 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10335 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10336 (SVE_COND_FP_CMP): New int iterators.
10337 (perm_hilo): Handle the new unpack unspecs.
10338 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10340 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10341 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10342 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10343 (aarch64_equality_operator, aarch64_constant_vector_operand)
10344 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10345 (aarch64_sve_nonimmediate_operand): Likewise.
10346 (aarch64_sve_general_operand): Likewise.
10347 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10348 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10349 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10350 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10351 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10352 (aarch64_sve_float_arith_immediate): Likewise.
10353 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10354 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10355 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10356 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10357 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10358 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10359 (aarch64_sve_float_arith_operand): Likewise.
10360 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10361 (aarch64_sve_float_mul_operand): Likewise.
10362 (aarch64_sve_vec_perm_operand): Likewise.
10363 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10364 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10365 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10366 as well as const_vector.
10367 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10368 in file. Use CONST0_RTX and CONSTM1_RTX.
10369 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10370 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10371 Use aarch64_simd_imm_zero.
10372 * config/aarch64/aarch64-sve.md: New file.
10373 * config/aarch64/aarch64.md: Include it.
10374 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10375 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10376 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10377 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10378 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10379 (sve): New attribute.
10380 (enabled): Disable instructions with the sve attribute unless
10382 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10383 aarch64_expand_mov_immediate.
10384 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10385 CNT[BHSD] immediates.
10386 (movti): Split CONST_POLY_INT moves into two halves.
10387 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10388 Split additions that need a temporary here if the destination
10389 is the stack pointer.
10390 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10391 (*add<mode>3_poly_1): New instruction.
10392 (set_clobber_cc): New expander.
10394 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10396 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10397 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10398 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10399 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10400 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10401 Change innermode from fixed_mode_size to machine_mode.
10402 (simplify_subreg): Update call accordingly. Handle a constant-sized
10403 subreg of a variable-length CONST_VECTOR.
10405 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10406 Alan Hayward <alan.hayward@arm.com>
10407 David Sherwood <david.sherwood@arm.com>
10409 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10410 (add_offset_to_base): New function, split out from...
10411 (create_mem_ref): ...here. When handling a scale other than 1,
10412 check first whether the address is valid without the offset.
10413 Add it into the base if so, leaving the index and scale as-is.
10415 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10418 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10419 fold_for_warn before checking if arg2 is INTEGER_CST.
10421 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10423 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10424 (store_multiple_operation): Delete.
10425 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10426 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10427 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10428 guarded by TARGET_STRING.
10429 (rs6000_output_load_multiple): Delete.
10430 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10431 OPTION_MASK_STRING / TARGET_STRING handling.
10432 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10433 (const rs6000_opt_masks) <"string">: Change mask to 0.
10434 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10435 (MASK_STRING): Delete.
10436 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10438 (load_multiple): Delete.
10445 (store_multiple): Delete.
10452 (movmemsi_8reg): Delete.
10453 (corresponding unnamed define_insn): Delete.
10454 (movmemsi_6reg): Delete.
10455 (corresponding unnamed define_insn): Delete.
10456 (movmemsi_4reg): Delete.
10457 (corresponding unnamed define_insn): Delete.
10458 (movmemsi_2reg): Delete.
10459 (corresponding unnamed define_insn): Delete.
10460 (movmemsi_1reg): Delete.
10461 (corresponding unnamed define_insn): Delete.
10462 * config/rs6000/rs6000.opt (mno-string): New.
10463 (mstring): Replace by deprecation warning stub.
10464 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10466 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10468 * regrename.c (regrename_do_replace): If replacing the same
10469 reg multiple times, try to reuse last created gen_raw_REG.
10472 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10473 main to workaround a bug in GDB.
10475 2018-01-12 Tom de Vries <tom@codesourcery.com>
10478 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10480 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10482 PR rtl-optimization/80481
10483 * ira-color.c (get_cap_member): New function.
10484 (allocnos_conflict_by_live_ranges_p): Use it.
10485 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10486 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10488 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10491 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10492 (*saddl_se_1): Ditto.
10493 (*ssubsi_1): Ditto.
10494 (*ssubl_se_1): Ditto.
10496 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10498 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10499 rather than wi::to_widest for DR_INITs.
10500 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10501 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10502 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10504 (vect_analyze_group_access_1): Note that here.
10506 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10508 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10509 polynomial type sizes.
10511 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10513 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10514 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10515 (gimple_add_tmp_var): Likewise.
10517 2018-01-12 Martin Liska <mliska@suse.cz>
10519 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10520 (gimple_alloc_sizes): Likewise.
10521 (dump_gimple_statistics): Use PRIu64 in printf format.
10522 * gimple.h: Change uint64_t to int.
10524 2018-01-12 Martin Liska <mliska@suse.cz>
10526 * tree-core.h: Use uint64_t instead of int.
10527 * tree.c (tree_node_counts): Likewise.
10528 (tree_node_sizes): Likewise.
10529 (dump_tree_statistics): Use PRIu64 in printf format.
10531 2018-01-12 Martin Liska <mliska@suse.cz>
10533 * Makefile.in: As qsort_chk is implemented in vec.c, add
10534 vec.o to linkage of gencfn-macros.
10535 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10536 passing the info to record_node_allocation_statistics.
10537 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10539 * ggc-common.c (struct ggc_usage): Add operator== and use
10540 it in operator< and compare function.
10541 * mem-stats.h (struct mem_usage): Likewise.
10542 * vec.c (struct vec_usage): Remove operator< and compare
10543 function. Can be simply inherited.
10545 2018-01-12 Martin Jambor <mjambor@suse.cz>
10548 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10549 * tree-ssa-math-opts.c: Include domwalk.h.
10550 (convert_mult_to_fma_1): New function.
10551 (fma_transformation_info): New type.
10552 (fma_deferring_state): Likewise.
10553 (cancel_fma_deferring): New function.
10554 (result_of_phi): Likewise.
10555 (last_fma_candidate_feeds_initial_phi): Likewise.
10556 (convert_mult_to_fma): Added deferring logic, split actual
10557 transformation to convert_mult_to_fma_1.
10558 (math_opts_dom_walker): New type.
10559 (math_opts_dom_walker::after_dom_children): New method, body moved
10560 here from pass_optimize_widening_mul::execute, added deferring logic
10562 (pass_optimize_widening_mul::execute): Moved most of code to
10563 math_opts_dom_walker::after_dom_children.
10564 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10565 * config/i386/i386.c (ix86_option_override_internal): Added
10566 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10568 2018-01-12 Richard Biener <rguenther@suse.de>
10571 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10572 inline instance vars.
10574 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10577 * config/rx/rx.c (rx_is_restricted_memory_address):
10578 Handle SUBREG case.
10580 2018-01-12 Richard Biener <rguenther@suse.de>
10582 PR tree-optimization/80846
10583 * target.def (split_reduction): New target hook.
10584 * targhooks.c (default_split_reduction): New function.
10585 * targhooks.h (default_split_reduction): Declare.
10586 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10587 target requests first reduce vectors by combining low and high
10589 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10590 (get_vectype_for_scalar_type_and_size): Export.
10591 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10592 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10593 * doc/tm.texi: Regenerate.
10594 * config/i386/i386.c (ix86_split_reduction): Implement
10595 TARGET_VECTORIZE_SPLIT_REDUCTION.
10597 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10600 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10601 in PIC mode except for TARGET_VXWORKS_RTP.
10602 * config/sparc/sparc.c: Include cfgrtl.h.
10603 (TARGET_INIT_PIC_REG): Define.
10604 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10605 (sparc_pic_register_p): New predicate.
10606 (sparc_legitimate_address_p): Use it.
10607 (sparc_legitimize_pic_address): Likewise.
10608 (sparc_delegitimize_address): Likewise.
10609 (sparc_mode_dependent_address_p): Likewise.
10610 (gen_load_pcrel_sym): Remove 4th parameter.
10611 (load_got_register): Adjust call to above. Remove obsolete stuff.
10612 (sparc_expand_prologue): Do not call load_got_register here.
10613 (sparc_flat_expand_prologue): Likewise.
10614 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10615 (sparc_use_pseudo_pic_reg): New function.
10616 (sparc_init_pic_reg): Likewise.
10617 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10618 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10620 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10622 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10623 Add item for branch_cost.
10625 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10627 PR rtl-optimization/83565
10628 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10629 not extend the result to a larger mode for rotate operations.
10630 (num_sign_bit_copies1): Likewise.
10632 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10635 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10637 Use values-Xc.o for -pedantic.
10638 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10640 2018-01-12 Martin Liska <mliska@suse.cz>
10643 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10645 (possible_polymorphic_call_targets): Use it.
10646 (ipa_devirt): Likewise.
10648 2018-01-12 Martin Liska <mliska@suse.cz>
10650 * profile-count.h (enum profile_quality): Use 0 as invalid
10651 enum value of profile_quality.
10653 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10655 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10656 -mext-string options.
10658 2018-01-12 Richard Biener <rguenther@suse.de>
10660 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10661 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10662 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10664 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10666 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10668 * configure.ac (--with-long-double-format): Add support for the
10669 configuration option to change the default long double format on
10671 * config.gcc (powerpc*-linux*-*): Likewise.
10672 * configure: Regenerate.
10673 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10674 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10675 used without modification.
10677 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10679 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10680 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10681 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10682 MISC_BUILTIN_SPEC_BARRIER.
10683 (rs6000_init_builtins): Likewise.
10684 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10686 (speculation_barrier): New define_insn.
10687 * doc/extend.texi: Document __builtin_speculation_barrier.
10689 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10692 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10693 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10694 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10696 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10697 integral modes instead of "ss" and "sd".
10698 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10699 vectors with 32-bit and 64-bit elements.
10700 (vecdupssescalarmodesuffix): New mode attribute.
10701 (vec_dup<mode>): Use it.
10703 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10706 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10707 frame if argument is passed on stack.
10709 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10712 * ree.c (combine_reaching_defs): Optimize also
10713 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10714 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10716 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10718 PR middle-end/83189
10719 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10721 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10723 PR middle-end/83718
10724 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10725 after they are computed.
10727 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10729 PR tree-optimization/83695
10730 * gimple-loop-linterchange.cc
10731 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10732 reset cached scev information after interchange.
10733 (pass_linterchange::execute): Remove call to scev_reset_htab.
10735 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10737 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10738 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10739 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10740 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10741 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10742 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10743 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10744 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10745 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10746 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10747 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10748 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10749 (V_lane_reg): Likewise.
10750 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10752 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10753 (vfmal_lane_low<mode>_intrinsic,
10754 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10755 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10756 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10757 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10758 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10759 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10761 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10763 * config/arm/arm-cpus.in (fp16fml): New feature.
10764 (ALL_SIMD): Add fp16fml.
10765 (armv8.2-a): Add fp16fml as an option.
10766 (armv8.3-a): Likewise.
10767 (armv8.4-a): Add fp16fml as part of fp16.
10768 * config/arm/arm.h (TARGET_FP16FML): Define.
10769 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10771 * config/arm/arm-modes.def (V2HF): Define.
10772 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10773 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10774 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10775 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10776 vfmsl_low, vfmsl_high): New set of builtins.
10777 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10778 (vfml_op): New code attribute.
10779 (VFMLHALVES): New int iterator.
10780 (VFML, VFMLSEL): New mode attributes.
10781 (V_reg): Define mapping for V2HF.
10782 (V_hi, V_lo): New mode attributes.
10783 (VF_constraint): Likewise.
10784 (vfml_half, vfml_half_selector): New int attributes.
10785 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10787 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10788 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10790 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10791 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10792 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10793 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10795 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10796 Document new effective target and option set.
