1 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3 * config/nds32/nds32-fpu.md: Update copyright year.
5 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
9 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11 * config/nds32/nds32.c
12 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
13 * config/nds32/nds32.opt (minline-asm-r15): New option.
15 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
17 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
19 * config/nds32/nds32.md (abssi2): New pattern.
21 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
23 * config/i386/i386.md (rex64namesuffix): New mode attribute.
24 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
25 Merge insn pattern from sse_cvtsi2ss<round_name> and
26 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
27 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
28 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
29 using SWI48 mode iterator.
30 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
31 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
32 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
33 pattern from sse_cvttss2si<round_saeonly_name>
34 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
35 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
36 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
37 using SWI48 mode iterator.
38 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
39 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
40 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
41 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
42 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
43 using SWI48 mode iterator.
44 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
45 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
46 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
47 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
48 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
50 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
51 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
52 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
53 pattern from sse_cvttsd2si<round_saeonly_name>
54 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
56 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
58 * config/nds32/nds32-md-auxiliary.c
59 (nds32_valid_smw_lwm_base_p): Refine.
60 (nds32_output_smw_single_word): Refine.
61 (nds32_output_smw_double_word): New.
62 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
64 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
66 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
67 (nds32_output_stack_pop): Refine.
68 (nds32_expand_unaligned_load): Refine.
69 (nds32_expand_unaligned_store): Refine.
71 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
72 Chung-Ju Wu <jasonwucj@gmail.com>
74 * config/nds32/constants.md: Add TP_REGNUM constant.
75 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
76 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
78 * config/nds32/nds32-doubleword.md: Consider flag_pic.
79 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
80 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
81 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
82 and PIC code generation.
83 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
85 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
87 * config/nds32/nds32.md: Support TLS and PIC.
88 * config/nds32/nds32.c: Support TLS and PIC.
89 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
90 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
93 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
95 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
98 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
99 Chung-Ju Wu <jasonwucj@gmail.com>
101 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
102 * config/nds32/nds32-md-auxiliary.c
103 (symbolic_reference_mentioned_p): New.
104 (nds32_legitimize_ict_address): New.
105 (nds32_expand_ict_move): New.
106 (nds32_indirect_call_referenced_p): New.
107 (nds32_symbol_binds_local_p): Delete.
108 (nds32_long_call_p): Modify.
109 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
110 * config/nds32/nds32-protos.h
111 (symbolic_reference_mentioned_p): Declare.
112 (nds32_legitimize_ict_address): Declare.
113 (nds32_expand_ict_move): Declare.
114 (nds32_indirect_call_referenced_p): Declare.
115 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
116 (nds32_relax_group): Use nds32_ict_const_p as condition.
117 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
118 (nds32_asm_file_start): Output ict_model directive in asm code.
119 (nds32_legitimate_address_p): Consider indirect call.
120 (nds32_print_operand): Consider indirect call.
121 (nds32_print_operand_address): Consider indirect call.
122 (nds32_insert_attributes): Handle "indirect_call" attribute.
123 (TARGET_LEGITIMATE_ADDRESS_P): Define.
124 (TARGET_LEGITIMATE_CONSTANT_P): Define.
125 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
126 (TARGET_DELEGITIMIZE_ADDRESS): Define.
127 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
128 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
129 (TARGET_ICT_MODEL_SMALL): Define.
130 (TARGET_ICT_MODEL_LARGE): Define.
131 * config/nds32/nds32.md (movsi): Consider ict model.
132 (call, call_value): Consider ict model.
133 (sibcall, sibcall_value): Consider ict model.
134 * config/nds32/nds32.opt (mict-model): New option.
135 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
138 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
139 Monk Chiang <sh.chiang04@gmail.com>
140 Jim Wilson <jimw@sifive.com>
142 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
143 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
144 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
145 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
146 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
147 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
148 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
149 compute save_libcall_adjustment properly.
150 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
151 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
152 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
153 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
154 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
155 (ABI_SPEC): Handle mabi=ilp32e.
156 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
158 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
159 <-march>: Add rv32e as an example.
161 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
164 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
165 (intra_create_variable_infos): Handle C++ constructors.
167 2018-05-18 Martin Liska <mliska@suse.cz>
169 * passes.def: Remove a redundant pass.
171 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
174 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
176 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
178 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
180 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
182 (armv2, armv2a, armv3, armv3m): Delete architectures.
183 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
184 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
185 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
187 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
188 (*mulsidi3adddi): Likewise.
189 (mulsidi3): Likewise.
190 (*mulsidi3_nov6): Likewise.
191 (umulsidi3): Likewise.
192 (umulsidi3_nov6): Likewise.
193 (umaddsidi4): Likewise.
194 (*umulsidi3adddi): Likewise.
195 (smulsi3_highpart): Likewise.
196 (*smulsi3_highpart_nov6): Likewise.
197 (umulsi3_highpart): Likewise.
198 (*umulsi3_highpart_nov6): Likewise.
199 * config/arm/arm.h (arm_arch3m): Delete.
200 * config/arm/arm.c (arm_arch3m): Delete.
201 (arm_option_override_internal): Update armv3-related comment.
202 (arm_configure_build_target): Delete use of isa_bit_mode32.
203 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
204 (arm_rtx_costs_internal): Delete check of arm_arch3m.
205 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
208 * config/arm/arm-protos.h (arm_arch3m): Delete.
209 * config/arm/arm-tables.opt: Regenerate.
210 * config/arm/arm-tune.md: Likewise.
211 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
212 deleted architectures.
214 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
216 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
217 (armv5t, armv5te): New features.
218 (ARMv5, ARMv5e): Delete fgroups.
219 (ARMv5t, ARMv5te): Adjust for above changes.
221 (armv5, armv5e): Delete arches.
222 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
224 (*call_reg_arm): Likewise.
225 (*call_value_reg_armv5): Likewise.
226 (*call_value_reg_arm): Likewise.
227 (*call_symbol): Likewise.
228 (*call_value_symbol): Likewise.
229 (*sibcall_insn): Likewise.
230 (*sibcall_value_insn): Likewise.
232 (prefetch): Likewise.
233 (define_split and define_peephole2 dependent on arm_arch5):
235 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
237 (TARGET_ARM_QBIT): Likewise.
238 (TARGET_DSP_MULTIPLY): Likewise.
239 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
240 (arm_arch5, arm_arch5e): Delete.
241 (arm_arch5t, arm_arch5te): Declare.
242 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
243 (arm_arch5t): Declare.
244 (arm_option_reconfigure_globals): Update for the above.
245 (arm_options_perform_arch_sanity_checks): Update comment, replace
246 use of arm_arch5 with arm_arch5t.
247 (use_return_insn): Likewise.
248 (arm_emit_call_insn): Likewise.
249 (output_return_instruction): Likewise.
250 (arm_final_prescan_insn): Likewise.
251 (arm_coproc_builtin_available): Likewise.
252 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
253 arm_arch5e with arm_arch5t and arm_arch5te.
254 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
255 (arm_arch5t, arm_arch5te): Declare.
256 * config/arm/arm-tables.opt: Regenerate.
257 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
258 * config/arm/t-multilib: Likewise.
259 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
260 instead of arm_arch5.
261 (*call_reg_thumb1): Likewise.
262 (*call_value_reg_thumb1_v5): Likewise.
263 (*call_value_reg_thumb1): Likewise.
264 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
266 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
268 2018-05-18 Martin Liska <mliska@suse.cz>
270 PR gcov-profile/84846
271 * doc/gcov.texi: Document -t option of gcov tool.
273 2018-05-18 Martin Liska <mliska@suse.cz>
275 PR gcov-profile/84846
276 * gcov.c (print_usage): Add new -t option.
277 (process_args): Handle the option.
278 (generate_results): Use stdout as output when requested by
281 2018-05-18 Martin Liska <mliska@suse.cz>
283 PR gcov-profile/84846
284 * coverage.c (coverage_init): Write PWD to .gcno file.
285 * doc/gcov.texi: Document how working directory is printed.
286 * gcov-dump.c (dump_gcov_file): Print PWD.
287 * gcov.c (output_intermediate_file): Likewise.
288 (read_graph_file): Read PWD string.
289 (output_lines): Print PWD.
291 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
294 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
295 for retval and return false if all args to phi are zero.
297 2018-05-18 Richard Biener <rguenther@suse.de>
299 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
301 (evrp_dom_walker::before_dom_children): Call it.
303 2018-05-18 Richard Biener <rguenther@suse.de>
305 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
306 results when processing array refs with variable index.
308 2018-05-18 Toon Moene <toon@moene.org>
310 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
311 directly after that of -floop-interchange. Indicate that both
312 options are enabled by default when specifying -O3.
314 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
316 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
317 iterator. Delete separate integer-mode vec_set<mode> expander.
318 (aarch64_simd_vec_setv2di): Delete.
319 (vec_setv2di): Delete.
320 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
321 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
322 the "w, r" alternative.
324 2018-05-18 Martin Liska <mliska@suse.cz>
326 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
327 * tree-pass.h (make_pass_lower_switch_O0): New function.
328 * tree-switch-conversion.c (node_has_low_bound): Remove.
329 (node_has_high_bound): Likewise.
330 (node_is_bounded): Likewise.
331 (class pass_lower_switch): Make it a template type and create
333 (pass_lower_switch::execute): Add template argument.
334 (make_pass_lower_switch): New function.
335 (make_pass_lower_switch_O0): New function.
336 (do_jump_if_equal): Remove.
337 (emit_case_nodes): Simplify to just handle all 3 cases and leave
338 all the hard work to tree optimization passes.
340 2018-05-18 Martin Liska <mliska@suse.cz>
342 * dbgcnt.c (limit_low): Renamed from limit.
343 (limit_high): New variable.
344 (dbg_cnt_is_enabled): Check for upper limit.
345 (dbg_cnt): Adjust dumping.
346 (dbg_cnt_set_limit_by_index): Add new argument for high
348 (dbg_cnt_set_limit_by_name): Likewise.
349 (dbg_cnt_process_single_pair): Parse new format.
350 (dbg_cnt_process_opt): Use strtok.
351 (dbg_cnt_list_all_counters): Remove 'value' and add
353 * doc/invoke.texi: Document changes.
355 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
357 * doc/sourcebuild.texi (scalar_all_fma): Document.
358 * tree.def (FMA_EXPR): Delete.
359 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
360 * internal-fn.c (ternary_direct): New macro.
361 (expand_ternary_optab_fn): Likewise.
362 (direct_ternary_optab_supported_p): Likewise.
363 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
364 * builtins.c (fold_builtin_fma): Delete.
365 (fold_builtin_3): Don't call it.
366 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
367 * expr.c (expand_expr_real_2): Likewise.
368 * fold-const.c (operand_equal_p): Likewise.
369 (fold_ternary_loc): Likewise.
370 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
371 * gimple.c (DEFTREECODE): Likewise.
372 * gimplify.c (gimplify_expr): Likewise.
373 * optabs-tree.c (optab_for_tree_code): Likewise.
374 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
375 * tree-eh.c (operation_could_trap_p): Likewise.
376 (stmt_could_throw_1_p): Likewise.
377 * tree-inline.c (estimate_operator_cost): Likewise.
378 * tree-pretty-print.c (dump_generic_node): Likewise.
379 (op_code_prio): Likewise.
380 * tree-ssa-loop-im.c (stmt_cost): Likewise.
381 * tree-ssa-operands.c (get_expr_operands): Likewise.
382 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
383 * fold-const-call.h (fold_fma): Delete.
384 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
385 CFN_FNMA and CFN_FNMS.
387 * genmatch.c (combined_fn): New enum.
388 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
389 (commutative_op): New function.
390 (commutate): Use it. Handle more than 2 operands.
391 (dt_operand::gen_gimple_expr): Use commutative_op.
392 (parser::parse_expr): Allow :c to be used with non-binary
393 operators if the commutative operand is known.
394 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
395 CFN_FMS, CFN_FNMA and CFN_FNMS.
396 (backprop::process_assign_use): Remove FMA_EXPR handling.
397 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
398 (gen_hsa_fma): New function.
399 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
400 IFN_FNMA and IFN_FNMS.
401 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
402 * gimple-fold.h (follow_all_ssa_edges): Declare.
403 * gimple-fold.c (follow_all_ssa_edges): New function.
404 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
405 gimple_build interface and use follow_all_ssa_edges to fold the result.
406 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
407 instead of checking for optabs directly.
408 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
409 rather than FMA_EXPRs.
410 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
411 call to IFN_FMA instead of an FMA_EXPR.
413 2018-05-17 Jim Wilson <jimw@sifive.com>
415 * expr.c (do_tablejump): When converting index to Pmode, if we have a
416 sign extended promoted subreg, and the range does not have the sign bit
417 set, then do a sign extend.
419 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
420 test, check for sign extended subreg and/or constant operands, and
421 do a sign extend in that case.
423 2018-05-17 Steve Ellcey <sellcey@cavium.com>
425 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
426 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
428 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
429 Change logics_shift_reg to logics_shift_imm.
430 (thunderx2t99_fp_loadpair_basic): Delete.
431 (thunderx2t99_fp_storepair_basic): Delete.
432 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
433 (thunderx2t99_asimd_polynomial): Delete.
434 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
435 and neon_fp_mul_d_scalar_q.
436 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
437 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
438 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
439 (thunderx2t99_asimd_lut): Add missing tbl types.
440 (thunderx2t99_asimd_ext): Delete.
441 (thunderx2t99_asimd_load1_1_mult): Delete.
442 (thunderx2t99_asimd_load1_2_mult): Delete.
443 (thunderx2t99_asimd_load1_ldp): New.
444 (thunderx2t99_asimd_load1): New.
445 (thunderx2t99_asimd_load2): Add missing *load2* types.
446 (thunderx2t99_asimd_load3): New.
447 (thunderx2t99_asimd_load4): New.
448 (thunderx2t99_asimd_store1_1_mult): Delete.
449 (thunderx2t99_asimd_store1_2_mult): Delete.
450 (thunderx2t99_asimd_store2_mult): Delete.
451 (thunderx2t99_asimd_store2_onelane): Delete.
452 (thunderx2t99_asimd_store_stp): New.
453 (thunderx2t99_asimd_store1): New.
454 (thunderx2t99_asimd_store2): New.
455 (thunderx2t99_asimd_store3): New.
456 (thunderx2t99_asimd_store4): New.
458 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
460 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
461 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
463 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
464 Segher Boessenkool <segher@kernel.crashing.org>
467 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
470 2018-05-17 Richard Biener <rguenther@suse.de>
472 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
473 for pruning loop and prune defs feeding only already visited PHIs.
475 2018-05-17 Richard Biener <rguenther@suse.de>
477 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
479 2018-05-17 Bin Cheng <bin.cheng@arm.com>
480 Richard Biener <rguenther@suse.de>
482 PR tree-optimization/85793
483 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
484 for VMAT_ELEMENTWISE.
486 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
488 * internal-fn.h (lookup_internal_fn): Declare
489 * internal-fn.c (lookup_internal_fn): New function.
490 * gimple.c (gimple_build_call_from_tree): Handle calls to
492 * gimple-pretty-print.c (dump_gimple_call): Print "." before
493 internal function names.
494 * tree-pretty-print.c (dump_generic_node): Likewise.
495 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
497 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
499 * gimple-fold.h (gimple_build): Make the function forms take
500 combined_fn rather than built_in_function.
501 (gimple_simplify): Likewise.
502 * gimple-match-head.c (gimple_simplify): Likewise.
503 * gimple-fold.c (gimple_build): Likewise.
504 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
505 rather than gimple_build_call_internal.
506 (get_initial_defs_for_reduction): Likewise.
507 (vect_create_epilog_for_reduction): Likewise.
508 (vectorizable_live_operation): Likewise.
510 2018-05-17 Martin Liska <mliska@suse.cz>
512 * gimple-ssa-sprintf.c (format_directive): Do not use
513 space in between 'G_' and '('.
515 2018-05-17 Jakub Jelinek <jakub@redhat.com>
518 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
519 even if the mask is not all ones.
522 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
524 (ix86_gimple_fold_builtin): Likewise.
527 * config/i386/i386.c: Include tree-vector-builder.h.
528 (ix86_vector_shift_count): New function.
529 (ix86_fold_builtin): Fold shift builtins by scalar count.
530 (ix86_gimple_fold_builtin): Likewise.
532 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
533 _mm512_setzero): New intrinsics.
535 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
536 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
538 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
539 code generation for cases where splatting a value is not useful.
540 * simplify-rtx.c (simplify_ternary_operation): Simplify
541 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
542 mode to a vec_concat.
544 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
546 * config.gcc: Support "goldmont-plus".
547 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
549 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
550 PROCESSOR_GOLDMONT_PLUS.
551 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
552 (processor_target_table): Add "goldmont-plus".
553 (PTA_GOLDMONT_PLUS): Define.
554 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
555 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
556 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
557 (fold_builtin_cpu): Add "goldmont-plus".
558 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
559 (ix86_option_override_internal): Add "goldmont-plus".
560 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
561 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
562 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
563 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
565 2018-05-17 Richard Biener <rguenther@suse.de>
567 PR tree-optimization/85757
568 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
569 remove defs that only feed that PHI from further processing.
571 2018-05-16 Jim Wilson <jimw@sifive.com>
573 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
575 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
577 2018-05-16 Mark Wielaard <mark@klomp.org>
579 * dwarf2out.c (count_index_strings): New function.
580 (output_indirect_strings): Call count_index_strings and generate
581 header for dwarf_version >= 5.
583 2018-05-16 Mark Wielaard <mark@klomp.org>
585 * dwarf2out.c (dwarf_FORM): New function.
586 (set_indirect_string): Use dwarf_FORM.
587 (reset_indirect_string): Likewise.
588 (size_of_die): Likewise.
589 (value_format): Likewise.
590 (output_die): Likewise.
591 (add_skeleton_AT_string): Likewise.
592 (output_macinfo_op): Likewise.
593 (index_string): Likewise.
594 (output_index_string_offset): Likewise.
595 (output_index_string): Likewise.
596 (count_index_strings): Likewise.
598 2018-05-16 Carl Love <cel@us.ibm.com>
600 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
601 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
603 2018-05-16 Martin Jambor <mjambor@suse.cz>
605 * ipa-prop.c (ipa_free_all_edge_args): Remove.
606 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
608 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
610 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
611 (fnma<mode>4): Likewise.
612 (fms<mode>4): Likewise.
613 (fnms<mode>4): Likewise.
614 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
615 (aarch64_fnma<mode>4): Likewise.
616 (aarch64_fms<mode>4): Likewise.
617 (aarch64_fnms<mode>4): Likewise.
618 (aarch64_fnmadd<mode>4): Likewise.
620 2018-05-16 Jason Merrill <jason@redhat.com>
622 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
624 2018-05-16 Richard Biener <rguenther@suse.de>
626 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
627 (dump_stmt_cost): Declare.
628 (add_stmt_cost): Dump cost we add.
629 (add_stmt_costs): New function.
630 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
632 (vect_analyze_stmt): Adjust prototype.
633 (vectorizable_condition): Likewise.
634 (vectorizable_live_operation): Likewise.
635 (vectorizable_reduction): Likewise.
636 (vectorizable_induction): Likewise.
637 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
638 cost vector to pass to vectorizable_ and record afterwards.
639 (vect_model_reduction_cost): Take cost vector argument and adjust.
640 (vect_model_induction_cost): Likewise.
641 (vectorizable_reduction): Likewise.
642 (vectorizable_induction): Likewise.
643 (vectorizable_live_operation): Likewise.
644 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
645 SLP_TREE_NUMBER_OF_VEC_STMTS.
646 (vect_analyze_slp_cost_1): Remove.
647 (vect_analyze_slp_cost): Likewise.
648 (vect_slp_analyze_node_operations): Take visited args and
649 a target cost vector. Avoid processing already visited stmt sets.
650 (vect_slp_analyze_operations): Use a local cost vector to gather
651 costs and register those of non-discarded instances.
652 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
653 (vect_schedule_slp_instance): Remove copying of
654 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
656 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
657 adding cost. Record cost entry location.
658 (vect_prologue_cost_for_slp_op): Function to compute cost of
659 a constant or invariant generated for SLP vect in the prologue,
660 split out from vect_analyze_slp_cost_1.
661 (vect_model_simple_cost): Make static. Adjust for SLP costing.
662 (vect_model_promotion_demotion_cost): Likewise.
663 (vect_model_store_cost): Likewise, make static.
664 (vect_model_load_cost): Likewise.
665 (vectorizable_bswap): Add cost vector arg and adjust.
666 (vectorizable_call): Likewise.
667 (vectorizable_simd_clone_call): Likewise.
668 (vectorizable_conversion): Likewise.
669 (vectorizable_assignment): Likewise.
670 (vectorizable_shift): Likewise.
671 (vectorizable_operation): Likewise.
672 (vectorizable_store): Likewise.
673 (vectorizable_load): Likewise.
674 (vectorizable_condition): Likewise.
675 (vectorizable_comparison): Likewise.
676 (can_vectorize_live_stmts): Likewise.
677 (vect_analyze_stmt): Likewise.
678 (vect_transform_stmt): Adjust calls to vectorizable_*.
679 * tree-vectorizer.c: Include gimple-pretty-print.h.
680 (dump_stmt_cost): New function.
682 2018-05-16 Richard Biener <rguenther@suse.de>
684 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
685 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
686 * tree-ssa-dse.c: Include tree-ssa-loop.h.
687 (check_name): New callback.
688 (dse_classify_store): Track cycles via a visited bitmap of PHI
689 defs and simplify handling of in-loop and across loop dead stores
690 and properly fail for loop-variant refs. Handle byte-tracking with
691 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
694 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
696 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
697 (vect_get_mask_type_for_stmt): Likewise.
698 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
700 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
701 to determine the statement's vector type and the vector type that
702 should be used for calculating nunits. Deal with cases in which
703 the type has to be deferred.
704 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
705 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
706 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
707 (vect_determine_vf_for_stmt): New functions, split out from...
708 (vect_determine_vectorization_factor): ...here.
709 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
710 (vect_get_mask_type_for_stmt): New functions, split out from
711 vect_determine_vectorization_factor.
713 2018-05-16 Richard Biener <rguenther@suse.de>
715 * tree-cfg.c (verify_gimple_assign_ternary): Properly
716 verify the [VEC_]COND_EXPR embedded comparison.
718 2018-05-15 Martin Sebor <msebor@redhat.com>
720 PR tree-optimization/85753
721 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
722 RECORD_TYPE in addition to ARRAY_TYPE.
724 2018-05-15 Martin Sebor <msebor@redhat.com>
727 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
729 2018-05-15 Richard Biener <rguenther@suse.de>
731 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
732 add by_clobber_p one. Change algorithm to collect all defs
733 representing uses we need to walk and try reducing them to
734 a single one before failing.
735 (dse_dom_walker::dse_optimize_stmt): Adjust.
737 2018-05-13 Mark Wielaard <mark@klomp.org>
739 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
740 (size_of_loc_descr): Likewise.
741 (output_loc_operands): Likewise.
742 (output_loc_operands_raw): Likewise.
743 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
744 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
745 (hash_loc_operands): Likewise.
746 (compare_loc_operands): Likewise.
748 2018-05-14 Mark Wielaard <mark@klomp.org>
750 * dwarf2out.c (count_index_addrs): New function.
751 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
753 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
755 PR tree-optimization/83648
756 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
757 return value as malloc candidate.
759 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
762 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
763 as true in call to suggest_attribute.
765 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
767 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
770 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
772 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
774 (aarch64_crypto_aesd_fused): Likewise.
776 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
778 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
779 (movsi_aarch64): Likewise.
780 (load_pairsi): Likewise.
781 (load_pairdi): Likewise.
782 (store_pairsi): Likewise.
783 (store_pairdi): Likewise.
784 (load_pairsf): Likewise.
785 (load_pairdf): Likewise.
786 (store_pairsf): Likewise.
787 (store_pairdf): Likewise.
788 (zero_extend): Likewise.
789 (trunc): Swap alternatives.
790 (fcvt_target): Add '?' to prefer w over r.
792 2018-05-14 Jakub Jelinek <jakub@redhat.com>
795 * config/i386/i386.md: Disallow non-commutative arithmetics in
796 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
797 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
798 in the peephole2 before it.
800 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
802 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
803 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
804 (ix86_handle_option): Handle -mcldemote.
805 * config.gcc: New header.