10798 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10800 * config/arm/arm-cpus.in (armv8_4): New feature.
10801 (ARMv8_4a): New fgroup.
10802 (armv8.4-a): New arch.
10803 * config/arm/arm-tables.opt: Regenerate.
10804 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10805 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10806 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10807 Add matching rules for -march=armv8.4-a and extensions.
10808 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10810 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10813 * config/rx/rx.md (BW): New mode attribute.
10814 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10816 2018-01-11 Richard Biener <rguenther@suse.de>
10818 PR tree-optimization/83435
10819 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10820 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10821 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10823 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10824 Alan Hayward <alan.hayward@arm.com>
10825 David Sherwood <david.sherwood@arm.com>
10827 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10829 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10830 (aarch64_print_address_internal): Use it to check for a zero offset.
10832 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10833 Alan Hayward <alan.hayward@arm.com>
10834 David Sherwood <david.sherwood@arm.com>
10836 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10837 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10838 Return a poly_int64 rather than a HOST_WIDE_INT.
10839 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10840 rather than a HOST_WIDE_INT.
10841 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10842 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10843 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10844 final_offset from HOST_WIDE_INT to poly_int64.
10845 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10846 to_constant when getting the number of units in an Advanced SIMD
10848 (aarch64_builtin_vectorized_function): Check for a constant number
10850 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10852 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10853 attribute instead of GET_MODE_NUNITS.
10854 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10855 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10856 GET_MODE_SIZE for fixed-size registers.
10857 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10858 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10859 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10860 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10861 (aarch64_print_operand, aarch64_print_address_internal)
10862 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10863 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10864 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10865 Handle polynomial GET_MODE_SIZE.
10866 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10867 wider than SImode without modification.
10868 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10869 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10870 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10871 passing and returning SVE modes.
10872 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10873 rather than GEN_INT.
10874 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10875 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10876 (aarch64_allocate_and_probe_stack_space): Likewise.
10877 (aarch64_layout_frame): Cope with polynomial offsets.
10878 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10879 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10880 polynomial offsets.
10881 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10882 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10883 poly_int64 rather than a HOST_WIDE_INT.
10884 (aarch64_get_separate_components, aarch64_process_components)
10885 (aarch64_expand_prologue, aarch64_expand_epilogue)
10886 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10887 (aarch64_anchor_offset): New function, split out from...
10888 (aarch64_legitimize_address): ...here.
10889 (aarch64_builtin_vectorization_cost): Handle polynomial
10890 TYPE_VECTOR_SUBPARTS.
10891 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10893 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10894 number of elements from the PARALLEL rather than the mode.
10895 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10896 rather than GET_MODE_BITSIZE.
10897 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10898 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10899 (aarch64_expand_vec_perm_const_1): Handle polynomial
10900 d->perm.length () and d->perm elements.
10901 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10902 Apply to_constant to d->perm elements.
10903 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10904 polynomial CONST_VECTOR_NUNITS.
10905 (aarch64_move_pointer): Take amount as a poly_int64 rather
10907 (aarch64_progress_pointer): Avoid temporary variable.
10908 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10909 the mode attribute instead of GET_MODE.
10911 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10912 Alan Hayward <alan.hayward@arm.com>
10913 David Sherwood <david.sherwood@arm.com>
10915 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10916 x exists before using it.
10917 (aarch64_add_constant_internal): Rename to...
10918 (aarch64_add_offset_1): ...this. Replace regnum with separate
10919 src and dest rtxes. Handle the case in which they're different,
10920 including when the offset is zero. Replace scratchreg with an rtx.
10921 Use 2 additions if there is no spare register into which we can
10922 move a 16-bit constant.
10923 (aarch64_add_constant): Delete.
10924 (aarch64_add_offset): Replace reg with separate src and dest
10925 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10926 Use aarch64_add_offset_1.
10927 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10928 an rtx rather than an int. Take the delta as a poly_int64
10929 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10930 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10931 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10932 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10933 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10934 and aarch64_add_sp.
10935 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10936 aarch64_add_constant.
10938 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10940 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10941 Use scalar_float_mode.
10943 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10945 * config/aarch64/aarch64-simd.md
10946 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10947 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10948 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10949 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10950 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10951 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10952 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10953 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10954 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10955 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10957 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10960 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10961 targ_options->x_arm_arch_string is non NULL.
10963 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10965 * config/aarch64/aarch64.h
10966 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10968 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10971 * expmed.c (emit_store_flag_force): Swap if const op0
10972 and change VOIDmode to mode of op0.
10974 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10976 PR rtl-optimization/83761
10977 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10978 than bytes to mode_for_size.
10980 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10982 PR middle-end/83189
10983 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10984 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10987 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10989 PR middle-end/83575
10990 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10991 when in layout mode.
10992 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10993 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10996 2018-01-10 Michael Collison <michael.collison@arm.com>
10998 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10999 * config/aarch64/aarch64-option-extension.def: Add
11000 AARCH64_OPT_EXTENSION of 'fp16fml'.
11001 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11002 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11003 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11004 * config/aarch64/constraints.md (Ui7): New constraint.
11005 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11006 (VFMLA_SEL_W): Ditto.
11009 (VFMLA16_LOW): New int iterator.
11010 (VFMLA16_HIGH): Ditto.
11011 (UNSPEC_FMLAL): New unspec.
11012 (UNSPEC_FMLSL): Ditto.
11013 (UNSPEC_FMLAL2): Ditto.
11014 (UNSPEC_FMLSL2): Ditto.
11015 (f16mac): New code attribute.
11016 * config/aarch64/aarch64-simd-builtins.def
11017 (aarch64_fmlal_lowv2sf): Ditto.
11018 (aarch64_fmlsl_lowv2sf): Ditto.
11019 (aarch64_fmlalq_lowv4sf): Ditto.
11020 (aarch64_fmlslq_lowv4sf): Ditto.
11021 (aarch64_fmlal_highv2sf): Ditto.
11022 (aarch64_fmlsl_highv2sf): Ditto.
11023 (aarch64_fmlalq_highv4sf): Ditto.
11024 (aarch64_fmlslq_highv4sf): Ditto.
11025 (aarch64_fmlal_lane_lowv2sf): Ditto.
11026 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11027 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11028 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11029 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11030 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11031 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11032 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11033 (aarch64_fmlal_lane_highv2sf): Ditto.
11034 (aarch64_fmlsl_lane_highv2sf): Ditto.
11035 (aarch64_fmlal_laneq_highv2sf): Ditto.
11036 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11037 (aarch64_fmlalq_lane_highv4sf): Ditto.
11038 (aarch64_fmlsl_lane_highv4sf): Ditto.
11039 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11040 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11041 * config/aarch64/aarch64-simd.md:
11042 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11043 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11044 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11045 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11046 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11047 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11048 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11049 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11050 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11051 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11052 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11053 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11054 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11055 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11056 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11057 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11058 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11059 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11060 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11061 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11062 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11063 (vfmlsl_low_u32): Ditto.
11064 (vfmlalq_low_u32): Ditto.
11065 (vfmlslq_low_u32): Ditto.
11066 (vfmlal_high_u32): Ditto.
11067 (vfmlsl_high_u32): Ditto.
11068 (vfmlalq_high_u32): Ditto.
11069 (vfmlslq_high_u32): Ditto.
11070 (vfmlal_lane_low_u32): Ditto.
11071 (vfmlsl_lane_low_u32): Ditto.
11072 (vfmlal_laneq_low_u32): Ditto.
11073 (vfmlsl_laneq_low_u32): Ditto.
11074 (vfmlalq_lane_low_u32): Ditto.
11075 (vfmlslq_lane_low_u32): Ditto.
11076 (vfmlalq_laneq_low_u32): Ditto.
11077 (vfmlslq_laneq_low_u32): Ditto.
11078 (vfmlal_lane_high_u32): Ditto.
11079 (vfmlsl_lane_high_u32): Ditto.
11080 (vfmlal_laneq_high_u32): Ditto.
11081 (vfmlsl_laneq_high_u32): Ditto.
11082 (vfmlalq_lane_high_u32): Ditto.
11083 (vfmlslq_lane_high_u32): Ditto.
11084 (vfmlalq_laneq_high_u32): Ditto.
11085 (vfmlslq_laneq_high_u32): Ditto.
11086 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11087 (AARCH64_FL_FOR_ARCH8_4): New.
11088 (AARCH64_ISA_F16FML): New ISA flag.
11089 (TARGET_F16FML): New feature flag for fp16fml.
11090 (doc/invoke.texi): Document new fp16fml option.
11092 2018-01-10 Michael Collison <michael.collison@arm.com>
11094 * config/aarch64/aarch64-builtins.c:
11095 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11096 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11097 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11098 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11099 (AARCH64_ISA_SHA3): New ISA flag.
11100 (TARGET_SHA3): New feature flag for sha3.
11101 * config/aarch64/iterators.md (sha512_op): New int attribute.
11102 (CRYPTO_SHA512): New int iterator.
11103 (UNSPEC_SHA512H): New unspec.
11104 (UNSPEC_SHA512H2): Ditto.
11105 (UNSPEC_SHA512SU0): Ditto.
11106 (UNSPEC_SHA512SU1): Ditto.
11107 * config/aarch64/aarch64-simd-builtins.def
11108 (aarch64_crypto_sha512hqv2di): New builtin.
11109 (aarch64_crypto_sha512h2qv2di): Ditto.
11110 (aarch64_crypto_sha512su0qv2di): Ditto.
11111 (aarch64_crypto_sha512su1qv2di): Ditto.
11112 (aarch64_eor3qv8hi): Ditto.
11113 (aarch64_rax1qv2di): Ditto.
11114 (aarch64_xarqv2di): Ditto.
11115 (aarch64_bcaxqv8hi): Ditto.
11116 * config/aarch64/aarch64-simd.md:
11117 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11118 (aarch64_crypto_sha512su0qv2di): Ditto.
11119 (aarch64_crypto_sha512su1qv2di): Ditto.
11120 (aarch64_eor3qv8hi): Ditto.
11121 (aarch64_rax1qv2di): Ditto.
11122 (aarch64_xarqv2di): Ditto.
11123 (aarch64_bcaxqv8hi): Ditto.
11124 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11125 (vsha512h2q_u64): Ditto.
11126 (vsha512su0q_u64): Ditto.
11127 (vsha512su1q_u64): Ditto.
11128 (veor3q_u16): Ditto.
11129 (vrax1q_u64): Ditto.
11130 (vxarq_u64): Ditto.
11131 (vbcaxq_u16): Ditto.
11132 * config/arm/types.md (crypto_sha512): New type attribute.
11133 (crypto_sha3): Ditto.
11134 (doc/invoke.texi): Document new sha3 option.
11136 2018-01-10 Michael Collison <michael.collison@arm.com>
11138 * config/aarch64/aarch64-builtins.c:
11139 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11140 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11141 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11142 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11143 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11144 (AARCH64_ISA_SM4): New ISA flag.
11145 (TARGET_SM4): New feature flag for sm4.
11146 * config/aarch64/aarch64-simd-builtins.def
11147 (aarch64_sm3ss1qv4si): Ditto.
11148 (aarch64_sm3tt1aq4si): Ditto.
11149 (aarch64_sm3tt1bq4si): Ditto.
11150 (aarch64_sm3tt2aq4si): Ditto.
11151 (aarch64_sm3tt2bq4si): Ditto.
11152 (aarch64_sm3partw1qv4si): Ditto.
11153 (aarch64_sm3partw2qv4si): Ditto.
11154 (aarch64_sm4eqv4si): Ditto.