806 * config/i386/cldemoteintrin.h: New file.
807 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
808 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
810 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
811 OPTION_MASK_ISA_CLDEMOTE.
812 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
813 (ix86_valid_target_attribute_inner_p): Ditto.
814 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
815 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
816 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
817 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
818 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
820 * config/i386/i386.opt: Add -mcldemote.
821 * config/i386/x86intrin.h: New header.
822 * doc/invoke.texi: Add -mcldemote.
824 2018-05-14 Richard Biener <rguenther@suse.de>
826 * doc/match-and-simplify.texi: Adjust :s documentation.
828 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
830 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
831 intended memcpy size.
832 (REORDER_45): Likewise.
834 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
837 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
838 * vec.c (qsort_chk): Use gcc_qsort.
839 * Makefile.in (OBJS-libcommon): Add sort.o.
840 (build/sort.o): New target. Use it...
841 (BUILD_RTL): ... here, and...
842 (build/gencfn-macros): ... here, and...
843 (build/genmatch): ... here.
845 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
846 Chung-Ju Wu <jasonwucj@gmail.com>
848 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
849 * config/nds32/nds32-graywolf.md: New file.
850 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
851 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
853 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
854 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
855 * config/nds32/nds32.md (pipeline_model): Add graywolf.
856 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
857 * config/nds32/pipelines.md: Include n15 settings.
859 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
860 Chung-Ju Wu <jasonwucj@gmail.com>
862 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
863 * config/nds32/nds32-n13.md: New file.
864 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
865 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
867 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
868 * config/nds32/nds32.md (pipeline_model): Add n13.
869 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
870 * config/nds32/pipelines.md: Include n13 settings.
872 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
873 Chung-Ju Wu <jasonwucj@gmail.com>
875 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
876 * config/nds32/nds32-n10.md: New file.
877 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
878 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
880 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
881 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
882 * config/nds32/nds32.md (pipeline_model): Add n10.
883 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
884 * config/nds32/pipelines.md: Include n10 settings.
886 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
887 Kito Cheng <kito.cheng@gmail.com>
888 Chung-Ju Wu <jasonwucj@gmail.com>
890 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
891 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
892 Add enum values for DSP extension instructions.
893 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
895 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
896 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
898 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
899 * config/nds32/nds32-dspext.md: New file for DSP implementation.
900 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
901 * config/nds32/nds32-intrinsic.md: Likewise.
902 * config/nds32/nds32_intrinsic.h: Likewise.
903 * config/nds32/nds32-md-auxiliary.c: Likewise.
904 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
905 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
906 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
907 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
908 * config/nds32/nds32-protos.h: New declarations for DSP extension.
909 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
910 TYPE_DMAC in switch statement.
911 * config/nds32/nds32.c: New checking and implementation for DSP
912 extension instructions.
913 * config/nds32/nds32.h: Likewise.
914 * config/nds32/nds32.md: Likewise.
915 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
916 * config/nds32/predicates.md: Implement new predicates for DSP
919 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
921 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
922 Reformat alternatives and attributes so it is easier to identify
923 which constraints/attributes go with which instruction.
924 (mov<mode>_hardfloat32, FMOVE64): Likewise.
925 (mov<mode>_softfloat32, FMOVE64): Likewise.
926 (mov<mode>_hardfloat64, FMOVE64): Likewise.
927 (mov<mode>_softfloat64, FMOVE64): Likewise.
929 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
931 * doc/extend.texi (PowerPC Built-in Functions): Rename this
933 (Basic PowerPC Built-in Functions): The new name of the
934 subsection previously known as "PowerPC Built-in Functions".
935 (Basic PowerPC Built-in Functions Available on all Configurations):
937 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
938 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
939 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
940 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
942 2018-05-11 Martin Jambor <mjambor@suse.cz>
945 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
948 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
951 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
953 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
955 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
956 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
957 (ix86_handle_option): Handle -mwaitpkg.
958 * config.gcc: New header.
959 * config/i386/cpuid.h (bit_WAITPKG): New bit.
960 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
961 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
963 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
964 OPTION_MASK_ISA_WAITPKG.
965 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
966 (ix86_option_override_internal): Add PTA_WAITPKG.
967 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
968 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
970 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
971 __builtin_ia32_umwait and __builtin_ia32_tpause.
972 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
973 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
974 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
975 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
976 UNSPECV_TPAUSE): New.
977 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
978 * config/i386/i386.opt: Add -mwaitpkg.
979 * config/i386/waitpkgintrin.h: New file.
980 * config/i386/x86intrin.h: New header.
981 * doc/invoke.texi: Add -mwaitpkg.
983 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
986 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
988 (cortex-m0): Use armv6s-m isa.
989 (cortex-m0plus): Likewise.
990 (cortex-m1): Likewise.
991 (cortex-m0.small-multiply): Likewise.
992 (cortex-m0plus.small-multiply): Likewise.
993 (cortex-m1.small-multiply): Likewise.
995 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
996 Jakub Jelinek <jakub@redhat.com>
998 PR tree-optimization/85692
999 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1000 source permute as well.
1002 2018-05-11 Martin Liska <mliska@suse.cz>
1005 * doc/extend.texi: Document LLVM style format for no_sanitize
1008 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1010 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1011 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1012 (mode_supports_vsx_dform_quad): Likewise.
1013 (mode_supports_vmx_dform): Move these functions to be next to the
1014 other mode_supports functions.
1015 (mode_supports_dq_form): Likewise.
1016 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1017 mode_supports_dq_form.
1018 (reg_offset_addressing_ok_p): Likewise.
1019 (offsettable_ok_by_alignment): Likewise.
1020 (rs6000_legitimate_offset_address_p): Likewise.
1021 (legitimate_lo_sum_address_p): Likewise.
1022 (rs6000_legitimize_address): Likewise.
1023 (rs6000_legitimize_reload_address): Likewise.
1024 (rs6000_secondary_reload_inner): Likewise.
1025 (rs6000_preferred_reload_class): Likewise.
1026 (rs6000_output_move_128bit): Likewise.
1028 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1030 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1031 Generate SImode target register for null target.
1032 <case IX86_BUILTIN_XGETBV>: Ditto.
1033 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1034 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1036 2018-05-10 Carl Love <cel@us.ibm.com>
1038 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1039 dcbtt and dcbtstt if operands[2] is 0.
1041 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1044 * config/i386/sse.md (usadv64qi): New expander.
1046 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1048 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1049 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1050 -maltivec=be support.
1051 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1052 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1053 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1054 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1055 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1056 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1057 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1058 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1059 altivec_vsumsws): Adjust.
1060 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1061 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1062 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1064 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1065 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1066 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1067 (altivec_lve<VI_char>x): Delete expand.
1068 (*altivec_lve<VI_char>x_internal): Rename to...
1069 (altivec_lve<VI_char>x): ... this.
1070 (altivec_lvxl_<mode>): Delete expand.
1071 (*altivec_lvxl_<mode>_internal): Rename to ...
1072 (altivec_lvxl_<mode>): ... this.
1073 (altivec_stvxl_<mode>): Delete expand.
1074 (*altivec_stvxl_<mode>_internal): Rename to ...
1075 (altivec_stvxl_<mode>): ... this.
1076 (altivec_stve<VI_char>x): Delete expand.
1077 (*altivec_stve<VI_char>x_internal): Rename to ...
1078 (altivec_stve<VI_char>x): ... this.
1079 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1080 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1081 reduc_plus_scal_<mode>): Adjust.
1082 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1084 (rs6000_cpu_cpp_builtins): Adjust.
1085 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1086 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1087 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1088 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1089 -maltivec=be support.
1090 (rs6000_split_vec_extract_var): Adjust.
1091 (rs6000_split_v4si_init): Adjust.
1092 (swap_selector_for_mode): Delete.
1093 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1094 altivec_expand_stvex_be): Delete.
1095 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1096 -maltivec=be support.
1097 (rs6000_gimple_fold_builtin): Ditto.
1098 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1100 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1101 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1102 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1103 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1104 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1105 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1106 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1107 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1108 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1109 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1110 anonymous split): Adjust.
1111 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1112 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1114 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1116 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1117 when --with-gxx-include-dir is also specified.
1118 * configure: Regenerate.
1120 2018-05-09 Jim Wilson <jimw@sifive.com>
1123 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1124 * config/riscv/t-withmultilib: New.
1125 * config/riscv/withmultilib.h: New.
1126 * doc/install.texi: Document RISC-V --with-multilib-list support.
1128 2018-05-09 Richard Biener <rguenther@suse.de>
1130 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1132 (vect_bb_vectorization_profitable_p): Adjust. Compute
1133 actual scalar cost using the cost vector and the add_stmt_cost
1136 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1138 PR rtl-optimization/85645
1139 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1140 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1142 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1144 PR rtl-optimization/85645
1145 * shrink-wrap.c (spread_components): Return a boolean saying if
1146 anything was changed.
1147 (try_shrink_wrapping_separate): Iterate spread_components until
1148 nothing changes anymore.
1150 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1152 PR rtl-optimization/85645
1153 * regrename.c (build_def_use): Also kill the chains that include the
1154 destination of a REG_CFA_REGISTER note.
1156 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1158 PR rtl-optimization/85645
1159 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1160 insn that has a REG_CFA_REGISTER note.
1162 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1164 * cfgexpand.c (expand_clobber): New function.
1165 (expand_gimple_stmt_1): Use it.
1166 * tree-vect-stmts.c (vect_clobber_variable): New function,
1168 (vectorizable_simd_clone_call): ...here.
1169 (vectorizable_store): Emit a clobber either side of an
1170 IFN_STORE_LANES sequence.
1171 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1173 2018-05-09 Tom de Vries <tom@codesourcery.com>
1176 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1177 (define_insn "trap_if_false"): Add exit after trap.
1179 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1181 PR rtl-optimization/85638
1182 * bb-reorder.c: Include common/common-target.h.
1183 (create_forwarder_block): New function extracted from...
1184 (fix_up_crossing_landing_pad): ...here. Rename into...
1185 (dw2_fix_up_crossing_landing_pad): ...this.
1186 (sjlj_fix_up_crossing_landing_pad): New function.
1187 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1188 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1189 from both partitions and exit the loop after one iteration.
1191 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1194 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1196 (Basic PowerPC Built-in Functions): The new name of the
1197 subsection previously known as "PowerPC Built-in Functions".
1198 (Basic PowerPC Built-in Functions Available on all Configurations):
1200 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1202 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1204 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1206 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1209 2018-05-08 Jim Wilson <jimw@sifive.com>
1211 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1212 (LD_EMUL_SUFFIX): New.
1213 (LINK_SPEC): Use it.
1215 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1217 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1219 (Basic PowerPC Built-in Functions): The new name of the
1220 subsection previously known as "PowerPC Built-in Functions".
1221 (Basic PowerPC Built-in Functions Available on all Configurations):
1223 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1225 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1227 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1229 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1232 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1235 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1236 after cmpelim optimization.
1238 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1240 * config.gcc: Support "goldmont".
1241 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1242 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1244 * config/i386/i386.c (m_GOLDMONT): Define.
1245 (processor_target_table): Add "goldmont".
1246 (PTA_GOLDMONT): Define.
1247 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1248 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1249 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1250 (fold_builtin_cpu): Add "goldmont".
1251 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1252 (ix86_option_override_internal): Add "goldmont".
1253 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1254 (processor_type): Add PROCESSOR_GOLDMONT.
1255 * config/i386/i386.md: Add CPU "glm".
1256 * config/i386/glm.md: New file.
1257 * config/i386/x86-tune.def: Add m_GOLDMONT.
1258 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1260 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1263 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1265 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1266 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1267 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1270 * config/i386/i386.c (ix86_fold_builtin): Handle
1271 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1274 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1275 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1277 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1280 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1281 (check_arch): Likewise.
1282 (check_fpu): Return the result rather than printing it.
1283 (end arch): Fix operator precedence.
1284 (end cpu): Likewise.
1285 (END): Print the result from check_fpu.
1287 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1288 Alan Hayward <alan.hayward@arm.com>
1289 David Sherwood <david.sherwood@arm.com>
1291 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1292 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1293 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1294 (*fcmuo<mode>_and): New patterns.
1296 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1298 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1299 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1300 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1301 (cmp_op, sve_imm_con): New code attributes.
1302 (SVE_COND_INT_CMP, imm_con): Delete.
1303 (cmp_op): Remove above unspecs from int attribute.
1304 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1306 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1307 comparison-specific unspecs.
1308 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1309 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1310 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1311 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1312 (*vec_fcm<cmp_op><mode>): Rename to...
1313 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1314 (*vec_fcmuo<mode>): Rename to...
1315 (*fcmuo<mode>): ...this and adjust likewise.
1316 (*pred_fcm<cmp_op><mode>): New pattern.
1317 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1318 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1320 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1322 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1323 (aarch64_emit_sve_predicated_cond): New function.
1324 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1325 (aarch64_emit_unspec_cond_or): Replace with...
1326 (aarch64_emit_sve_or_conds): ...this new function. Use
1327 aarch64_emit_sve_ptrue_op for the individual comparisons and
1328 aarch64_emit_binop to OR them together.
1329 (aarch64_emit_inverted_unspec_cond): Replace with...
1330 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1331 aarch64_emit_sve_ptrue_op for the comparison and
1332 aarch64_emit_unop to invert the result.
1333 (aarch64_expand_sve_vec_cmp_float): Update after the above
1334 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1336 2018-05-07 Nathan Sidwell <nathan@acm.org>
1338 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1339 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1340 (Backwards Compatibility): Likewise.
1342 2018-05-07 Luis Machado <luis.machado@linaro.org>
1346 2018-05-07 Luis Machado <luis.machado@linaro.org>
1348 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1349 <prefetch_dynamic_strides>: New const bool field.
1350 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1351 prefetch_dynamic_strides.
1352 (exynosm1_prefetch_tune): Likewise.
1353 (thunderxt88_prefetch_tune): Likewise.
1354 (thunderx_prefetch_tune): Likewise.
1355 (thunderx2t99_prefetch_tune): Likewise.
1356 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1357 (aarch64_override_options_internal): Update to set
1358 PARAM_PREFETCH_DYNAMIC_STRIDES.
1359 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1360 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1361 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1362 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1363 prefetch-dynamic-strides setting.
1365 2018-05-07 Luis Machado <luis.machado@linaro.org>
1367 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1368 <minimum_stride>: New const int field.
1369 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1370 minimum_stride field.
1371 (exynosm1_prefetch_tune): Likewise.
1372 (thunderxt88_prefetch_tune): Likewise.
1373 (thunderx_prefetch_tune): Likewise.
1374 (thunderx2t99_prefetch_tune): Likewise.
1375 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1376 (aarch64_override_options_internal): Update to set
1377 PARAM_PREFETCH_MINIMUM_STRIDE.
1378 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1379 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1380 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1381 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1382 stride is constant and is below the minimum stride threshold.
1384 2018-05-07 Luis Machado <luis.machado@linaro.org>
1386 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1389 2018-05-07 Luis Machado <luis.machado@linaro.org>
1391 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1392 <prefetch_dynamic_strides>: New const bool field.
1393 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1394 prefetch_dynamic_strides.
1395 (exynosm1_prefetch_tune): Likewise.
1396 (thunderxt88_prefetch_tune): Likewise.
1397 (thunderx_prefetch_tune): Likewise.
1398 (thunderx2t99_prefetch_tune): Likewise.
1399 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1400 (aarch64_override_options_internal): Update to set
1401 PARAM_PREFETCH_DYNAMIC_STRIDES.
1402 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1403 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1404 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1405 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1406 prefetch-dynamic-strides setting.
1408 2018-05-07 Luis Machado <luis.machado@linaro.org>
1410 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1411 <minimum_stride>: New const int field.
1412 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1413 minimum_stride field.
1414 (exynosm1_prefetch_tune): Likewise.
1415 (thunderxt88_prefetch_tune): Likewise.
1416 (thunderx_prefetch_tune): Likewise.
1417 (thunderx2t99_prefetch_tune): Likewise.
1418 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1419 (aarch64_override_options_internal): Update to set
1420 PARAM_PREFETCH_MINIMUM_STRIDE.
1421 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1422 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1423 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1424 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1425 stride is constant and is below the minimum stride threshold.
1427 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1430 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1431 the type is addressable. Don't force op into register if it has
1434 2018-05-05 Roland McGrath <mcgrathr@google.com>
1437 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1438 any section for which we don't know a specific type it should have,
1439 regardless of name. Previously this was done only for the exact
1440 names ".init_array", ".fini_array", and ".preinit_array".
1441 (default_elf_asm_named_section): Add comment about
1442 relationship with default_section_type_flags and SECTION_NOTYPE.
1443 (get_section): Don't consider it a type conflict if one side has
1444 SECTION_NOTYPE and the other doesn't, as long as neither has the
1445 SECTION_BSS et al used in the default_section_type_flags logic.
1447 2018-05-05 Tom de Vries <tom@codesourcery.com>
1450 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1451 (workaround_barsyncs): New function.
1452 (nvptx_reorg): Use workaround_barsyncs.
1453 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1454 (define_expand "nvptx_membar_cta"): New define_expand.
1455 (define_insn "*nvptx_membar_cta"): New insn.
1457 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1459 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1460 To improve optimization opportunities.
1461 * builtin-types.def: The new needed builtin types for the above.
1463 2018-05-04 Richard Biener <rguenther@suse.de>
1465 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1466 * gimple-ssa-store-merging.c
1467 (imm_store_chain_info::output_merged_store): Remove redundant create,
1468 release split_store vector contents on failure.
1469 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1470 scalar stmt vector on cache hit.
1472 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1474 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1476 * config.gcc (powerpc-xilinx-eabi*): Remove.
1477 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1479 (fusion_addis_mem_combo_load): Ditto.
1480 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1482 (rs6000_cpu_cpp_builtins): Ditto.
1483 * config/rs6000/rs6000-linux.c
1484 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1485 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1486 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1488 (rs6000_setup_reg_addr_masks): Ditto.
1489 (rs6000_init_hard_regno_mode_ok): Ditto.
1490 (rs6000_option_override_internal): Ditto.
1491 (legitimate_lo_sum_address_p): Ditto.
1492 (rs6000_legitimize_address): Ditto.
1493 (rs6000_legitimize_reload_address): Ditto.
1494 (rs6000_legitimate_address_p): Ditto.
1495 (abi_v4_pass_in_fpr): Ditto.
1496 (setup_incoming_varargs): Ditto.
1497 (rs6000_gimplify_va_arg): Ditto.
1498 (rs6000_split_multireg_move): Ditto.
1499 (rs6000_savres_strategy): Ditto.
1500 (rs6000_emit_prologue_components): Ditto.
1501 (rs6000_emit_epilogue_components): Ditto.
1502 (rs6000_emit_prologue): Ditto.
1503 (rs6000_emit_epilogue): Ditto.
1504 (rs6000_elf_file_end): Ditto.
1505 (rs6000_function_value): Ditto.
1506 (rs6000_libcall_value): Ditto.
1507 * config/rs6000/rs6000.h: Ditto.
1508 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1509 (TARGET_MINMAX): ... this. New.
1510 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1511 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1512 (*movsi_internal1_single): Delete.
1513 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1514 mfpu=, mxilinx-fpu): Delete.
1515 * config/rs6000/singlefp.h: Delete.
1516 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1517 * config/rs6000/t-rs6000: Ditto.
1518 * config/rs6000/t-xilinx: Delete.
1519 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1520 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1521 (VStype_simple): Delete.
1522 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1523 * config/rs6000/xfpu.h: Delete.
1524 * config/rs6000/xfpu.md: Delete.
1525 * config/rs6000/xilinx.h: Delete.
1526 * config/rs6000/xilinx.opt: Delete.
1527 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1528 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1530 2018-05-04 Tom de Vries <tom@codesourcery.com>
1533 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1536 2018-05-04 Richard Biener <rguenther@suse.de>
1539 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1540 (expand_complex_div_wide): Likewise.
1541 (expand_complex_operations_1): Likewise.
1542 (expand_complex_libcall): Preserve EH info of the original stmt.
1543 (tree_lower_complex): Handle removed blocks.
1544 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1545 on complex multiplication and division libcall builtins.
1547 2018-05-04 Richard Biener <rguenther@suse.de>
1550 * fold-const.c (negate_expr_p): Restrict negation of operand
1551 zero of a division to when we know that can happen without
1553 (fold_negate_expr_1): Likewise.
1555 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1558 * real.h (real_nextafter): Declare.
1559 * real.c (real_nextafter): New function.
1560 * fold-const-call.c (fold_const_nextafter): New function.
1561 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1562 CASE_CFN_NEXTTOWARD.
1563 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1564 even when arg1_mode is different from arg0_mode.
1566 2018-05-03 Nathan Sidwell <nathan@acm.org>
1568 * doc/extend.texi (Deprecated Features): Remove
1570 (Backwards Compatibility): Likewise.
1571 * doc/invoke.texi (C++ Language Options): Likewise.
1572 (C++ Dialect Options): Likewise.
1574 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1577 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1578 _mm512_mask_mullox_epi64): New intrinsics.
1580 2018-05-03 Tom de Vries <tom@codesourcery.com>
1583 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1584 dump files): Add offload-tree.
1586 2018-05-03 Richard Biener <rguenther@suse.de>
1588 PR tree-optimization/85615
1589 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1590 to loops not nested in BBs loop father to avoid creating multi-entry
1593 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1595 PR tree-optimization/70291
1596 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1597 arguments. Change return type to tree. Emit libcall as a new
1598 statement rather than replacing existing one when inplace_p is true.
1599 (expand_complex_multiplication_components): New function.
1600 (expand_complex_multiplication): Expand floating-point complex
1601 multiplication using the above.
1602 (expand_complex_division): Rename inner_type parameter to type.
1603 Update expand_complex_libcall call-site.
1604 (expand_complex_operations_1): Update expand_complex_multiplication
1605 and expand_complex_division call-sites.
1607 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1610 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1611 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1612 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1613 the highest significant bit of the shift count mask is clear. In
1614 check whether and[sq]i3 is needed verify that all significant bits
1615 of the shift count other than the highest are set.
1617 2018-05-02 Tom de Vries <tom@codesourcery.com>
1620 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1621 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1622 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1623 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1624 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1625 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1626 __builtin_goacc_parlevel_size.
1628 2018-05-02 Richard Biener <rguenther@suse.de>
1630 PR tree-optimization/85597
1631 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1632 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1635 2018-05-02 Tom de Vries <tom@codesourcery.com>
1638 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1639 dump files): Add ltrans-tree.
1641 2018-05-02 Tom de Vries <tom@codesourcery.com>
1644 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1645 dump files): Add wpa-ipa.
1647 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1649 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1650 powerpc*-*-linux*paired* target.
1651 * config/rs6000/750cl.h: Delete.
1652 * config/rs6000/paired.h: Delete.
1653 * config/rs6000/paired.md: Delete.
1654 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1656 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1657 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1658 comment. Remove paired float support.
1659 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1660 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1662 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1663 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1665 * config/rs6000/rs6000.c: Remove paired float support.
1666 (paired_expand_vector_init, paired_expand_vector_move,
1667 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1668 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1669 paired_expand_builtin, paired_expand_predicate_builtin,
1670 paired_init_builtins): Delete.
1671 * config/rs6000/rs6000.h: Remove paired float support.
1672 * config/rs6000/rs6000.md: Remove paired float support.
1673 (move_from_CR_ov_bit): Delete.
1674 * config/rs6000/rs6000.opt (mpaired): Delete.
1675 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1676 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1678 2018-05-02 Richard Biener <rguenther@suse.de>
1681 * gimplify.c (gimplify_save_expr): When in SSA form allow
1682 SAVE_EXPRs to compute to SSA vars.
1684 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1687 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1688 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1689 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1690 clobber operands[2], instead use a new pseudo. Formatting fixes.
1692 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1694 PR tree-optimization/85586
1695 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1696 exit early for statements in the same group if the accesses are
1699 2018-05-02 Tom de Vries <tom@codesourcery.com>
1702 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1705 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1707 PR tree-optimization/85143
1708 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1710 2018-05-01 Tom de Vries <tom@codesourcery.com>
1713 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1714 not found" error message.
1716 2018-05-01 Tom de Vries <tom@codesourcery.com>
1719 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1720 * vec.c (test_ordered_remove_if): New function.
1721 (vec_c_tests): Call test_ordered_remove_if.
1722 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1723 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1724 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1725 VEC_ORDERED_REMOVE_IF.