11155 (aarch64_sm4ekeyqv4si): Ditto.
11156 * config/aarch64/aarch64-simd.md:
11157 (aarch64_sm3ss1qv4si): Ditto.
11158 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11159 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11160 (aarch64_sm4eqv4si): Ditto.
11161 (aarch64_sm4ekeyqv4si): Ditto.
11162 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11163 (sm3part_op): Ditto.
11164 (CRYPTO_SM3TT): Ditto.
11165 (CRYPTO_SM3PART): Ditto.
11166 (UNSPEC_SM3SS1): New unspec.
11167 (UNSPEC_SM3TT1A): Ditto.
11168 (UNSPEC_SM3TT1B): Ditto.
11169 (UNSPEC_SM3TT2A): Ditto.
11170 (UNSPEC_SM3TT2B): Ditto.
11171 (UNSPEC_SM3PARTW1): Ditto.
11172 (UNSPEC_SM3PARTW2): Ditto.
11173 (UNSPEC_SM4E): Ditto.
11174 (UNSPEC_SM4EKEY): Ditto.
11175 * config/aarch64/constraints.md (Ui2): New constraint.
11176 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11177 * config/arm/types.md (crypto_sm3): New type attribute.
11178 (crypto_sm4): Ditto.
11179 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11180 (vsm3tt1aq_u32): Ditto.
11181 (vsm3tt1bq_u32): Ditto.
11182 (vsm3tt2aq_u32): Ditto.
11183 (vsm3tt2bq_u32): Ditto.
11184 (vsm3partw1q_u32): Ditto.
11185 (vsm3partw2q_u32): Ditto.
11186 (vsm4eq_u32): Ditto.
11187 (vsm4ekeyq_u32): Ditto.
11188 (doc/invoke.texi): Document new sm4 option.
11190 2018-01-10 Michael Collison <michael.collison@arm.com>
11192 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11193 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11194 (AARCH64_FL_FOR_ARCH8_4): New.
11195 (AARCH64_FL_V8_4): New flag.
11196 (doc/invoke.texi): Document new armv8.4-a option.
11198 2018-01-10 Michael Collison <michael.collison@arm.com>
11200 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11201 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11202 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11203 * config/aarch64/aarch64-option-extension.def: Add
11204 AARCH64_OPT_EXTENSION of 'sha2'.
11205 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11206 (crypto): Disable sha2 and aes if crypto disabled.
11207 (crypto): Enable aes and sha2 if enabled.
11208 (simd): Disable sha2 and aes if simd disabled.
11209 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11211 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11212 (TARGET_SHA2): New feature flag for sha2.
11213 (TARGET_AES): New feature flag for aes.
11214 * config/aarch64/aarch64-simd.md:
11215 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11216 conditional on TARGET_AES.
11217 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11218 (aarch64_crypto_sha1hsi): Make pattern conditional
11220 (aarch64_crypto_sha1hv4si): Ditto.
11221 (aarch64_be_crypto_sha1hv4si): Ditto.
11222 (aarch64_crypto_sha1su1v4si): Ditto.
11223 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11224 (aarch64_crypto_sha1su0v4si): Ditto.
11225 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11226 (aarch64_crypto_sha256su0v4si): Ditto.
11227 (aarch64_crypto_sha256su1v4si): Ditto.
11228 (doc/invoke.texi): Document new aes and sha2 options.
11230 2018-01-10 Martin Sebor <msebor@redhat.com>
11232 PR tree-optimization/83781
11233 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11236 2018-01-11 Martin Sebor <msebor@gmail.com>
11237 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11239 PR tree-optimization/83501
11240 PR tree-optimization/81703
11242 * tree-ssa-strlen.c (get_string_cst): Rename...
11243 (get_string_len): ...to this. Handle global constants.
11244 (handle_char_store): Adjust.
11246 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11247 Jim Wilson <jimw@sifive.com>
11249 * config/riscv/riscv-protos.h (riscv_output_return): New.
11250 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11251 (riscv_attribute_table, riscv_output_return),
11252 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11253 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11254 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11256 (riscv_expand_prologue): Add early return for naked function.
11257 (riscv_expand_epilogue): Likewise.
11258 (riscv_function_ok_for_sibcall): Return false for naked function.
11259 (riscv_set_current_function): New.
11260 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11261 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11262 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11263 * doc/extend.texi (RISC-V Function Attributes): New.
11265 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11267 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11268 check for 128-bit long double before checking TCmode.
11269 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11270 128-bit long doubles before checking TFmode or TCmode.
11271 (FLOAT128_IBM_P): Likewise.
11273 2018-01-10 Martin Sebor <msebor@redhat.com>
11275 PR tree-optimization/83671
11276 * builtins.c (c_strlen): Unconditionally return zero for the empty
11278 Use -Warray-bounds for warnings.
11279 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11280 for non-constant array indices with COMPONENT_REF, arrays of
11281 arrays, and pointers to arrays.
11282 (gimple_fold_builtin_strlen): Determine and set length range for
11283 non-constant character arrays.
11285 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11287 PR middle-end/81897
11288 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11291 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11293 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11295 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11298 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11299 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11300 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11301 indexed_or_indirect_operand predicate.
11302 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11303 (*vsx_le_perm_load_v8hi): Likewise.
11304 (*vsx_le_perm_load_v16qi): Likewise.
11305 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11306 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11307 (*vsx_le_perm_store_v8hi): Likewise.
11308 (*vsx_le_perm_store_v16qi): Likewise.
11309 (eight unnamed splitters): Likewise.
11311 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11313 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11314 * config/rs6000/emmintrin.h: Likewise.
11315 * config/rs6000/mmintrin.h: Likewise.
11316 * config/rs6000/xmmintrin.h: Likewise.
11318 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11321 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11323 * tree.c (tree_nop_conversion): Return true for location wrapper
11325 (maybe_wrap_with_location): New function.
11326 (selftest::check_strip_nops): New function.
11327 (selftest::test_location_wrappers): New function.
11328 (selftest::tree_c_tests): Call it.
11329 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11330 (maybe_wrap_with_location): New decl.
11331 (EXPR_LOCATION_WRAPPER_P): New macro.
11332 (location_wrapper_p): New inline function.
11333 (tree_strip_any_location_wrapper): New inline function.
11335 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11338 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11339 stack_realign_offset for the largest alignment of stack slot
11341 (ix86_find_max_used_stack_alignment): New function.
11342 (ix86_finalize_stack_frame_flags): Use it. Set
11343 max_used_stack_alignment if we don't realign stack.
11344 * config/i386/i386.h (machine_function): Add
11345 max_used_stack_alignment.
11347 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11349 * config/arm/arm.opt (-mbranch-cost): New option.
11350 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11353 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11356 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11357 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11359 2018-01-10 Richard Biener <rguenther@suse.de>
11362 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11363 early out so it also covers the case where we have a non-NULL
11366 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11368 PR tree-optimization/83753
11369 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11370 for non-strided grouped accesses if the number of elements is 1.
11372 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11375 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11376 * i386.h (TARGET_USE_GATHER): Define.
11377 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11379 2018-01-10 Martin Liska <mliska@suse.cz>
11382 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11383 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11385 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11386 CLEANUP_NO_PARTITIONING is not set.
11388 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11390 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11391 for vectors, as a partial revert of r254296.
11392 * rtl.h (const_vec_p): Delete.
11393 (const_vec_duplicate_p): Don't test for vector CONSTs.
11394 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11395 * expmed.c (make_tree): Likewise.
11398 * common.md (E, F): Use CONSTANT_P instead of checking for
11400 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11401 checking for CONST_VECTOR.
11403 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11405 PR middle-end/83575
11406 * predict.c (force_edge_cold): Handle in more sane way edges
11407 with no prediction.
11409 2018-01-09 Carl Love <cel@us.ibm.com>
11411 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11413 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11414 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11415 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11416 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11417 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11418 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11419 * config/rs6000/rs6000-protos.h: Add extern defition for
11420 rs6000_generate_float2_double_code.
11421 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11423 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11424 (float2_v2df): Add define_expand.
11426 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11429 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11430 op_mode in the force_to_mode call.
11432 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11434 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11435 instead of checking each element individually.
11436 (aarch64_evpc_uzp): Likewise.
11437 (aarch64_evpc_zip): Likewise.
11438 (aarch64_evpc_ext): Likewise.
11439 (aarch64_evpc_rev): Likewise.
11440 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11441 instead of checking each element individually. Return true without
11443 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11444 whether all selected elements come from the same input, instead of
11445 checking each element individually. Remove calls to gen_rtx_REG,
11446 start_sequence and end_sequence and instead assert that no rtl is
11449 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11451 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11452 order of HIGH and CONST checks.
11454 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11456 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11457 if the destination isn't an SSA_NAME.
11459 2018-01-09 Richard Biener <rguenther@suse.de>
11461 PR tree-optimization/83668
11462 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11464 (canonicalize_loop_form): ... here, renamed from ...
11465 (canonicalize_loop_closed_ssa_form): ... this and amended to
11466 swap successor edges for loop exit blocks to make us use
11467 the RPO order we need for initial schedule generation.
11469 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11471 PR tree-optimization/64811
11472 * match.pd: When optimizing comparisons with Inf, avoid
11473 introducing or losing exceptions from comparisons with NaN.
11475 2018-01-09 Martin Liska <mliska@suse.cz>
11478 * asan.c (shadow_mem_size): Add gcc_assert.
11480 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11482 Don't save registers in main().
11485 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11486 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11487 * config/avr/avr.c (avr_set_current_function): Don't error if
11488 naked, OS_task or OS_main are specified at the same time.
11489 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11491 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11493 * common/config/avr/avr-common.c (avr_option_optimization_table):
11494 Switch on -mmain-is-OS_task for optimizing compilations.
11496 2018-01-09 Richard Biener <rguenther@suse.de>
11498 PR tree-optimization/83572
11499 * graphite.c: Include cfganal.h.
11500 (graphite_transform_loops): Connect infinite loops to exit
11501 and remove fake edges at the end.
11503 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11505 * ipa-inline.c (edge_badness): Revert accidental checkin.
11507 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11510 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11511 symbols; not inline clones.
11513 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11516 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11517 hard registers. Formatting fixes.
11519 PR preprocessor/83722
11520 * gcc.c (try_generate_repro): Pass
11521 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11522 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11525 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11526 Kito Cheng <kito.cheng@gmail.com>
11528 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11529 (riscv_leaf_function_p): Delete.
11530 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11532 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11534 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11536 (do_ifelse): New function.
11537 (do_isel): New function.
11538 (do_sub3): New function.
11539 (do_add3): New function.
11540 (do_load_mask_compare): New function.
11541 (do_overlap_load_compare): New function.
11542 (expand_compare_loop): New function.
11543 (expand_block_compare): Call expand_compare_loop() when appropriate.
11544 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11545 option description.
11546 (-mblock-compare-inline-loop-limit): New option.
11548 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11551 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11552 Reverse order of second and third operands in first alternative.
11553 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11554 of first and second elements in UNSPEC_VPERMR vector.
11555 (altivec_expand_vec_perm_le): Likewise.
11557 2018-01-08 Jeff Law <law@redhat.com>
11559 PR rtl-optimizatin/81308
11560 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11561 (process_switch): If group_case_labels makes a change, then set
11563 (pass_convert_switch::execute): If a switch is converted, then
11564 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11566 PR rtl-optimization/81308
11567 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11570 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11572 PR target/83663 - Revert r255946
11573 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11574 generation for cases where splatting a value is not useful.
11575 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11576 across a vec_duplicate and a paradoxical subreg forming a vector
11577 mode to a vec_concat.