1727 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1729 PR tree-optimization/82665
1730 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1731 pointer subtraction where arguments come from a memchr call.
1733 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1735 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1736 --push-state --as-needed and --pop-state instead of --as-needed and
1737 --no-as-needed if ld supports it.
1738 * configure: Regenerated.
1741 * doc/install.texi2html: Replace _002d with - and _002a with * in
1742 generated html files using sed.
1744 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1747 * gcc-rich-location.c (blank_line_before_p): New function.
1748 (use_new_line): New function.
1749 (gcc_rich_location::add_fixit_insert_formatted): New function.
1750 * gcc-rich-location.h
1751 (gcc_rich_location::add_fixit_insert_formatted): New function.
1753 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1755 * selftest.c (assert_streq): Rename "expected" and "actual" to
1756 "val1" and "val2". Extend NULL-handling to cover both inputs
1757 symmetrically, while still requiring both to be non-NULL for a pass.
1758 * selftest.h (assert_streq): Rename "expected" and "actual" to
1760 (ASSERT_EQ): Likewise.
1761 (ASSERT_EQ_AT): Likewise.
1762 (ASSERT_KNOWN_EQ): Likewise.
1763 (ASSERT_KNOWN_EQ_AT): Likewise.
1764 (ASSERT_NE): Likewise.
1765 (ASSERT_MAYBE_NE): Likewise.
1766 (ASSERT_MAYBE_NE_AT): Likewise.
1767 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1768 the assertion to pass.
1769 (ASSERT_STREQ_AT): Likewise.
1771 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1773 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1774 interaction with -pie.
1776 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1778 * selftest.h: Fix alphabetization of per-source-file selftest
1781 2018-04-30 Jason Merrill <jason@redhat.com>
1783 PR c++/61982 - dead stores to destroyed objects.
1784 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1787 2018-04-30 Jason Merrill <jason@redhat.com>
1789 * tree.c (build_clobber): New.
1790 * tree.h: Declare it.
1791 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1793 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1795 * diagnostic-show-locus.c (layout::layout): Update for
1796 location_get_source_line returning a char_span.
1797 (struct char_span): Move to input.h.
1798 (struct correction): Update for fields in char_span becoming
1800 (struct source_line): Update for location_get_source_line
1801 returning a char_span.
1802 (layout::print_line): Likewise.
1803 * edit-context.c (edited_file::print_content): Likewise.
1804 (edited_file::print_diff_hunk): Likewise.
1805 (edited_file::print_run_of_changed_lines): Likewise.
1806 (edited_file::get_num_lines): Likewise.
1807 (edited_line::edited_line): Likewise.
1808 * final.c (asm_show_source): Likewise.
1809 * input.c (location_get_source_line): Convert return type
1810 from const char * to char_span, losing the final "line_len"
1812 (dump_location_info): Update for the above.
1813 (get_substring_ranges_for_loc): Likewise. Use a char_span
1814 when handling the literal within the line.
1815 (test_reading_source_line): Update for location_get_source_line
1816 returning a char_span.
1817 * input.h (class char_span): Move here from
1818 diagnostic-show-locus.c, converting from a struct to a class.
1819 Make data members private.
1820 (char_span::operator bool): New.
1821 (char_span::length): New.
1822 (char_span::get_buffer): New.
1823 (char_span::operator[]): New.
1824 (char_span::subspan): Make const.
1825 (char_span::xstrdup): New.
1826 (location_get_source_line): Convert return type from const char *
1827 to char_span, losing the final "line_size" param.
1829 2018-04-30 Jan Hubicka <jh@suse.cz>
1831 * lto-wrapper.c (ltrans_priorities): New static var.
1832 (cmp_priority): New.
1833 (run_gcc): Read priorities and if doing parallel build order
1834 the Makefile by them.
1836 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1838 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1840 2018-04-30 Richard Biener <rguenther@suse.de>
1842 * tree-cfg.c (verify_address): Remove base argument, add
1843 flag whether to check TREE_ADDRESSABLE and do that.
1844 (verify_expr): Remove.
1845 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1846 (verify_gimple_assign_single): Likewise.
1847 (verify_gimple_switch): Likewise.
1848 (verify_expr_location_1): Dereference tp once. Add (disabled)
1849 piece from verify_expr.
1850 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1852 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1854 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1856 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1858 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1859 (small_data_pattern): Likewise.
1860 (arc_rewrite_small_data): Likewise.
1861 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1862 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1863 (get_symbol_alignment): New function.
1864 (legitimate_small_data_address_p): Likewise.
1865 (legitimate_scaled_address): Update, call
1866 legitimate_small_data_address_p.
1867 (output_sdata): New static variable.
1868 (arc_print_operand): Update how we handle small data operands.
1869 (arc_print_operand_address): Likewise.
1870 (arc_legitimate_address_p): Update, use
1871 legitimate_small_data_address_p.
1872 (arc_rewrite_small_data_p): Remove.
1873 (arc_rewrite_small_data_1): Likewise.
1874 (arc_rewrite_small_data): Likewise.
1875 (small_data_pattern): Likewise.
1876 (compact_sda_memory_operand): Update to use
1877 legitimate_small_data_address_p and get_symbol_alignment.
1878 (prepare_move_operands): Don't rewite sdata pattern.
1879 (prepare_extend_operands): Remove.
1880 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1882 (zero_extendqisi2): Likewise.
1883 (zero_extendhisi2): Likewise.
1884 (extendqihi2): Likewise.
1885 (extendqisi2): Likewise.
1886 (extendhisi2): Likewise.
1890 * config/arc/constraints.md (Usd): Change it to memory constraint.
1892 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1894 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1895 as source of std instructions.
1896 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1897 allow 6-bit constants as source for store instructions.
1898 (movdi_insn): Update instruction pattern to allow 6-bit constants
1899 as source for store instructions.
1901 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1903 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1905 2018-04-30 Nathan Sidwell <nathan@acm.org>
1906 Sandra Loosemore <sandra@codesourcery.com>
1908 * dumpfile.c (dump_open): Allow '-' for stdout.
1909 * doc/invoke.texi (Developer Options): Document dump filename
1910 determination early. Document stdin/stdout selection.
1912 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1914 Microblaze Target: PIC data text relative
1916 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1917 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1919 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1920 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1921 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1922 New addressing mode for data-text relative position indepenedent code.
1923 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1924 'ADDRESS_SYMBOLIC_TXT_REL'.
1925 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1926 (microblaze_legitimate_pic_operand): Exclude function calls from
1927 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1928 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1930 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1931 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1932 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1933 for 'address + offset'.
1934 (microblaze_expand_prologue): Add new function prologue call for
1936 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1937 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1938 table in case of TARGET_PIC_DATA_TEXT_REL.
1939 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1940 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1941 Add new macros 'UNSPEC_TEXT',
1942 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1943 + exclude function calls from 'UNSPEC_PLT' in case of data text
1945 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1946 new target hook for generating address diff vector tables in case of
1948 * doc/tm.texi : Regenerate.
1949 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1950 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1951 of addr diff vector generation.
1952 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1953 target hook definition.
1954 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1955 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1956 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1957 option for data text relative.
1959 2018-04-30 Richard Biener <rguenther@suse.de>
1961 * tree-chrec.h (evolution_function_is_constant_p): Remove
1963 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1965 2018-04-30 Richard Biener <rguenther@suse.de>
1968 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1970 2018-04-30 Richard Biener <rguenther@suse.de>
1972 PR tree-optimization/28364
1973 PR tree-optimization/85275
1974 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1975 copying first exit test.
1977 2018-04-28 Mark Wielaard <mark@klomp.org>
1979 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1981 (dwarf_AT): Handle DW_AT_addr_base.
1982 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1984 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1987 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1988 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1989 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1990 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1992 2018-04-28 Richard Biener <rguenther@suse.de>
1994 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1995 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1996 to reflect use. Only add interesting stmts.
1998 2018-04-27 Martin Jambor <mjambor@suse.cz>
2001 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2002 the jump function allows for passing through aggregate values.
2004 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2006 * input.h (in_system_header_at): Convert from macro to inline
2008 (from_macro_expansion_at): Likewise.
2009 (from_macro_definition_at): Likewise.
2011 2018-04-27 Jeff Law <law@redhat.com>
2013 * config.gcc: Mark tile* targets as deprecated/obsolete.
2015 2018-04-27 Richard Biener <rguenther@suse.de>
2017 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2020 2018-04-27 Richard Biener <rguenther@suse.de>
2022 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2024 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2026 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2027 with Yd constraint. Set "preferred_for_speed" attribute from
2028 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2030 (*movdi_internal): Ditto.
2031 (movti_interunit splitters): Remove
2032 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2033 (movdi_interunit splitters): Ditto.
2034 * config/i386/constraints.md (Ye): Remove.
2035 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2037 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2040 * config/aarch64/constraints.md (Usg): Limit to 31.
2043 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2045 PR tree-optimization/85529
2046 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2047 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2048 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2049 zero extension or masking of the MSB bit.
2050 (optimize_range_tests): Add FIRST_BB argument, pass it through
2051 to optimize_range_tests_var_bound.
2052 (maybe_optimize_range_tests, reassociate_bb): Adjust
2053 optimize_range_tests callers.
2055 2018-04-26 Richard Biener <rguenther@suse.de>
2056 Jakub Jelinek <jakub@redhat.com>
2058 * cgraph.h (symbol_table): Just declare debug method here.
2059 * symtab.c (symbol_table::debug): Define.
2061 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2063 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2065 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2067 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2068 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2069 (*movdi_internal): Substitute Yi and Yj constraint with x
2070 and Ym and Yn constraint with y constraint. Update "isa"
2071 attribute and set "preferred_for_speed" attribute from
2072 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2073 (*movsi_internal): Ditto.
2074 (*movdf_internal): Ditto.
2075 (*movsf_internal): Ditto.
2076 (*zero_extendsidi2): Ditto.
2077 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2078 (sse2_loadld): Ditto.
2079 (*vec_extract<ssevecmodelower>_0): Ditto.
2080 (*vec_extractv4si_0_zext_sse4): Ditto.
2081 (vec_concatv2di): Ditto.
2082 (*vec_dup<mode>): Ditto.
2083 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2084 * config/i386/constraints.md (Yi): Remove.
2089 2018-04-26 Nathan Sidwell <nathan@acm.org>
2091 * dumpfile.c (dump_open): New.
2092 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2093 (dump_finish): Detect stdio/stderr by value not name.
2095 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2097 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2099 2018-04-26 Tom de Vries <tom@codesourcery.com>
2102 * config/nvptx/nvptx.c (verify_neutering_jumps)
2103 (verify_neutering_labels): New function
2104 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2106 2018-04-26 Tom de Vries <tom@codesourcery.com>
2109 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2110 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2113 2018-04-26 Richard Biener <rguenther@suse.de>
2114 Tom de Vries <tom@codesourcery.com>
2117 * lto-streamer-out.c (output_function): Fixup loops if required to match
2118 discovery done in the reader.
2120 2018-04-26 Richard Biener <rguenther@suse.de>
2122 PR tree-optimization/85116
2123 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2124 have a loop exit from the single latch predecessor. Remove
2125 case of header with just condition.
2126 (ch_base::copy_headers): Exclude infinite loops from any
2128 (pass_ch::execute): Record exits.
2130 2018-04-26 Richard Biener <rguenther@suse.de>
2132 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2133 prologue cost vector and pass it to vect_get_load_cost.
2134 (vect_get_peeling_costs_all_drs): Likewise.
2135 (vect_peeling_hash_get_lowest_cost): Likewise.
2136 (vect_enhance_data_refs_alignment): Likewise.
2138 2018-04-26 Richard Biener <rguenther@suse.de>
2141 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2142 checking of integer<->pointer conversions.
2143 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2144 sign-/zero-extending pointer types.
2145 (expand_omp_for_static_chunk): Likewise.
2147 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2148 Jean Lee <xiaoyur347@gmail.com>
2150 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2151 (TARGET_ASAN_SHADOW_OFFSET): Define.
2152 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2153 true for -fsanitize=address.
2155 2018-04-25 Mark Wielaard <mark@klomp.org>
2157 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2160 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2162 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2163 than "alu", remove explicit "memory" and "imm_disp" attributes.
2164 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2167 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2168 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2169 gen_lowpart_no_emit.