11579 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11581 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11582 -march=armv8.3-a variants.
11583 * config/arm/t-multilib: Likewise.
11584 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11586 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11588 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11590 (cceq_ior_compare_complement): Give it a name so I can use it, and
11591 change boolean_or_operator predicate to boolean_operator so it can
11592 be used to generate a crand.
11593 (eqne): New code iterator.
11594 (bd/bd_neg): New code_attrs.
11595 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11596 a single define_insn.
11597 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11598 decrement (bdnzt/bdnzf/bdzt/bdzf).
11599 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11600 with the new names of the branch decrement patterns, and added the
11601 names of the branch decrement conditional patterns.
11603 2018-01-08 Richard Biener <rguenther@suse.de>
11605 PR tree-optimization/83563
11606 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11609 2018-01-08 Richard Biener <rguenther@suse.de>
11611 PR middle-end/83713
11612 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11614 2018-01-08 Richard Biener <rguenther@suse.de>
11616 PR tree-optimization/83685
11617 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11618 references to abnormals.
11620 2018-01-08 Richard Biener <rguenther@suse.de>
11623 * dwarf2out.c (output_indirect_strings): Handle empty
11624 skeleton_debug_str_hash.
11625 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11627 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11629 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11630 (emit_store_direct): Likewise.
11631 (arc_trampoline_adjust_address): Likewise.
11632 (arc_asm_trampoline_template): New function.
11633 (arc_initialize_trampoline): Use asm_trampoline_template.
11634 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11635 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11636 * config/arc/arc.md (flush_icache): Delete pattern.
11638 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11640 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11641 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11644 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11647 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11648 by not USED_FOR_TARGET.
11649 (make_pass_resolve_sw_modes): Likewise.
11651 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11653 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11656 2018-01-08 Richard Biener <rguenther@suse.de>
11658 PR middle-end/83580
11659 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11661 2018-01-08 Richard Biener <rguenther@suse.de>
11663 PR middle-end/83517
11664 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11666 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11668 PR middle-end/81897
11669 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11670 basic blocks with a small number of successors.
11671 (convert_control_dep_chain_into_preds): Improve handling of
11673 (dump_predicates): Split apart into...
11674 (dump_pred_chain): ...here...
11675 (dump_pred_info): ...and here.
11676 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11677 (can_chain_union_be_invalidated_p): Improve check for invalidation
11679 (uninit_uses_cannot_happen): Avoid unnecessary if
11680 convert_control_dep_chain_into_preds yielded nothing.
11682 2018-01-06 Martin Sebor <msebor@redhat.com>
11684 PR tree-optimization/83640
11685 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11686 subtracting negative offset from size.
11687 (builtin_access::overlap): Adjust offset bounds of the access to fall
11688 within the size of the object if possible.
11690 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11692 PR rtl-optimization/83699
11693 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11694 extract_bit_field_as_subreg to cases in which the extracted
11695 value is also a vector.
11697 * lra-constraints.c (process_alt_operands): Test for the equivalence
11698 substitutions when detecting a possible reload cycle.
11700 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11703 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11704 by default if flag_selective_schedling{,2}. Formatting fixes.
11706 PR rtl-optimization/83682
11707 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11708 if it has non-VECTOR_MODE element mode.
11709 (vec_duplicate_p): Likewise.
11711 PR middle-end/83694
11712 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11713 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11715 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11718 * config/i386/i386-builtin.def
11719 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11720 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11721 Require also OPTION_MASK_ISA_AVX512F in addition to
11722 OPTION_MASK_ISA_GFNI.
11723 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11724 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11725 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11726 to OPTION_MASK_ISA_GFNI.
11727 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11728 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11729 OPTION_MASK_ISA_AVX512BW.
11730 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11731 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11732 addition to OPTION_MASK_ISA_GFNI.
11733 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11734 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11735 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11736 to OPTION_MASK_ISA_GFNI.
11737 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11738 a requirement for all ISAs rather than any of them with a few
11740 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11742 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11743 bitmasks to be enabled with 3 exceptions, instead of requiring any
11744 enabled ISA with lots of exceptions.
11745 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11746 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11747 Change avx512bw in isa attribute to avx512f.
11748 * config/i386/sgxintrin.h: Add license boilerplate.
11749 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11750 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11751 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11752 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11754 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11755 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11756 temporarily sse2 rather than sse if not enabled already.
11759 * config/i386/sse.md (VI248_VLBW): Rename to ...
11760 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11761 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11762 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11763 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11764 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11765 mode iterator instead of VI248_VLBW.
11767 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11769 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11770 (record_modified): Skip clobbers; add debug output.
11771 (param_change_prob): Use sreal frequencies.
11773 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11775 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11776 punt for user-aligned variables.
11778 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11780 * tree-chrec.c (chrec_contains_symbols): Return true for
11783 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11786 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11787 of (x|y) == x for BICS pattern.
11789 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11791 PR tree-optimization/83605
11792 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11793 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11796 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11798 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11799 * config/epiphany/rtems.h: New file.
11801 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11802 Uros Bizjak <ubizjak@gmail.com>
11805 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11806 QIreg_operand instead of register_operand predicate.
11807 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11808 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11809 comments instead of -fmitigate[-_]rop.
11811 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11814 * cgraphunit.c (symbol_table::compile): Switch to text_section
11815 before calling assembly_start debug hook.
11816 * run-rtl-passes.c (run_rtl_passes): Likewise.
11819 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11821 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11822 range_int_cst_p rather than !symbolic_range_p before calling
11823 extract_range_from_multiplicative_op_1.
11825 2018-01-04 Jeff Law <law@redhat.com>
11827 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11828 redundant test in assertion.
11830 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11832 * doc/rtl.texi: Document machine_mode wrapper classes.
11834 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11836 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11837 using tree_to_uhwi.
11839 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11841 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11842 the VEC_PERM_EXPR fold to fail.
11844 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11847 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11848 to switched_sections.
11850 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11853 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11854 test for d.testing.
11856 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11859 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11860 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11862 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11865 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11866 is BLKmode and bitpos not zero or mode change is needed.
11868 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11871 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11874 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11877 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11878 instead of MULT rtx. Update all corresponding splitters.
11879 (*saddl_se): Ditto.
11880 (*ssub<modesuffix>): Ditto.
11881 (*ssubl_se): Ditto.
11882 (*cmp_sadd_di): Update split patterns.
11883 (*cmp_sadd_si): Ditto.
11884 (*cmp_sadd_sidi): Ditto.
11885 (*cmp_ssub_di): Ditto.
11886 (*cmp_ssub_si): Ditto.
11887 (*cmp_ssub_sidi): Ditto.
11888 * config/alpha/predicates.md (const23_operand): New predicate.
11889 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11890 Look for ASHIFT, not MULT inner operand.
11891 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11893 2018-01-04 Martin Liska <mliska@suse.cz>
11895 PR gcov-profile/83669
11896 * gcov.c (output_intermediate_file): Add version to intermediate
11898 * doc/gcov.texi: Document new field 'version' in intermediate
11899 file format. Fix location of '-k' option of gcov command.
11901 2018-01-04 Martin Liska <mliska@suse.cz>
11904 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11906 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11908 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11910 2018-01-03 Martin Sebor <msebor@redhat.com>
11912 PR tree-optimization/83655
11913 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11914 checking calls with invalid arguments.
11916 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11918 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11919 (vectorizable_mask_load_store): Delete.
11920 (vectorizable_call): Return false for masked loads and stores.
11921 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11922 instead of gimple_assign_rhs1.
11923 (vectorizable_load): Handle IFN_MASK_LOAD.
11924 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11928 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11930 (vectorizable_mask_load_store): ...here.
11931 (vectorizable_load): ...and here.
11933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11935 * tree-vect-stmts.c (vect_build_all_ones_mask)
11936 (vect_build_zero_merge_argument): New functions, split out from...
11937 (vectorizable_load): ...here.
11939 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11941 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11943 (vectorizable_mask_load_store): ...here.
11944 (vectorizable_store): ...and here.
11946 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11948 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11950 (vectorizable_mask_load_store): ...here.
11952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11954 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11955 (vect_model_store_cost): Take a vec_load_store_type instead of a
11957 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11958 (vect_model_store_cost): Take a vec_load_store_type instead of a
11960 (vectorizable_mask_load_store): Update accordingly.
11961 (vectorizable_store): Likewise.
11962 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11966 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11967 IFN_MASK_LOAD calls here rather than...
11968 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11971 Alan Hayward <alan.hayward@arm.com>
11972 David Sherwood <david.sherwood@arm.com>
11974 * expmed.c (extract_bit_field_1): For vector extracts,
11975 fall back to extract_bit_field_as_subreg if vec_extract
11978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11979 Alan Hayward <alan.hayward@arm.com>
11980 David Sherwood <david.sherwood@arm.com>
11982 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11983 they are variable or constant sized.
11984 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11985 slots for constant-sized data.
11987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11988 Alan Hayward <alan.hayward@arm.com>
11989 David Sherwood <david.sherwood@arm.com>
11991 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11992 handling COND_EXPRs with boolean comparisons, try to find a better
11993 basis for the mask type than the boolean itself.
11995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11997 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11998 is calculated and how it can be overridden.
11999 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12000 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12002 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12006 Alan Hayward <alan.hayward@arm.com>
12007 David Sherwood <david.sherwood@arm.com>
12009 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12010 Remove the mode argument.
12011 (aarch64_simd_valid_immediate): Remove the mode and inverse
12013 * config/aarch64/iterators.md (bitsize): New iterator.
12014 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12015 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12016 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12017 aarch64_simd_valid_immediate.
12018 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12019 (aarch64_reg_or_bic_imm): Likewise.
12020 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12021 with an insn_type enum and msl with a modifier_type enum.
12022 Replace element_width with a scalar_mode. Change the shift
12023 to unsigned int. Add constructors for scalar_float_mode and
12024 scalar_int_mode elements.
12025 (aarch64_vect_float_const_representable_p): Delete.
12026 (aarch64_can_const_movi_rtx_p)
12027 (aarch64_simd_scalar_immediate_valid_for_move)
12028 (aarch64_simd_make_constant): Update call to
12029 aarch64_simd_valid_immediate.
12030 (aarch64_advsimd_valid_immediate_hs): New function.
12031 (aarch64_advsimd_valid_immediate): Likewise.
12032 (aarch64_simd_valid_immediate): Remove mode and inverse
12033 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12034 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12035 and aarch64_float_const_representable_p on the result.
12036 (aarch64_output_simd_mov_immediate): Remove mode argument.
12037 Update call to aarch64_simd_valid_immediate and use of
12038 simd_immediate_info.
12039 (aarch64_output_scalar_simd_mov_immediate): Update call
12042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12043 Alan Hayward <alan.hayward@arm.com>
12044 David Sherwood <david.sherwood@arm.com>
12046 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12047 (mode_nunits): Likewise CONST_MODE_NUNITS.
12048 * machmode.def (ADJUST_NUNITS): Document.
12049 * genmodes.c (mode_data::need_nunits_adj): New field.
12050 (blank_mode): Update accordingly.
12051 (adj_nunits): New variable.
12052 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12054 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12055 listed in adj_nunits.
12056 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12057 listed in adj_nunits. Don't emit case statements for such modes.
12058 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12059 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12060 nothing if adj_nunits is nonnull.
12061 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12062 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12063 (emit_mode_fbit): Update use of print_maybe_const_decl.
12064 (emit_move_size): Likewise. Treat the array as non-const
12066 (emit_mode_adjustments): Handle adj_nunits.
12068 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12070 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12071 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12072 (VECTOR_MODES): Use it.
12073 (make_vector_modes): Take the prefix as an argument.
12075 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12076 Alan Hayward <alan.hayward@arm.com>
12077 David Sherwood <david.sherwood@arm.com>
12079 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12080 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12081 for MODE_VECTOR_BOOL.
12082 * machmode.def (VECTOR_BOOL_MODE): Document.
12083 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12084 (make_vector_bool_mode): New function.
12085 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12087 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12088 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12090 * stor-layout.c (int_mode_for_mode): Likewise.
12091 * tree.c (build_vector_type_for_mode): Likewise.
12092 * varasm.c (output_constant_pool_2): Likewise.
12093 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12094 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12095 for MODE_VECTOR_BOOL.
12096 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12097 of mode class checks.
12098 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12099 instead of a list of mode class checks.
12100 (expand_vector_scalar_condition): Likewise.
12101 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12104 Alan Hayward <alan.hayward@arm.com>
12105 David Sherwood <david.sherwood@arm.com>
12107 * machmode.h (mode_size): Change from unsigned short to
12109 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12110 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12111 or if measurement_type is not polynomial.
12112 (fixed_size_mode::includes_p): Check for constant-sized modes.
12113 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12114 return a poly_uint16 rather than an unsigned short.
12115 (emit_mode_size): Change the type of mode_size from unsigned short
12116 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12117 (emit_mode_adjustments): Cope with polynomial vector sizes.
12118 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12120 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12122 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12123 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12124 * caller-save.c (setup_save_areas): Likewise.
12125 (replace_reg_with_saved_mem): Likewise.
12126 * calls.c (emit_library_call_value_1): Likewise.
12127 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12128 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12129 (gen_lowpart_for_combine): Likewise.
12130 * convert.c (convert_to_integer_1): Likewise.
12131 * cse.c (equiv_constant, cse_insn): Likewise.
12132 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12133 (cselib_subst_to_values): Likewise.
12134 * dce.c (word_dce_process_block): Likewise.
12135 * df-problems.c (df_word_lr_mark_ref): Likewise.
12136 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12137 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12138 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12139 (rtl_for_decl_location): Likewise.
12140 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12141 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12142 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12143 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12144 (expand_expr_real_1): Likewise.
12145 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12146 (pad_below): Likewise.
12147 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12148 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12149 * ira.c (get_subreg_tracking_sizes): Likewise.
12150 * ira-build.c (ira_create_allocno_objects): Likewise.
12151 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12152 (ira_sort_regnos_for_alter_reg): Likewise.
12153 * ira-costs.c (record_operand_costs): Likewise.
12154 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12155 (resolve_simple_move): Likewise.
12156 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12157 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12158 (lra_constraints): Likewise.
12159 (CONST_POOL_OK_P): Reject variable-sized modes.
12160 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12161 (add_pseudo_to_slot, lra_spill): Likewise.
12162 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12163 * optabs-query.c (get_best_extraction_insn): Likewise.
12164 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12165 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12166 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12167 * recog.c (offsettable_address_addr_space_p): Likewise.
12168 * regcprop.c (maybe_mode_change): Likewise.
12169 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12170 * regrename.c (build_def_use): Likewise.
12171 * regstat.c (dump_reg_info): Likewise.
12172 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12173 (find_reloads, find_reloads_subreg_address): Likewise.
12174 * reload1.c (eliminate_regs_1): Likewise.
12175 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12176 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12177 (simplify_binary_operation_1, simplify_subreg): Likewise.
12178 * targhooks.c (default_function_arg_padding): Likewise.
12179 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12180 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12181 (verify_gimple_assign_ternary): Likewise.
12182 * tree-inline.c (estimate_move_cost): Likewise.
12183 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12184 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12185 (get_address_cost_ainc): Likewise.
12186 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12187 (vect_supportable_dr_alignment): Likewise.
12188 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12189 (vectorizable_reduction): Likewise.
12190 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12191 (vectorizable_operation, vectorizable_load): Likewise.
12192 * tree.c (build_same_sized_truth_vector_type): Likewise.
12193 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12194 * var-tracking.c (emit_note_insn_var_location): Likewise.
12195 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12196 (ADDR_VEC_ALIGN): Likewise.
12198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12199 Alan Hayward <alan.hayward@arm.com>
12200 David Sherwood <david.sherwood@arm.com>
12202 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12204 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12205 or if measurement_type is polynomial.
12206 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12207 * combine.c (make_extraction): Likewise.
12208 * dse.c (find_shift_sequence): Likewise.
12209 * dwarf2out.c (mem_loc_descriptor): Likewise.
12210 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12211 (extract_bit_field, extract_low_bits): Likewise.
12212 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12213 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12214 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12215 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12216 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12217 * reload.c (find_reloads): Likewise.
12218 * reload1.c (alter_reg): Likewise.
12219 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12220 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12221 * tree-if-conv.c (predicate_mem_writes): Likewise.
12222 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12223 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12224 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12225 * valtrack.c (dead_debug_insert_temp): Likewise.
12226 * varasm.c (mergeable_constant_section): Likewise.
12227 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12230 Alan Hayward <alan.hayward@arm.com>
12231 David Sherwood <david.sherwood@arm.com>
12233 * expr.c (expand_assignment): Cope with polynomial mode sizes
12234 when assigning to a CONCAT.
12236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12237 Alan Hayward <alan.hayward@arm.com>
12238 David Sherwood <david.sherwood@arm.com>
12240 * machmode.h (mode_precision): Change from unsigned short to
12242 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12244 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12245 or if measurement_type is not polynomial.
12246 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12247 in which the mode is already known to be a scalar_int_mode.
12248 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12249 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12251 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12252 for GET_MODE_PRECISION.
12253 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12254 for GET_MODE_PRECISION.
12255 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12257 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12258 (expand_field_assignment, make_extraction): Likewise.
12259 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12260 (get_last_value): Likewise.
12261 * convert.c (convert_to_integer_1): Likewise.
12262 * cse.c (cse_insn): Likewise.
12263 * expr.c (expand_expr_real_1): Likewise.
12264 * lra-constraints.c (simplify_operand_subreg): Likewise.
12265 * optabs-query.c (can_atomic_load_p): Likewise.
12266 * optabs.c (expand_atomic_load): Likewise.
12267 (expand_atomic_store): Likewise.
12268 * ree.c (combine_reaching_defs): Likewise.
12269 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12270 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12271 * tree.h (type_has_mode_precision_p): Likewise.
12272 * ubsan.c (instrument_si_overflow): Likewise.
12274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12275 Alan Hayward <alan.hayward@arm.com>
12276 David Sherwood <david.sherwood@arm.com>
12278 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12279 polynomial numbers of units.
12280 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12281 (valid_vector_subparts_p): New function.
12282 (build_vector_type): Remove temporary shim and take the number
12283 of units as a poly_uint64 rather than an int.
12284 (build_opaque_vector_type): Take the number of units as a
12285 poly_uint64 rather than an int.
12286 * tree.c (build_vector_from_ctor): Handle polynomial
12287 TYPE_VECTOR_SUBPARTS.
12288 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12289 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12290 (build_vector_from_val): If the number of units is variable,
12291 use build_vec_duplicate_cst for constant operands and
12292 VEC_DUPLICATE_EXPR otherwise.
12293 (make_vector_type): Remove temporary is_constant ().
12294 (build_vector_type, build_opaque_vector_type): Take the number of
12295 units as a poly_uint64 rather than an int.
12296 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12298 * cfgexpand.c (expand_debug_expr): Likewise.
12299 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12300 (store_constructor, expand_expr_real_1): Likewise.
12301 (const_scalar_mask_from_tree): Likewise.
12302 * fold-const-call.c (fold_const_reduction): Likewise.
12303 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12304 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12305 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12306 (fold_relational_const): Likewise.
12307 (native_interpret_vector): Likewise. Change the size from an
12308 int to an unsigned int.
12309 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12310 TYPE_VECTOR_SUBPARTS.
12311 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12312 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12313 duplicating a non-constant operand into a variable-length vector.
12314 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12315 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12316 * ipa-icf.c (sem_variable::equals): Likewise.
12317 * match.pd: Likewise.
12318 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12319 * print-tree.c (print_node): Likewise.
12320 * stor-layout.c (layout_type): Likewise.
12321 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12322 * tree-cfg.c (verify_gimple_comparison): Likewise.
12323 (verify_gimple_assign_binary): Likewise.
12324 (verify_gimple_assign_ternary): Likewise.
12325 (verify_gimple_assign_single): Likewise.
12326 * tree-pretty-print.c (dump_generic_node): Likewise.
12327 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12328 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12329 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12330 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12331 (vect_shift_permute_load_chain): Likewise.
12332 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12333 (expand_vector_condition, optimize_vector_constructor): Likewise.
12334 (lower_vec_perm, get_compute_type): Likewise.
12335 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12336 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12337 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12338 (vect_recog_mask_conversion_pattern): Likewise.
12339 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12340 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12341 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12342 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12343 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12344 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12345 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12346 (supportable_widening_operation): Likewise.
12347 (supportable_narrowing_operation): Likewise.
12348 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12350 * varasm.c (output_constant): Likewise.
12352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12353 Alan Hayward <alan.hayward@arm.com>
12354 David Sherwood <david.sherwood@arm.com>
12356 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12357 so that both the length == 3 and length != 3 cases set up their
12358 own permute vectors. Add comments explaining why we know the
12359 number of elements is constant.
12360 (vect_permute_load_chain): Likewise.
12362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12363 Alan Hayward <alan.hayward@arm.com>
12364 David Sherwood <david.sherwood@arm.com>
12366 * machmode.h (mode_nunits): Change from unsigned char to
12368 (ONLY_FIXED_SIZE_MODES): New macro.
12369 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12370 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12371 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12373 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12374 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12375 or if measurement_type is not polynomial.
12376 * genmodes.c (ZERO_COEFFS): New macro.
12377 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12379 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12380 Use ZERO_COEFFS when emitting initializers.
12381 * data-streamer.h (bp_pack_poly_value): New function.
12382 (bp_unpack_poly_value): Likewise.
12383 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12384 for GET_MODE_NUNITS.
12385 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12386 for GET_MODE_NUNITS.
12387 * tree.c (make_vector_type): Remove temporary shim and make
12388 the real function take the number of units as a poly_uint64
12389 rather than an int.
12390 (build_vector_type_for_mode): Handle polynomial nunits.
12391 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12392 * emit-rtl.c (const_vec_series_p_1): Likewise.
12393 (gen_rtx_CONST_VECTOR): Likewise.
12394 * fold-const.c (test_vec_duplicate_folding): Likewise.
12395 * genrecog.c (validate_pattern): Likewise.
12396 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12397 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12398 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12399 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12400 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12401 * rtlanal.c (subreg_get_info): Likewise.
12402 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12403 (vect_grouped_load_supported): Likewise.
12404 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12405 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12406 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12407 (simplify_const_unary_operation, simplify_binary_operation_1)
12408 (simplify_const_binary_operation, simplify_ternary_operation)
12409 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12410 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12411 instead of CONST_VECTOR_NUNITS.
12412 * varasm.c (output_constant_pool_2): Likewise.
12413 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12414 explicit-encoded elements in the XVEC for variable-length vectors.
12416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12418 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12421 Alan Hayward <alan.hayward@arm.com>
12422 David Sherwood <david.sherwood@arm.com>
12424 * coretypes.h (fixed_size_mode): Declare.
12425 (fixed_size_mode_pod): New typedef.
12426 * builtins.h (target_builtins::x_apply_args_mode)
12427 (target_builtins::x_apply_result_mode): Change type to
12428 fixed_size_mode_pod.