2171 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2174 * config/i386/i386.c (ix86_expand_builtin): Change memory
2175 operand to XI, extend p0 to Pmode.
2176 * config/i386/i386.md: Change unspec volatile and operand
2177 1 mode to XI, change operand 0 mode to P.
2179 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2181 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2182 GET_MODE_MASK before any checking.
2183 (nds32_can_use_bset_p): Likewise.
2184 (nds32_can_use_btgl_p): Likewise.
2186 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2188 * config/nds32/nds32-doubleword.md: New define_split pattern for
2189 illegal register number.
2191 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2193 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2195 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2197 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2199 2018-04-25 Richard Biener <rguenther@suse.de>
2201 * lto-streamer.h (LTO_major_version): Bump to 8.
2203 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2205 * BASE-VER: Set to 9.0.0.
2207 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2209 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2210 in __abskf2 and __powikf2.
2212 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2215 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2216 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2217 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2218 Use the above on operand 2. Reindent.
2219 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2221 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2224 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2226 * config/i386/i386.opt (mcet): Removed.
2227 * doc/install.texi: Remove -mcet documentation.
2228 * doc/invoke.texi: Likewise.
2230 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2233 * doc/install.texi: Remove -mcet from bootstrap-cet.
2235 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2238 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2239 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2243 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2244 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2245 containing a CONST_VECTOR.
2247 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2249 * doc/install.texi: Update newlib dependency for nvptx.
2251 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2254 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2255 instead of INTVAL when shifting x left.
2257 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2259 PR tree-optimization/85478
2260 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2261 vect_grouped_store_supported for single element vectors.
2263 2018-04-24 Richard Biener <rguenther@suse.de>
2266 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2267 load cost increase to the case of non-constant step.
2269 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2272 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2273 destination if any_malformed_asm.
2275 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2278 * expr.c (store_field): In the bitfield case, if the value comes from
2279 a function call and is returned in registers by means of a PARALLEL,
2280 do not change the mode of the temporary unless BLKmode and VOIDmode.
2282 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2284 PR rtl-optimization/85423
2285 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2286 dependencies to debug insns when the previous insn is non-debug.
2288 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2290 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2291 enums into a single definition.
2292 (fls): Fix predicates and printing.
2295 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2297 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2298 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2299 and short u6 immediate.
2300 (check_if_valid_sleep_operand): Remove.
2301 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2303 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2305 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2306 flag_always_save_lp condition.
2307 * config/nds32/nds32.opt (malways-save-lp): New option.
2309 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2311 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2312 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2313 * config/nds32/nds32.h
2314 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2315 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2317 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2319 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2320 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2322 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2323 Chung-Ju Wu <jasonwucj@gmail.com>
2325 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2327 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2328 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2330 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2332 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2334 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2336 * config/nds32/nds32-protos.h (nds32_data_alignment,
2337 nds32_local_alignment): Declare.
2338 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2339 nds32_local_alignment): New functions.
2340 (TARGET_CONSTANT_ALIGNMENT): Define.
2341 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2343 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2345 * config/nds32/nds32.c
2346 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2347 (TARGET_MODES_TIEABLE_P): Likewise.
2349 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2351 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2354 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2355 Chung-Ju Wu <jasonwucj@gmail.com>
2357 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2358 for unaligned access.
2359 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2361 * config/nds32/nds32-intrinsic.md: Likewise.
2362 * config/nds32/nds32_intrinsic.h: Likewise.
2363 * config/nds32/nds32.h (nds32_builtins): Likewise.
2364 * config/nds32/nds32.opt (munaligned-access): New option.
2365 * config/nds32/nds32.c (nds32_asm_file_start): Display
2366 flag_unaligned_access status.
2368 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2370 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2371 -mno-relax is present.
2372 * config/riscv/linux.h (LINK_SPEC): Ditto.
2374 2018-04-20 Martin Sebor <msebor@redhat.com>
2377 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2379 (gimple_fold_builtin_stxcpy_chk): Same.
2380 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2382 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2385 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2386 __powikf2 when long double is IEEE 128-bit.
2388 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2390 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2391 step to make sure stack always aligned.
2393 2018-04-20 Carl Love <cel@us.ibm.com>
2396 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2397 size check for arg0.
2399 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2400 Tom de Vries <tom@codesourcery.com>
2403 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2404 Emit insns for calls too.
2405 (nvptx_find_par): Always look for worker-level predecessor insn.
2406 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2408 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2409 (nvptx_process_pars): Propagate frames for calls.
2411 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2414 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2416 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2417 (ix86_handle_option): Don't handle OPT_mibt.
2418 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2420 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2422 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2423 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2424 (ix86_target_macros): Define __CET__ with flag_cf_protection
2425 for -fcf-protection.
2426 * config/i386/i386.c (isa2_opts): Remove -mibt.
2427 * config/i386/i386.h (TARGET_IBT): Removed.
2428 (TARGET_IBT_P): Likewise.
2429 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2430 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2431 * config/i386/i386.opt (mcet): Update help message.
2434 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2435 -mcet as an alias for -mshstk.
2437 2018-04-20 Richard Biener <rguenther@suse.de>
2440 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2441 complexity by forcing a single use of the multiply operand.
2443 2018-04-20 Martin Jambor <mjambor@suse.cz>
2446 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2447 recursion dependency to only apply to non-clones.
2449 2018-04-20 Martin Jambor <mjambor@suse.cz>
2452 * ipa-cp.c (create_specialized_node): Check that clones of
2453 self-recursive edges exist during IPA-CP.
2455 2018-04-19 Toon Moene <toon@moene.org>
2457 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2460 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2462 PR tree-optimization/85467
2463 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2464 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2465 VECTOR_CST element to type.
2467 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2470 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2471 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2472 (builtin_longjmp): Likewise.
2473 (save_stack_nonlocal): New pattern.
2474 (restore_stack_nonlocal): Likewise.
2476 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2479 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2480 Replace ASM_OUTPUT_LABEL with fprintf.
2482 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2485 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2486 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2487 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2488 define __IBT__ and __SHSTK__ for -fcf-protection.
2489 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2491 (ix86_trampoline_init): Likewise.
2492 (x86_output_mi_thunk): Likewise.
2493 (ix86_notrack_prefixed_insn_p): Likewise.
2494 (ix86_option_override_internal): Don't disallow -fcf-protection.
2495 * config/i386/i386.md (rdssp<mode>): Also enable for
2497 (incssp<mode>): Likewise.
2498 (nop_endbr): Likewise.
2499 * config/i386/i386.opt (mcet): Change help message to built-in
2503 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2504 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2505 enable CET built-in functions.
2507 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2509 * common/config/i386/i386-common.c
2510 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2511 OPTION_MASK_ISA_MOVDIRI_UNSET,
2512 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2513 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2514 * config.gcc (movdirintrin.h): New header.
2515 * config/i386/cpuid.h (bit_MOVDIRI,
2516 bit_MOVDIR64B): New bits.
2517 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2519 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2520 (VOID, PVOID, PCVOID)): New function types.
2521 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2522 __builtin_ia32_directstoreu_u64,
2523 __builtin_ia32_movdir64b): New builtins.
2524 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2525 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2527 (ix86_valid_target_attribute_inner_p): Ditto.
2528 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2529 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2530 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2531 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2532 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2533 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2534 (movdiri<mode>, movdir64b_<mode>): New.
2535 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2536 * config/i386/immintrin.h: Include movdirintrin.h.
2537 * config/i386/movdirintrin.h: New file.
2538 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2540 2018-04-19 Richard Biener <rguenther@suse.de>
2543 * cfg.c (clear_bb_flags): When loop state says we have
2544 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2546 2018-04-19 Richard Biener <rguenther@suse.de>
2548 PR tree-optimization/84737
2549 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2550 copying restrict info.
2551 (vect_setup_realignment): Use it.
2552 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2553 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2554 the first DR to all generated stores.
2555 (vectorizable_load): Likewise for loads.
2557 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2559 PR tree-optimization/85446
2560 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2561 the integral and pointer types to have the same precision.
2563 * doc/install.texi: Document --disable-cet being the default and
2566 2018-04-18 Martin Liska <mliska@suse.cz>
2568 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2571 2018-04-18 Martin Liska <mliska@suse.cz>
2574 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2577 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2578 arguments if they are comparable.
2580 2018-04-18 Martin Liska <mliska@suse.cz>
2583 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2586 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2589 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2592 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2593 ENDBR after calling __morestack.
2595 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2598 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2599 by using gcc_base_ver to generate a gcc_driver_version, and use
2600 it when generating GCC_DRIVER_NAME.
2601 * configure: Regenerate.
2603 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2606 * config.gcc: Obsolete powerpc*-*-*spe*.
2608 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2611 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2612 (stabstr_D): Change type of unum from unsigned int to
2613 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2616 2018-04-17 Jim Wilson <jimw@sifive.com>
2619 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2620 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2621 Set arg_pointer_offset after using pretend_args_size.
2623 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2625 PR rtl-optimization/85431
2626 * dse.c (record_store): Ignore zero width stores.
2629 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2630 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2631 __builtin_stack_restore rather than after it.
2632 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2633 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2634 argument instead of virtual_dynamic_stack_rtx.
2636 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2638 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2640 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2641 Add note to error message to explain internal mapping of overloaded
2642 built-in function name to non-overloaded built-in function name.
2643 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2646 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2649 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2650 where the inputs overlap with the output.
2652 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2654 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2655 (=v, v) alternative and explicit "memory" attribute.
2656 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2657 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2659 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2660 "sselog1" type instead of "sselog".
2661 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2662 "sselog". Remove explicit "memory" attribute.
2663 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2664 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2666 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2667 "sselog1" type instead of "sselog". Remove explicit "memory"
2669 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2670 use "sselog1" type instead of "sselog". Remove explicit "memory"
2672 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2673 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2675 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2676 "sselog1" type instead of "sselog". Remove explicit "memory"
2678 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2679 use "sselog1" type instead of "sselog". Remove explicit "memory"
2683 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2686 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2689 2018-04-17 Martin Jambor <mjambor@suse.cz>
2692 * ipa-cp.c (create_specialized_node): Call
2693 expand_all_artificial_thunks if necessary.
2695 2018-04-17 Martin Liska <mliska@suse.cz>
2698 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2699 in message, remote space in between '_G' and '('.
2701 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2704 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2705 avx512f_vmcmp<mode>3<round_saeonly_name>,
2706 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2707 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2708 avx512f_rndscale<mode><round_saeonly_name>,
2709 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2710 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2711 Use %<iptr>2 instead of %2 for -masm=intel.
2712 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2713 avx512f_vcvttss2usi<round_saeonly_name>,
2714 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2716 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2717 avx512f_vcvttsd2usi<round_saeonly_name>,
2718 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2719 Use %q1 instead of %1 for -masm=intel.
2720 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2721 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2722 of %3 for -masm=intel.
2723 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2725 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2727 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2729 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2730 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2732 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2734 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2735 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2736 %g1 and one with %0 and %1.
2737 (avx512er_vmrcp28<mode><round_saeonly_name>,
2738 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2740 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2741 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2742 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2743 of %0 and %{%4%} for -masm=intel.
2744 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2745 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2746 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2747 order of %0 and %{%5%}%{z%} for -masm=intel.
2749 2018-04-17 Jan Hubicka <jh@suse.cz>
2752 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2754 2018-04-17 Martin Liska <mliska@suse.cz>
2757 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2758 for target_clone error message. Make default implementation
2759 clone to be a local declaration.
2760 (separate_attrs): Add new argument and check for an empty
2762 (expand_target_clones): Handle it.
2763 (ipa_target_clone): Make redirection just for target_clones
2766 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2767 Tom de Vries <tom@codesourcery.com>
2770 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2771 tiled basic blocks without omp continue statements.
2773 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2776 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2777 vec_extract expression as having side effects to make sure it gets
2780 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2783 * config/i386/i386.c (get_builtin_code_for_version): Check
2786 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2789 * gcc/config.gcc: Support "skylake".
2790 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2792 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2793 (processor_target_table): Add "skylake".
2794 (ix86_option_override_internal): Add "skylake".
2795 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2796 PROCESSOR_CANNONLAKE.
2797 (get_builtin_code_for_version): Fix priority for
2798 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2799 PROCESSOR_SKYLAKE-AVX512.
2800 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2801 (processor_type): Add PROCESSOR_SKYLAKE.