12429 * builtins.c (apply_args_size, apply_result_size, result_vector)
12430 (expand_builtin_apply_args_1, expand_builtin_apply)
12431 (expand_builtin_return): Update accordingly.
12433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12435 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12436 * cselib.c (cselib_hash_rtx): Likewise.
12437 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12438 CONST_VECTOR encoding.
12440 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12441 Jeff Law <law@redhat.com>
12444 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12445 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12446 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12447 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12450 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12451 explicitly probe *sp in a noreturn function if there were any callee
12452 register saves or frame pointer is needed.
12454 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12457 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12458 BLKmode for ternary, binary or unary expressions.
12461 * var-tracking.c (delete_vta_debug_insn): New inline function.
12462 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12463 insns from get_insns () to NULL instead of each bb separately.
12464 Use delete_vta_debug_insn. No longer static.
12465 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12466 delete_vta_debug_insns callers.
12467 * rtl.h (delete_vta_debug_insns): Declare.
12468 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12469 instead of variable_tracking_main.
12471 2018-01-03 Martin Sebor <msebor@redhat.com>
12473 PR tree-optimization/83603
12474 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12475 arguments past the endof the argument list in functions declared
12476 without a prototype.
12477 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12478 Avoid checking when arguments are null.
12480 2018-01-03 Martin Sebor <msebor@redhat.com>
12483 * doc/extend.texi (attribute const): Fix a typo.
12484 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12485 issuing -Wsuggest-attribute for void functions.
12487 2018-01-03 Martin Sebor <msebor@redhat.com>
12489 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12490 offset_int::from instead of wide_int::to_shwi.
12491 (maybe_diag_overlap): Remove assertion.
12492 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12493 * gimple-ssa-sprintf.c (format_directive): Same.
12494 (parse_directive): Same.
12495 (sprintf_dom_walker::compute_format_length): Same.
12496 (try_substitute_return_value): Same.
12498 2018-01-03 Jeff Law <law@redhat.com>
12500 PR middle-end/83654
12501 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12502 non-constant residual for zero at runtime and avoid probing in
12503 that case. Reorganize code for trailing problem to mirror handling
12506 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12508 PR tree-optimization/83501
12509 * tree-ssa-strlen.c (get_string_cst): New.
12510 (handle_char_store): Call get_string_cst.
12512 2018-01-03 Martin Liska <mliska@suse.cz>
12514 PR tree-optimization/83593
12515 * tree-ssa-strlen.c: Include tree-cfg.h.
12516 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12517 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12518 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12520 (strlen_dom_walker::before_dom_children): Call
12521 gimple_purge_dead_eh_edges. Dump tranformation with details
12523 (strlen_dom_walker::before_dom_children): Update call by adding
12524 new argument cleanup_eh.
12525 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12527 2018-01-03 Martin Liska <mliska@suse.cz>
12530 * cif-code.def (VARIADIC_THUNK): New enum value.
12531 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12534 2018-01-03 Jan Beulich <jbeulich@suse.com>
12536 * sse.md (mov<mode>_internal): Tighten condition for when to use
12537 vmovdqu<ssescalarsize> for TI and OI modes.
12539 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12541 Update copyright years.
12543 2018-01-03 Martin Liska <mliska@suse.cz>
12546 * ipa-visibility.c (function_and_variable_visibility): Skip
12547 functions with noipa attribure.
12549 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12551 * gcc.c (process_command): Update copyright notice dates.
12552 * gcov-dump.c (print_version): Ditto.
12553 * gcov.c (print_version): Ditto.
12554 * gcov-tool.c (print_version): Ditto.
12555 * gengtype.c (create_file): Ditto.
12556 * doc/cpp.texi: Bump @copying's copyright year.
12557 * doc/cppinternals.texi: Ditto.
12558 * doc/gcc.texi: Ditto.
12559 * doc/gccint.texi: Ditto.
12560 * doc/gcov.texi: Ditto.
12561 * doc/install.texi: Ditto.
12562 * doc/invoke.texi: Ditto.
12564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12566 * vector-builder.h (vector_builder::m_full_nelts): Change from
12567 unsigned int to poly_uint64.
12568 (vector_builder::full_nelts): Update prototype accordingly.
12569 (vector_builder::new_vector): Likewise.
12570 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12571 (vector_builder::operator ==): Likewise.
12572 (vector_builder::finalize): Likewise.
12573 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12574 Take the number of elements as a poly_uint64 rather than an
12576 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12577 from unsigned int to poly_uint64.
12578 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12579 (vec_perm_indices::new_vector): Likewise.
12580 (vec_perm_indices::length): Likewise.
12581 (vec_perm_indices::nelts_per_input): Likewise.
12582 (vec_perm_indices::input_nelts): Likewise.
12583 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12584 number of elements per input as a poly_uint64 rather than an
12585 unsigned int. Use the original encoding for variable-length
12586 vectors, rather than clamping each individual element.
12587 For the second and subsequent elements in each pattern,
12588 clamp the step and base before clamping their sum.
12589 (vec_perm_indices::series_p): Handle polynomial element counts.
12590 (vec_perm_indices::all_in_range_p): Likewise.
12591 (vec_perm_indices_to_tree): Likewise.
12592 (vec_perm_indices_to_rtx): Likewise.
12593 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12594 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12595 (tree_vector_builder::new_binary_operation): Handle polynomial
12596 element counts. Return false if we need to know the number
12597 of elements at compile time.
12598 * fold-const.c (fold_vec_perm): Punt if the number of elements
12599 isn't known at compile time.
12601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12603 * vec-perm-indices.h (vec_perm_builder): Change element type
12604 from HOST_WIDE_INT to poly_int64.
12605 (vec_perm_indices::element_type): Update accordingly.
12606 (vec_perm_indices::clamp): Handle polynomial element_types.
12607 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12608 (vec_perm_indices::all_in_range_p): Likewise.
12609 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12611 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12612 polynomial vec_perm_indices element types.
12613 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12614 * fold-const.c (fold_vec_perm): Likewise.
12615 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12616 * tree-vect-generic.c (lower_vec_perm): Likewise.
12617 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12618 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12619 element type to HOST_WIDE_INT.
12621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12622 Alan Hayward <alan.hayward@arm.com>
12623 David Sherwood <david.sherwood@arm.com>
12625 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12626 rather than an int. Use plus_constant.
12627 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12628 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12631 Alan Hayward <alan.hayward@arm.com>
12632 David Sherwood <david.sherwood@arm.com>
12634 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12635 a HOST_WIDE_INT to a poly_int64.
12637 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12638 Alan Hayward <alan.hayward@arm.com>
12639 David Sherwood <david.sherwood@arm.com>
12641 * calls.c (load_register_parameters): Cope with polynomial
12642 mode sizes. Require a constant size for BLKmode parameters
12643 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12644 forces a parameter to be padded at the lsb end in order to
12645 fill a complete number of words, require the parameter size
12646 to be ordered wrt UNITS_PER_WORD.
12648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12649 Alan Hayward <alan.hayward@arm.com>
12650 David Sherwood <david.sherwood@arm.com>
12652 * reload1.c (spill_stack_slot_width): Change element type
12653 from unsigned int to poly_uint64_pod.
12654 (alter_reg): Treat mode sizes as polynomial.
12656 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12657 Alan Hayward <alan.hayward@arm.com>
12658 David Sherwood <david.sherwood@arm.com>
12660 * reload.c (complex_word_subreg_p): New function.
12661 (reload_inner_reg_of_subreg, push_reload): Use it.
12663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12664 Alan Hayward <alan.hayward@arm.com>
12665 David Sherwood <david.sherwood@arm.com>
12667 * lra-constraints.c (process_alt_operands): Reject matched
12668 operands whose sizes aren't ordered.
12669 (match_reload): Refer to this check here.
12671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12672 Alan Hayward <alan.hayward@arm.com>
12673 David Sherwood <david.sherwood@arm.com>
12675 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12676 that the mode size is in the set {1, 2, 4, 8, 16}.
12678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12679 Alan Hayward <alan.hayward@arm.com>
12680 David Sherwood <david.sherwood@arm.com>
12682 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12683 Use plus_constant instead of gen_rtx_PLUS.
12685 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12686 Alan Hayward <alan.hayward@arm.com>
12687 David Sherwood <david.sherwood@arm.com>
12689 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12690 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12691 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12692 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12693 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12694 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12695 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12696 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12697 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12698 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12700 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12701 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12702 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12703 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12704 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12705 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12706 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12707 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12708 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12709 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12711 * expr.c (emit_move_resolve_push): Treat the input and result
12712 of PUSH_ROUNDING as a poly_int64.
12713 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12714 (emit_push_insn): Likewise.
12715 * lra-eliminations.c (mark_not_eliminable): Likewise.
12716 * recog.c (push_operand): Likewise.
12717 * reload1.c (elimination_effects): Likewise.
12718 * rtlanal.c (nonzero_bits1): Likewise.
12719 * calls.c (store_one_arg): Likewise. Require the padding to be
12720 known at compile time.
12722 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12723 Alan Hayward <alan.hayward@arm.com>
12724 David Sherwood <david.sherwood@arm.com>
12726 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12727 Use plus_constant instead of gen_rtx_PLUS.
12729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12730 Alan Hayward <alan.hayward@arm.com>
12731 David Sherwood <david.sherwood@arm.com>
12733 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12734 rather than an int.
12736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12737 Alan Hayward <alan.hayward@arm.com>
12738 David Sherwood <david.sherwood@arm.com>
12740 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12741 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12742 via stack temporaries. Treat the mode size as polynomial too.
12744 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12745 Alan Hayward <alan.hayward@arm.com>
12746 David Sherwood <david.sherwood@arm.com>
12748 * expr.c (expand_expr_real_2): When handling conversions involving
12749 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12750 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12751 as a poly_uint64 too.
12753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12754 Alan Hayward <alan.hayward@arm.com>
12755 David Sherwood <david.sherwood@arm.com>
12757 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12759 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12760 Alan Hayward <alan.hayward@arm.com>
12761 David Sherwood <david.sherwood@arm.com>
12763 * combine.c (can_change_dest_mode): Handle polynomial
12764 REGMODE_NATURAL_SIZE.
12765 * expmed.c (store_bit_field_1): Likewise.
12766 * expr.c (store_constructor): Likewise.
12767 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12768 and polynomial REGMODE_NATURAL_SIZE.
12769 (gen_lowpart_common): Likewise.
12770 * reginfo.c (record_subregs_of_mode): Likewise.
12771 * rtlanal.c (read_modify_subreg_p): Likewise.
12773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12774 Alan Hayward <alan.hayward@arm.com>
12775 David Sherwood <david.sherwood@arm.com>
12777 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12778 numbers of elements.
12780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12781 Alan Hayward <alan.hayward@arm.com>
12782 David Sherwood <david.sherwood@arm.com>
12784 * match.pd: Cope with polynomial numbers of vector elements.
12786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12787 Alan Hayward <alan.hayward@arm.com>
12788 David Sherwood <david.sherwood@arm.com>
12790 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12791 in a POINTER_PLUS_EXPR.
12793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12794 Alan Hayward <alan.hayward@arm.com>
12795 David Sherwood <david.sherwood@arm.com>
12797 * omp-simd-clone.c (simd_clone_subparts): New function.
12798 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12799 (ipa_simd_modify_function_body): Likewise.
12801 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12802 Alan Hayward <alan.hayward@arm.com>
12803 David Sherwood <david.sherwood@arm.com>
12805 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12806 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12807 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12808 (expand_vector_condition, vector_element): Likewise.
12809 (subparts_gt): New function.
12810 (get_compute_type): Use subparts_gt.