2803 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2804 Jason Merrill <jason@redhat.com>
2807 * convert.c (convert_to_integer_1): Use direct recursion for
2808 enumeral types and types with a precision less than the number
2809 of bits in their mode.
2811 2018-04-16 Julia Koval <julia.koval@intel.com>
2814 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2815 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2817 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2820 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2821 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2822 and -mno-direct-move.
2824 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2827 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2828 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2829 Ensure negative shifts result in {0}.
2831 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2833 PR rtl-optimization/79916
2834 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2835 regs (if any) to define how to gnerate SD moves when LRA is in
2838 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2840 PR rtl-optimization/85393
2841 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2842 * except.c (expand_dw2_landing_pad_for_region): Make static.
2843 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2844 a label and unconditional jump to old_bb, rather than
2845 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2848 PR rtl-optimization/85376
2849 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2850 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2851 instead of a specific value.
2853 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2854 Bin Cheng <bin.cheng@arm.com>
2856 PR tree-optimization/82965
2857 PR tree-optimization/83991
2858 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2859 by_profile_only parameter.
2860 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2861 information if the loop was predicted to iterate too many times.
2862 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2864 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2867 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2870 2018-04-13 Martin Liska <mliska@suse.cz>
2871 Jakub Jelinek <jakub@redhat.com>
2874 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2875 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2876 * builtins.c (expand_builtin_memory_copy_args): Use
2877 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2878 handle dest_addr == pc_rtx.
2880 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2883 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2884 asked to not generate direct moves.
2885 (fix_trunc<mode>si2_stfiwx): Similar.
2886 (fix_trunc<mode>si2_internal): Similar.
2888 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2891 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2892 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2893 lookup if dest in some wider mode is known to be const0_rtx and
2894 if so, record permanent equivalence for it to be ZERO_EXTEND of
2895 the narrower mode destination.
2897 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2899 * lto-streamer-out.c (output_function): Revert 259346.
2900 * omp-expand.c (expand_oacc_for): Likewise.
2902 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2904 PR rtl-optimization/85354
2905 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2906 * sel-sched.c (sel_global_init): ... here.
2908 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2911 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2912 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2913 mode for PE-COFF targets.
2914 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2915 (i386_pe_asm_lto_end): Likewise.
2916 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2917 (TARGET_ASM_LTO_END): Likewise.
2918 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2919 (i386_pe_asm_lto_start): New function.
2920 (i386_pe_asm_lto_end): Likewise.
2922 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2923 Richard Biener <rguenther@suse.de>
2926 * lto-streamer-out.c (output_function): Fix CFG loop state before
2928 * omp-expand.c (expand_oacc_for): Handle calls to internal
2929 functions like regular functions.
2931 2018-04-12 Richard Biener <rguenther@suse.de>
2934 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2935 for the early LTO debug to properly generate references to it
2936 during DIE emission. Do not re-use that for the skeleton for
2938 (dwarf2out_early_finish): Likewise.
2940 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2943 * config/i386/sse.md
2944 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2945 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2946 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2947 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2948 and output is a reg, avoid creating invalid lowpart subreg, but
2949 instead split into a 512-bit move. Don't split if not AVX512VL,
2950 input is xmm16+ reg and output is a mem.
2951 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2952 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2953 xmm16+ reg and output is a mem.
2955 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2957 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2958 also for flag_dwarf2_cfi_asm.
2960 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2962 PR rtl-optimization/85342
2963 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2964 a bool scalar var inside of the loop instead. Don't try to update
2965 recog_data.operand after failed apply_change_group.
2967 2018-04-12 Tom de Vries <tom@codesourcery.com>
2970 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2971 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2972 array with flexible array member as array without given dimension.
2973 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2974 argument for undefined param to true.
2976 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2979 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2980 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2981 from PowerPC section.
2982 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2983 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2984 help text that is too long.
2985 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2986 help text that is too long.
2987 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2988 help text that is too long.
2990 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2992 * config/alpha/alpha.md (stack_probe_internal): Rename
2993 from "probe_stack". Update all callers.
2995 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2997 PR rtl-optimization/84566
2998 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2999 sched_macro_fuse_insns.
3001 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3004 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3005 (compute_block_dependences): ... from here.
3007 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3009 PR tree-optimization/85331
3010 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3011 from int to HOST_WIDE_INT.
3013 2018-04-11 Martin Jambor <mjambor@suse.cz>
3016 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3017 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3018 not the same as the source val.
3019 (cgraph_edge_brings_value_p): New parameter.
3020 (gather_edges_for_value): Pass destination value to
3021 cgraph_edge_brings_value_p.
3022 (perhaps_add_new_callers): Likewise.
3023 (get_info_about_necessary_edges): Likewise and exclude values brought
3024 only by self-recursive edges.
3025 (create_specialized_node): Redirect only clones of self-calling edges.
3026 (+self_recursive_pass_through_p): New function.
3027 (find_more_scalar_values_for_callers_subset): Use it.
3028 (find_aggregate_values_for_callers_subset): Likewise.
3029 (known_aggs_to_agg_replacement_list): Removed.
3030 (decide_whether_version_node): Re-calculate known constants for all
3031 remaining context clones.
3033 2018-04-11 Richard Biener <rguenther@suse.de>
3036 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3037 from early DWARF output.
3038 (dwarf2out_early_finish): Output line info unconditionally into
3039 early DWARF and add reference to it.
3041 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3044 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3045 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3046 other than V2DFmode using iptr mode attribute.
3047 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3049 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3051 PR rtl-optimization/84659
3052 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3054 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3057 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3059 (output_loc_list): Pass address of a dummy size variable even in the
3060 locview handling loop.
3061 (index_location_lists): Add comment on why skip_loc_list_entry can't
3064 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3067 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3070 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3073 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3074 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3075 and -mstring-compare-inline-limit.
3077 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3080 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3081 for stack clash protection in a register whenever we need it to be in
3084 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3086 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3087 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3089 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3092 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3094 (mlong-double-): Ditto.
3095 * config/rs6000/sysv4.opt (msdata=): Ditto.
3096 (mtls-size=): Ditto.
3098 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3100 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3101 erroneous entries for
3102 "vector int vec_ldl (int, long int *)", and
3103 "vector unsigned int vec_ldl (int, unsigned long int *)".
3104 Add comments and entries for
3105 "vector bool char vec_ldl (int, bool char *)",
3106 "vector bool short vec_ldl (int, bool short *)",
3107 "vector bool int vec_ldl (int, bool int *)",
3108 "vector bool long long vec_ldl (int, bool long long *)",
3109 "vector pixel vec_ldl (int, pixel *)",
3110 "vector long long vec_ldl (int, long long *)",
3111 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3112 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3113 type tree bool_long_long_type_node and correct definition of
3114 bool_V2DI_type_node to make reference to this new type tree.
3115 (rs6000_mangle_type): Replace erroneous reference to
3116 bool_long_type_node with bool_long_long_type_node.
3117 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3118 comments to emphasize sign distinctions for char and int types and
3119 replace RS6000_BTI_bool_long constant with
3120 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3121 use of RS6000_BTI_pixel.
3122 (bool_long_type_node): Remove this macro definition.
3123 (bool_long_long_type_node): New macro definition
3125 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3127 PR rtl-optimization/85300
3128 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3129 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3130 simplify_unary_operation fails.
3132 2018-04-10 Martin Liska <mliska@suse.cz>
3134 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3135 cgraph_edge and ipa_ref.
3137 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3141 * config/i386/sse.md
3142 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3143 computation of the VEC_MERGE selector from mask.
3144 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3145 Fix decoding of the VEC_MERGE selector into mask.
3147 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3149 PR tree-optimization/85286
3150 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3152 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3154 * final.c (final_1): Set insn_last_address as well as
3155 insn_current_address.
3157 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3160 * explow.c (emit_stack_probe): Call validize_mem on memory location
3161 before passing it to gen_probe_stack. Create address operand and
3162 legitimize it for the probe_stack_address case.
3164 2018-04-09 Jan Hubicka <jh@suse.cz>
3167 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3168 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3169 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3171 (free_lang_data): Rebuild type inheritance graph.
3173 2018-04-09 Martin Sebor <msebor@redhat.com>
3175 * invoke.texi (-finline-small-functions): Mention other optimization
3177 (-findirect-inlining, -fpartial-inlining): Same.
3178 (-finline-functions-called-once): Same.
3179 (-freorder-blocks-and-partition): Same.
3181 2018-04-09 Jan Hubicka <jh@suse.cz>
3184 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3185 jumps; choose last target that matches the criteria (i.e.
3186 no partition changes for non-crossing jumps).
3187 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3188 support for redirecting crossing jumps to non-crossing.
3190 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3192 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3193 also for naked functions.
3195 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3197 * config/arc/arc.md (add_shift): New pattern.
3198 (add_shift2): Likewise.
3199 (sub_shift): Likewise.
3200 (sub_shift_cmp0_noout): Likewise.
3201 (compare_si_ashiftsi): Likewise.
3202 (xbfu_cmp0_noout): New combine pattern.
3203 (xbfu_cmp0"): Likewise.
3204 (movsi_set_cc_insn): Place the predicable variant first.
3205 (commutative_binary_cmp0_noout): Remove clobber.
3206 (commutative_binary_cmp0): New pattern.
3207 (noncommutative_binary_cmp0): Likewise.
3208 (noncommutative_binary_cmp0_noout): Likewise.
3209 (noncommutative_binary_comparison_result_used): Removed.
3210 (rsub_cmp0): New pattern.
3211 (rsub_cmp0_noout): Likewise.
3212 (extzvsi): Changed, keep only meaningful variants.
3213 (SQH, SEZ): New iterators.
3214 (SQH_postfix): New mode attribute.
3215 (SEZ_prefix): New code attribute.
3216 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3217 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3218 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3220 (noncommutative_operator): Check the availability of barrel
3223 2018-04-09 Richard Biener <rguenther@suse.de>
3225 PR tree-optimization/85284
3226 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3227 Only use the niter constraining form of simple_iv when the exit
3230 2018-04-09 Tom de Vries <tom@codesourcery.com>
3233 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3234 (define_expand "*memory_barrier"): New define_expand.
3235 (define_insn "memory_barrier"): New insn.
3237 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3239 PR rtl-optimization/80463
3240 PR rtl-optimization/83972
3241 PR rtl-optimization/83480
3243 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3244 correct producer for the insn.
3245 (tidy_control_flow): Fixup seqnos in case of debug insns.
3247 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3249 PR rtl-optimization/83913
3251 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3252 different sched-times when merging exprs.
3254 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3256 PR rtl-optimization/83962
3258 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3259 tidy_fallthru_edge and tidy_control_flow.
3261 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3263 PR rtl-optimization/83530
3265 * sel-sched.c (force_next_insn): New global variable.
3266 (remove_insn_for_debug): When force_next_insn is true, also leave only
3267 next insn in the ready list.
3268 (sel_sched_region): When the region wasn't scheduled, make another pass
3269 over it with force_next_insn set to 1.
3271 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3273 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3275 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3276 for interrupt control.
3277 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3278 functions for interrupt control.
3279 * config/nds32/nds32-intrinsic.md: Likewise.
3280 * config/nds32/nds32_intrinsic.h: Likewise.
3281 * config/nds32/nds32.h (nds32_builtins): Likewise.
3283 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3285 * config/nds32/nds32.c (nds32_init_machine_status,
3286 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3287 strict_aligned_p field.
3288 (nds32_expand_to_rtl_hook): New function.
3289 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3290 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3292 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3293 Chung-Ju Wu <jasonwucj@gmail.com>
3295 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3296 * config/nds32/nds32-n7.md: New file.
3297 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3298 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3300 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3301 * config/nds32/nds32.md (pipeline_model): Add n7.
3302 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3303 * config/nds32/pipelines.md: Include n7 settings.
3305 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3306 Chung-Ju Wu <jasonwucj@gmail.com>
3308 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3309 * config/nds32/nds32-e8.md: New file.
3310 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3311 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3313 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3314 * config/nds32/nds32.md (pipeline_model): Add e8.
3315 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3316 * config/nds32/pipelines.md: Include e8 settings.
3318 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3319 Chung-Ju Wu <jasonwucj@gmail.com>
3321 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3322 * config/nds32/nds32-n8.md: New file.