12811 (count_type_subparts): Delete.
12812 (expand_vector_operations_1): Use subparts_gt instead of
12813 count_type_subparts.
12815 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12816 Alan Hayward <alan.hayward@arm.com>
12817 David Sherwood <david.sherwood@arm.com>
12819 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12820 (vect_compile_time_alias): ...this new function. Do the calculation
12821 on poly_ints rather than trees.
12822 (vect_prune_runtime_alias_test_list): Update call accordingly.
12824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12825 Alan Hayward <alan.hayward@arm.com>
12826 David Sherwood <david.sherwood@arm.com>
12828 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12830 (vect_schedule_slp_instance): Likewise.
12832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12833 Alan Hayward <alan.hayward@arm.com>
12834 David Sherwood <david.sherwood@arm.com>
12836 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12837 constant and extern definitions for variable-length vectors.
12838 (vect_get_constant_vectors): Note that the number of units
12839 is known to be constant.
12841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12842 Alan Hayward <alan.hayward@arm.com>
12843 David Sherwood <david.sherwood@arm.com>
12845 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12846 of units as polynomial. Choose between WIDE and NARROW based
12849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12850 Alan Hayward <alan.hayward@arm.com>
12851 David Sherwood <david.sherwood@arm.com>
12853 * tree-vect-stmts.c (simd_clone_subparts): New function.
12854 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12857 Alan Hayward <alan.hayward@arm.com>
12858 David Sherwood <david.sherwood@arm.com>
12860 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12861 vectors as polynomial. Use build_index_vector for
12862 IFN_GOMP_SIMD_LANE.
12864 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12865 Alan Hayward <alan.hayward@arm.com>
12866 David Sherwood <david.sherwood@arm.com>
12868 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12869 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12870 for variable-length vectors.
12871 (vectorizable_mask_load_store): Treat the number of units as
12872 polynomial, asserting that it is constant if the condition has
12873 already been enforced.
12874 (vectorizable_store, vectorizable_load): Likewise.
12876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12877 Alan Hayward <alan.hayward@arm.com>
12878 David Sherwood <david.sherwood@arm.com>
12880 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12881 of units as polynomial. Punt if we can't tell at compile time
12882 which vector contains the final result.
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-loop.c (vectorizable_induction): Treat the number
12889 of units as polynomial. Punt on SLP inductions. Use an integer
12890 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12891 cast of such a series for variable-length floating-point
12894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12895 Alan Hayward <alan.hayward@arm.com>
12896 David Sherwood <david.sherwood@arm.com>
12898 * tree.h (build_index_vector): Declare.
12899 * tree.c (build_index_vector): New function.
12900 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12901 of units as polynomial, forcibly converting it to a constant if
12902 vectorizable_reduction has already enforced the condition.
12903 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12904 to create a {1,2,3,...} vector.
12905 (vectorizable_reduction): Treat the number of units as polynomial.
12906 Choose vectype_in based on the largest scalar element size rather
12907 than the smallest number of units. Enforce the restrictions
12910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12911 Alan Hayward <alan.hayward@arm.com>
12912 David Sherwood <david.sherwood@arm.com>
12914 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12915 number of units as polynomial.
12917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12918 Alan Hayward <alan.hayward@arm.com>
12919 David Sherwood <david.sherwood@arm.com>
12921 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12922 * target.def (autovectorize_vector_sizes): Return the vector sizes
12923 by pointer, using vector_sizes rather than a bitmask.
12924 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12925 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12926 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12928 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12929 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12930 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12931 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12932 * omp-general.c (omp_max_vf): Likewise.
12933 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12934 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12935 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12936 * tree-vect-slp.c (vect_slp_bb): Likewise.
12937 * doc/tm.texi: Regenerate.
12938 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12940 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12941 the vector size as a poly_uint64 rather than an unsigned int.
12942 (current_vector_size): Change from an unsigned int to a poly_uint64.
12943 (get_vectype_for_scalar_type): Update accordingly.
12944 * tree.h (build_truth_vector_type): Take the size and number of
12945 units as a poly_uint64 rather than an unsigned int.
12946 (build_vector_type): Add a temporary overload that takes
12947 the number of units as a poly_uint64 rather than an unsigned int.
12948 * tree.c (make_vector_type): Likewise.
12949 (build_truth_vector_type): Take the number of units as a poly_uint64
12950 rather than an unsigned int.
12952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12953 Alan Hayward <alan.hayward@arm.com>
12954 David Sherwood <david.sherwood@arm.com>
12956 * target.def (get_mask_mode): Take the number of units and length
12957 as poly_uint64s rather than unsigned ints.
12958 * targhooks.h (default_get_mask_mode): Update accordingly.
12959 * targhooks.c (default_get_mask_mode): Likewise.
12960 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12961 * doc/tm.texi: Regenerate.
12963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12964 Alan Hayward <alan.hayward@arm.com>
12965 David Sherwood <david.sherwood@arm.com>
12967 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12968 * omp-general.c (omp_max_vf): Likewise.
12969 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12970 (expand_omp_simd): Handle polynomial safelen.
12971 * omp-low.c (omplow_simd_context): Add a default constructor.
12972 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12973 (lower_rec_simd_input_clauses): Update accordingly.
12974 (lower_rec_input_clauses): Likewise.
12976 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12977 Alan Hayward <alan.hayward@arm.com>
12978 David Sherwood <david.sherwood@arm.com>
12980 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12981 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12982 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12983 (vect_analyze_slp_cost): Likewise.
12984 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12985 (vect_model_load_cost): Likewise.
12987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12988 Alan Hayward <alan.hayward@arm.com>
12989 David Sherwood <david.sherwood@arm.com>
12991 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12992 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12993 from an unsigned int * to a poly_uint64_pod *.
12994 (calculate_unrolling_factor): New function.
12995 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12998 Alan Hayward <alan.hayward@arm.com>
12999 David Sherwood <david.sherwood@arm.com>
13001 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13002 from an unsigned int to a poly_uint64.
13003 (_loop_vec_info::slp_unrolling_factor): Likewise.
13004 (_loop_vec_info::vectorization_factor): Change from an int
13006 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13007 (vect_get_num_vectors): New function.
13008 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13009 (vect_get_num_copies): Use vect_get_num_vectors.
13010 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13011 to an unsigned int *.
13012 (vect_analyze_data_refs): Change min_vf from an int * to a
13014 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13015 than an unsigned HOST_WIDE_INT.
13016 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13017 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13018 to an unsigned int *.
13019 (vect_analyze_data_ref_dependences): Likewise.
13020 (vect_compute_data_ref_alignment): Handle polynomial vf.
13021 (vect_enhance_data_refs_alignment): Likewise.
13022 (vect_prune_runtime_alias_test_list): Likewise.
13023 (vect_shift_permute_load_chain): Likewise.
13024 (vect_supportable_dr_alignment): Likewise.
13025 (dependence_distance_ge_vf): Take the vectorization factor as a
13026 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13027 (vect_analyze_data_refs): Change min_vf from an int * to a
13029 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13030 vfm1 as a poly_uint64 rather than an int. Make the same change
13031 for the returned bound_scalar.
13032 (vect_gen_vector_loop_niters): Handle polynomial vf.
13033 (vect_do_peeling): Likewise. Update call to
13034 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13035 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13037 * tree-vect-loop.c (vect_determine_vectorization_factor)
13038 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13039 (vect_get_known_peeling_cost): Likewise.
13040 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13041 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13042 (vect_transform_loop): Likewise. Use the lowest possible VF when
13043 updating the upper bounds of the loop.
13044 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13045 rather than an int.
13046 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13047 polynomial unroll factors.
13048 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13049 (vect_make_slp_decision): Likewise.
13050 (vect_supported_load_permutation_p): Likewise, and polynomial
13052 (vect_analyze_slp_cost): Handle polynomial vf.
13053 (vect_slp_analyze_node_operations): Likewise.
13054 (vect_slp_analyze_bb_1): Likewise.
13055 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13056 than an unsigned HOST_WIDE_INT.
13057 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13058 (vectorizable_load): Handle polynomial vf.
13059 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13061 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13063 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13064 Alan Hayward <alan.hayward@arm.com>
13065 David Sherwood <david.sherwood@arm.com>
13067 * match.pd: Handle bit operations involving three constants
13068 and try to fold one pair.
13070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13072 * tree-vect-loop-manip.c: Include gimple-fold.h.
13073 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13074 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13075 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13076 Add a path that uses a step of VF instead of 1, but disable it
13078 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13079 and niters_no_overflow parameters. Update calls to
13080 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13081 Create a new SSA name if the latter choses to use a ste other
13082 than zero, and return it via niters_vector_mult_vf_var.
13083 * tree-vect-loop.c (vect_transform_loop): Update calls to
13084 vect_do_peeling, vect_gen_vector_loop_niters and
13085 slpeel_make_loop_iterate_ntimes.
13086 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13087 (vect_gen_vector_loop_niters): Update declarations after above changes.
13089 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13091 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13092 128-bit round to integer instructions.
13093 (ceil<mode>2): Likewise.
13094 (btrunc<mode>2): Likewise.
13095 (round<mode>2): Likewise.
13097 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13099 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13100 unaligned VSX load/store on P8/P9.
13101 (expand_block_clear): Allow the use of unaligned VSX
13102 load/store on P8/P9.
13104 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13106 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13108 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13109 swap associated with both a load and a store.
13111 2018-01-02 Andrew Waterman <andrew@sifive.com>
13113 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13114 * config/riscv/riscv.md (clear_cache): Use it.
13116 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13118 * web.c: Remove out-of-date comment.
13120 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13122 * expr.c (fixup_args_size_notes): Check that any existing
13123 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13124 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13125 (emit_single_push_insn): ...here.
13127 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13129 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13130 (const_vector_encoded_nelts): New function.
13131 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13132 (const_vector_int_elt, const_vector_elt): Declare.
13133 * emit-rtl.c (const_vector_int_elt_1): New function.
13134 (const_vector_elt): Likewise.
13135 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13136 of CONST_VECTOR_ELT.
13138 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13140 * expr.c: Include rtx-vector-builder.h.
13141 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13142 directly on the tree encoding.
13143 (const_vector_from_tree): Likewise.
13144 * optabs.c: Include rtx-vector-builder.h.
13145 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13146 sequence of "u" values.
13147 * vec-perm-indices.c: Include rtx-vector-builder.h.
13148 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13149 directly on the vec_perm_indices encoding.
13151 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13153 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13154 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13155 * rtx-vector-builder.h: New file.
13156 * rtx-vector-builder.c: Likewise.
13157 * rtl.h (rtx_def::u2): Add a const_vector field.
13158 (CONST_VECTOR_NPATTERNS): New macro.
13159 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13160 (CONST_VECTOR_DUPLICATE_P): Likewise.
13161 (CONST_VECTOR_STEPPED_P): Likewise.
13162 (CONST_VECTOR_ENCODED_ELT): Likewise.
13163 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13164 (unwrap_const_vec_duplicate): Likewise.
13165 (const_vec_series_p): Check for a non-duplicated vector encoding.
13166 Say that the function only returns true for integer vectors.
13167 * emit-rtl.c: Include rtx-vector-builder.h.
13168 (gen_const_vec_duplicate_1): Delete.
13169 (gen_const_vector): Call gen_const_vec_duplicate instead of
13170 gen_const_vec_duplicate_1.
13171 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13172 (gen_const_vec_duplicate): Use rtx_vector_builder.
13173 (gen_const_vec_series): Likewise.
13174 (gen_rtx_CONST_VECTOR): Likewise.
13175 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13176 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13177 Build a new vector rather than modifying a CONST_VECTOR in-place.
13178 (handle_special_swappables): Update call accordingly.
13179 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13180 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13181 Build a new vector rather than modifying a CONST_VECTOR in-place.
13182 (handle_special_swappables): Update call accordingly.
13184 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13186 * simplify-rtx.c (simplify_const_binary_operation): Use
13187 CONST_VECTOR_ELT instead of XVECEXP.
13189 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13191 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13192 the selector elements to be different from the data elements
13193 if the selector is a VECTOR_CST.
13194 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13195 ssizetype for the selector.
13197 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13199 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13200 before testing each element individually.
13201 * tree-vect-generic.c (lower_vec_perm): Likewise.
13203 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13205 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13206 * selftest-run-tests.c (selftest::run_tests): Call it.
13207 * vector-builder.h (vector_builder::operator ==): New function.
13208 (vector_builder::operator !=): Likewise.
13209 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13210 (vec_perm_indices::all_from_input_p): New function.
13211 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13212 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13213 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13214 instead of reading the VECTOR_CST directly. Detect whether both
13215 vector inputs are the same before constructing the vec_perm_indices,
13216 and update the number of inputs argument accordingly. Use the
13217 utility functions added above. Only construct sel2 if we need to.
13219 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13221 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13222 the broadcast of the low byte.
13223 (expand_mult_highpart): Use an explicit encoding for the permutes.
13224 * optabs-query.c (can_mult_highpart_p): Likewise.
13225 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13226 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13227 (vectorizable_bswap): Likewise.
13228 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13229 explicit encoding for the power-of-2 permutes.
13230 (vect_permute_store_chain): Likewise.
13231 (vect_grouped_load_supported): Likewise.
13232 (vect_permute_load_chain): Likewise.
13234 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13236 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13237 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13238 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13239 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13240 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13241 (vect_gen_perm_mask_any): Likewise.
13243 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13245 * int-vector-builder.h: New file.
13246 * vec-perm-indices.h: Include int-vector-builder.h.
13247 (vec_perm_indices): Redefine as an int_vector_builder.
13248 (auto_vec_perm_indices): Delete.
13249 (vec_perm_builder): Redefine as a stand-alone class.
13250 (vec_perm_indices::vec_perm_indices): New function.
13251 (vec_perm_indices::clamp): Likewise.
13252 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13253 (vec_perm_indices::new_vector): New function.
13254 (vec_perm_indices::new_expanded_vector): Update for new
13255 vec_perm_indices class.
13256 (vec_perm_indices::rotate_inputs): New function.
13257 (vec_perm_indices::all_in_range_p): Operate directly on the
13258 encoded form, without computing elided elements.
13259 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13260 encoding. Update for new vec_perm_indices class.
13261 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13262 the given vec_perm_builder.
13263 (expand_vec_perm_var): Update vec_perm_builder constructor.
13264 (expand_mult_highpart): Use vec_perm_builder instead of
13265 auto_vec_perm_indices.
13266 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13267 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13268 or double series encoding as appropriate.
13269 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13270 vec_perm_indices instead of auto_vec_perm_indices.
13271 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13272 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13273 (vect_permute_store_chain): Likewise.
13274 (vect_grouped_load_supported): Likewise.
13275 (vect_permute_load_chain): Likewise.
13276 (vect_shift_permute_load_chain): Likewise.
13277 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13278 (vect_transform_slp_perm_load): Likewise.
13279 (vect_schedule_slp_instance): Likewise.
13280 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13281 (vectorizable_mask_load_store): Likewise.
13282 (vectorizable_bswap): Likewise.
13283 (vectorizable_store): Likewise.
13284 (vectorizable_load): Likewise.
13285 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13286 vec_perm_indices instead of auto_vec_perm_indices. Use
13287 tree_to_vec_perm_builder to read the vector from a tree.
13288 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13289 vec_perm_builder instead of a vec_perm_indices.
13290 (have_whole_vector_shift): Use vec_perm_builder and
13291 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13292 truncation to calc_vec_perm_mask_for_shift.
13293 (vect_create_epilog_for_reduction): Likewise.
13294 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13295 from auto_vec_perm_indices to vec_perm_indices.
13296 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13297 instead of changing individual elements.
13298 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13299 the vector in d.perm.
13300 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13301 from auto_vec_perm_indices to vec_perm_indices.
13302 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13303 instead of changing individual elements.
13304 (arm_vectorize_vec_perm_const): Use new_vector to install
13305 the vector in d.perm.
13306 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13307 Update vec_perm_builder constructor.
13308 (rs6000_expand_interleave): Likewise.
13309 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13310 (rs6000_expand_interleave): Likewise.
13312 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13314 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13315 to qimode could truncate the indices.
13316 * optabs.c (expand_vec_perm_var): Likewise.
13318 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13320 * Makefile.in (OBJS): Add vec-perm-indices.o.
13321 * vec-perm-indices.h: New file.
13322 * vec-perm-indices.c: Likewise.
13323 * target.h (vec_perm_indices): Replace with a forward class
13325 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13326 * optabs.h: Include vec-perm-indices.h.
13327 (expand_vec_perm): Delete.
13328 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13329 (expand_vec_perm_const): Declare.
13330 * target.def (vec_perm_const_ok): Replace with...
13331 (vec_perm_const): ...this new hook.
13332 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13333 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13334 * doc/tm.texi: Regenerate.
13335 * optabs.def (vec_perm_const): Delete.
13336 * doc/md.texi (vec_perm_const): Likewise.
13337 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13338 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13339 expand_vec_perm for constant permutation vectors. Assert that
13340 the mode of variable permutation vectors is the integer equivalent
13341 of the mode that is being permuted.
13342 * optabs-query.h (selector_fits_mode_p): Declare.
13343 * optabs-query.c: Include vec-perm-indices.h.
13344 (selector_fits_mode_p): New function.
13345 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13346 is defined, instead of checking whether the vec_perm_const_optab
13347 exists. Use targetm.vectorize.vec_perm_const instead of
13348 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13349 fit in the vector mode before using a variable permute.
13350 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13351 vec_perm_indices instead of an rtx.
13352 (expand_vec_perm): Replace with...
13353 (expand_vec_perm_const): ...this new function. Take the selector
13354 as a vec_perm_indices rather than an rtx. Also take the mode of
13355 the selector. Update call to shift_amt_for_vec_perm_mask.
13356 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13357 Use vec_perm_indices::new_expanded_vector to expand the original
13358 selector into bytes. Check whether the indices fit in the vector
13359 mode before using a variable permute.
13360 (expand_vec_perm_var): Make global.
13361 (expand_mult_highpart): Use expand_vec_perm_const.
13362 * fold-const.c: Includes vec-perm-indices.h.
13363 * tree-ssa-forwprop.c: Likewise.
13364 * tree-vect-data-refs.c: Likewise.
13365 * tree-vect-generic.c: Likewise.
13366 * tree-vect-loop.c: Likewise.
13367 * tree-vect-slp.c: Likewise.
13368 * tree-vect-stmts.c: Likewise.
13369 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13371 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13372 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13373 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13374 (aarch64_vectorize_vec_perm_const): ...this new function.
13375 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13376 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13377 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13378 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13379 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13380 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13381 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13383 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13384 check for NEON modes.
13385 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13386 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13387 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13388 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13390 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13391 the old VEC_PERM_CONST conditions.
13392 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13393 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13394 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13395 (ia64_vectorize_vec_perm_const_ok): Merge into...
13396 (ia64_vectorize_vec_perm_const): ...this new function.
13397 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13398 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13399 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13400 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13401 * config/mips/mips.c (mips_expand_vec_perm_const)
13402 (mips_vectorize_vec_perm_const_ok): Merge into...
13403 (mips_vectorize_vec_perm_const): ...this new function.
13404 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13405 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13406 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13407 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13408 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13409 (rs6000_expand_vec_perm_const): Delete.
13410 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13412 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13413 (altivec_expand_vec_perm_const_le): Take each operand individually.
13414 Operate on constant selectors rather than rtxes.
13415 (altivec_expand_vec_perm_const): Likewise. Update call to
13416 altivec_expand_vec_perm_const_le.
13417 (rs6000_expand_vec_perm_const): Delete.
13418 (rs6000_vectorize_vec_perm_const_ok): Delete.
13419 (rs6000_vectorize_vec_perm_const): New function.
13420 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13421 an element count and rtx array.
13422 (rs6000_expand_extract_even): Update call accordingly.
13423 (rs6000_expand_interleave): Likewise.
13424 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13425 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13426 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13427 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13428 (rs6000_expand_vec_perm_const): Delete.
13429 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13430 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13431 (altivec_expand_vec_perm_const_le): Take each operand individually.
13432 Operate on constant selectors rather than rtxes.
13433 (altivec_expand_vec_perm_const): Likewise. Update call to
13434 altivec_expand_vec_perm_const_le.
13435 (rs6000_expand_vec_perm_const): Delete.
13436 (rs6000_vectorize_vec_perm_const_ok): Delete.
13437 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13438 reference to the SPE evmerge intructions.
13439 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13440 an element count and rtx array.
13441 (rs6000_expand_extract_even): Update call accordingly.
13442 (rs6000_expand_interleave): Likewise.
13443 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13444 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13446 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13448 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13450 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13451 vector mode and that that mode matches the mode of the data
13453 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13454 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13455 directly using expand_vec_perm_1 when forcing selectors into
13457 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13459 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13461 * optabs-query.h (can_vec_perm_p): Delete.
13462 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13463 * optabs-query.c (can_vec_perm_p): Split into...
13464 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13465 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13466 particular selector is valid.
13467 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13468 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13469 (vect_grouped_load_supported): Likewise.
13470 (vect_shift_permute_load_chain): Likewise.
13471 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13472 (vect_transform_slp_perm_load): Likewise.
13473 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13474 (vectorizable_bswap): Likewise.
13475 (vect_gen_perm_mask_checked): Likewise.
13476 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13477 implementations of variable permutation vectors into account
13478 when deciding which selector to use.
13479 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13480 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13481 with a false third argument.
13482 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13483 to test whether the constant selector is valid and can_vec_perm_var_p
13484 to test whether a variable selector is valid.
13486 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13488 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13489 * optabs-query.c (can_vec_perm_p): Likewise.
13490 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13491 instead of vec_perm_indices.
13492 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13493 (vect_gen_perm_mask_checked): Likewise,
13494 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13495 (vect_gen_perm_mask_checked): Likewise,
13497 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13499 * optabs-query.h (qimode_for_vec_perm): Declare.
13500 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13501 (qimode_for_vec_perm): ...this new function.
13502 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13504 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13506 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13507 does not have a conditional at the top.
13509 2018-01-02 Richard Biener <rguenther@suse.de>
13511 * ipa-inline.c (big_speedup_p): Fix expression.
13513 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13516 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13519 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13523 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13524 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13525 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13526 cond_taken_branch_cost 3->4.
13528 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13530 PR tree-optimization/83581
13531 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13532 TODO_cleanup_cfg if any changes have been made.
13534 PR middle-end/83608
13535 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13536 convert_modes if target mode has the right side, but different mode
13539 PR middle-end/83609
13540 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13541 last argument when extracting from CONCAT. If either from_real or
13542 from_imag is NULL, use expansion through memory. If result is not
13543 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13544 the parts directly to inner mode, if even that fails, use expansion
13547 PR middle-end/83623
13548 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13549 check for bswap in mode rather than HImode and use that in expand_unop
13552 Copyright (C) 2018 Free Software Foundation, Inc.
13554 Copying and distribution of this file, with or without modification,
13555 are permitted in any medium without royalty provided the copyright
13556 notice and this notice are preserved.