3323 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3324 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3326 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3327 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3328 * config/nds32/nds32.md (pipeline_model): Add n8.
3329 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3330 * config/nds32/pipelines.md: Include n8 settings.
3332 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3333 Chung-Ju Wu <jasonwucj@gmail.com>
3335 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3336 * config/nds32/nds32-n9-2r1w.md: New file.
3337 * config/nds32/nds32-n9-3r2w.md: New file.
3338 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3339 nds32_register_ports): New or modify for cpu n9.
3340 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3342 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3343 * config/nds32/nds32-utils.c: New file.
3344 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3345 TARGET_MUL_SLOW): Define.
3346 * config/nds32/nds32.md (pipeline_model): New attribute.
3347 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3348 New options that support cpu n9.
3349 * config/nds32/pipelines.md: Include n9 settings.
3350 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3352 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3354 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3355 information if necessary.
3356 (output_cond_branch_compare_zero): Likewise.
3357 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3358 (nds32_target_alignment): Refine for alignment.
3359 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3360 (FUNCTION_BOUNDARY): Modify.
3361 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3363 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3365 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3367 * config/nds32/constants.md (unspec_volatile_element): Add values for
3368 TLB operation and data prefetch.
3369 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3370 functions for TLB operation and data prefetch.
3371 * config/nds32/nds32-intrinsic.md: Likewise.
3372 * config/nds32/nds32_intrinsic.h: Likewise.
3373 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3374 (nds32_print_operand): Likewise.
3375 * config/nds32/nds32.h (nds32_builtins): Likewise.
3377 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3378 Andrew Pinski <pinsika@gcc.gnu.org>
3381 * match.pd: Use constant_boolean_node of correct type instead of
3382 boolean_true_node or boolean_false_node for simplifying
3383 pointer comparisons to zero.
3385 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3387 PR tree-optimization/80021
3388 * tree.c (verify_type_variant): Make error call in verify_variant_match
3389 translatable and remove final full stop.
3391 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3393 * config/nds32/constants.md (unspec_volatile_element): Add
3394 UNSPEC_VOLATILE_EH_RETURN.
3395 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3396 nds32_output_stack_pop): Support dwarf exception handling process.
3397 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3398 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3399 exception handling process.
3400 (nds32_compute_stack_frame): Likewise.
3401 (nds32_return_addr_rtx): Likewise.
3402 (nds32_initial_elimination_offset): Likewise.
3403 (nds32_expand_prologue): Likewise.
3404 (nds32_expand_epilogue): Likewise.
3405 (nds32_dynamic_chain_address): New function.
3406 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3408 (DYNAMIC_CHAIN_ADDRESS): Define.
3409 (EH_RETURN_DATA_REGNO): Define.
3410 (EH_RETURN_STACKADJ_RTX): Define.
3411 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3412 patterns for dwarf exception handling.
3414 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3416 * config/nds32/nds32.h: Clean up obsolete macros.
3418 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3420 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3421 Add enum values for particular instructions.
3422 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3423 particular intrinsic functions.
3424 * config/nds32/nds32-intrinsic.md: Likewise.
3425 * config/nds32/nds32_intrinsic.h: Likewise.
3426 * config/nds32/nds32.h (nds32_builtins): Likewise.
3427 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3428 (btst, ave): New patterns for particular instructions.
3430 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3432 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3433 Add enum values for atomic load/store and memory sync.
3434 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3436 * config/nds32/nds32-intrinsic.md: Likewise.
3437 * config/nds32/nds32_intrinsic.h: Likewise.
3438 * config/nds32/nds32.h (nds32_builtins): Likewise.
3440 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3442 PR tree-optimization/85257
3443 * fold-const.c (native_encode_vector): If not all elts could fit
3444 and off is -1, return 0 rather than offset.
3445 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3446 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3447 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3448 adjust buffer in native_interpret_expr call.
3450 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3452 * config/nds32/constants.md (unspec_volatile_element): Add cache
3453 control enum values.
3454 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3455 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3456 * config/nds32/nds32.c (nds32_cctl_names): New.
3457 (nds32_print_operand): Handle cache control register names.
3458 * config/nds32/nds32.h (nds32_builtins): New enum values.
3459 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3461 * config/nds32/nds32.md (type): Add mmu.
3462 * config/nds32/pipelines.md (simple_insn): Add mmu.
3464 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3466 * config/nds32/nds32.md (type): Remove call.
3467 * config/nds32/pipelines.md (simple_insn): Likewise.
3469 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3471 * config/nds32/constants.md (unspec_volatile_element): Add
3472 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3473 UNSPEC_VOLATILE_FMFCFG.
3474 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3475 description for fmfcfg and fmfcsr.
3476 (bdesc_1arg): Add fmtcsr.
3477 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3478 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3479 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3480 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3481 unspec_fmfcfg): New patterns.
3482 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3483 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3484 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3485 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3486 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3487 __nds32__fmfcfg): Define.
3489 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3491 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3492 intrinsic register names.
3493 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3494 intrinsic register enum values and macros.
3496 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3498 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3499 for load/store addressing form.
3500 (nds32_print_operand_address): Likewise.
3502 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3505 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3506 based on LABEL_REF. Remove useless assertion.
3507 (pic_address_needs_scratch): Fix formatting.
3508 (sparc_legitimize_pic_address): Minor tweaks.
3509 (sparc_delegitimize_address): Adjust assertion accordingly.
3510 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3511 into symbolic_operand.
3512 (movsi_high_pic_label_ref): Likewise.
3513 (movsi_lo_sum_pic_label_ref): Likewise.
3514 (movdi_pic_label_ref): Likewise.
3515 (movdi_high_pic_label_ref): Likewise.
3516 (movdi_lo_sum_pic_label_ref): Likewise.
3518 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3520 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3521 custom LIB_SPEC setup.
3523 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3524 Kito Cheng <kito.cheng@gmail.com>
3526 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3527 * config/riscv/freebsd.h: New.
3529 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3531 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3532 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3535 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3536 Kito Cheng <kito.cheng@gmail.com>
3538 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3539 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3540 * config/nds32/nds32-protos.h (nds32_output_call,
3541 nds32_output_return): Declare.
3542 * config/nds32/nds32.md: Refine all the call and return patterns.
3544 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3547 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3548 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3550 PR rtl-optimization/84872
3551 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3552 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3555 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3557 * expr.c (copy_blkmode_to_reg): Revert 254862.
3558 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3560 2018-04-06 Richard Biener <rguenther@suse.de>
3563 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3564 after seeing a component reference with an adjacent field. Treat
3565 refs to arrays at struct end of external decls similar to
3566 refs to unconstrained commons.
3568 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3571 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3572 look through SAVE_EXPRs with non-side-effects argument. Adjust
3574 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3577 2018-04-06 Richard Biener <rguenther@suse.de>
3580 * alias.c (find_base_term): New wrapper around find_base_term
3581 unwinding CSELIB_VAL_PTR changes.
3582 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3585 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3587 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3589 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3590 constant definitions.
3591 ("nop"): lr 0,0 -> nopr r0
3592 ("nop_lr0", "nop_lr1"): New insn definitions.
3594 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3596 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3597 NDS32_V3PUSH_AVAILABLE_P macro.
3599 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3600 Chung-Ju Wu <jasonwucj@gmail.com>
3602 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3603 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3604 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3605 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3606 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3607 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3608 * config/nds32/constraints.md: New constraints and checking for hard
3609 float configuration.
3610 * config/nds32/iterators.md: New mode iterator and attribute for hard
3611 float configuration.
3612 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3614 * config/nds32/nds32-fpu.md: New file.
3615 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3616 deal with hard float code generation.
3617 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3619 (abi_type, float_reg_number): New enum type.
3620 * config/nds32/nds32-predicates.c: New predicates for hard float.
3621 * config/nds32/nds32-protos.h: Declare functions for hard float.
3622 * config/nds32/nds32.c: Implementation for hard float configuration.
3623 * config/nds32/nds32.h: Definitions for hard float configuration.
3624 * config/nds32/nds32.md: Include hard float machine description and
3625 modify patterns for hard float configuration.
3626 * config/nds32/nds32.opt: New options for hard float configuration.
3627 * config/nds32/predicates.md: New predicates for hard float
3630 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3632 * common/config/nds32/nds32-common.c
3633 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3635 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3638 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3639 CONSTRUCTOR_ELT (ctor, ...)->value.
3641 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3644 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3646 2018-04-05 Tom de Vries <tom@codesourcery.com>
3649 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3652 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3653 Kito Cheng <kito.cheng@gmail.com>
3655 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3656 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3657 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3658 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3660 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3661 Kito Cheng <kito.cheng@gmail.com>
3663 * config/nds32/constraints.md (Ufe): New memory constraint.
3664 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3665 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3666 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3668 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3669 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3671 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3673 * config/nds32/nds32.md: Use optimize_size in the condition for
3674 alu-shift instructions.
3676 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3678 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3680 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3682 * config/nds32/nds32.md (negsi2): Refine pattern.
3684 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3685 Chung-Ju Wu <jasonwucj@gmail.com>
3687 * config/nds32/iterators.md (shift_rotate): New code iterator.
3688 (shift): New code attribute.
3689 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3690 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3691 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3692 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3693 bit-wise operations.
3694 (andsi3, *andsi3): Ditto.
3695 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3696 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3697 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3698 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3699 nds32_ior_operand, nds32_xor_operand): New predicates.
3701 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3703 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3704 (addsi3, subsi3): ... this.
3706 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3708 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3710 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3712 * config/nds32/nds32.md: Adjust indention.
3714 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3716 * config/nds32/nds32.md (feature): New attribute.
3718 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3720 * config/nds32/nds32.md (subtype): New attribute.
3722 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3725 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3726 expansion to perform a bitwise AND of the argument followed by a
3727 boolean negation of the result.
3729 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3731 PR rtl-optimization/84878
3732 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3733 the basic block. Assert the use reference is not artificial and that
3734 it has an associated insn.
3736 2018-04-04 Michael Matz <matz@suse.de>
3738 * builtins.c (compute_objsize): Pass correct operand
3739 to array_at_struct_end_p.
3741 2018-04-04 Richard Biener <rguenther@suse.de>
3744 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3745 from contexts for DINFO_LEVEL_TERSE and below.
3747 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3749 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3750 resiter_operand condition.
3751 * config/nds32/nds32.md (*move<mode>): Ditto.
3753 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3754 Monk Chiang <sh.chiang04@gmail.com>
3756 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3758 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3760 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3762 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3763 Kito Cheng <kito.cheng@gmail.com>
3765 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3766 nds32_cond_code_str, output_cond_branch,
3767 output_cond_branch_compare_zero, nds32_expand_cbranch,
3768 nds32_expand_cstore, nds32_expand_movcc,
3769 nds32_output_cbranchsi4_equality_zero,
3770 nds32_output_cbranchsi4_equality_reg,
3771 nds32_output_cbranchsi4_equality_reg_or_const_int,
3772 nds32_output_cbranchsi4_greater_less_zero: New functions.
3773 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3774 nds32_expand_cstore, nds32_expand_movcc,
3775 nds32_output_cbranchsi4_equality_zero,
3776 nds32_output_cbranchsi4_equality_reg,
3777 nds32_output_cbranchsi4_equality_reg_or_const_int,
3778 nds32_output_cbranchsi4_greater_less_zero): Declare.
3779 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3780 nds32_rimm11s_operand): New predicates.
3781 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3782 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3785 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3787 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3788 * config/nds32/nds32.md: Ditto.
3789 * config/nds32/pipelines.md: Ditto.
3791 2018-04-04 Richard Biener <rguenther@suse.de>
3793 PR tree-optimization/85168
3794 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3795 propagating abnormals.
3797 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3799 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3801 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3802 Kito Cheng <kito.cheng@gmail.com>
3804 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3805 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3806 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3807 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3808 * config/nds32/nds32.md (sibcall_internal): New.
3809 (sibcall_register): Remove.
3810 (sibcall_immediate): Remove.
3811 (sibcall_value_internal): New.
3812 (sibcall_value_register): Remove.
3813 (sibcall_value_immediate): Remove.
3814 * config/nds32/predicates.md (nds32_general_register_operand): New.
3815 (nds32_call_address_operand): New.
3817 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3819 PR rtl-optimization/85167
3820 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3821 bb_defs if *split_p, instead preinitialize it to NULL.
3823 PR tree-optimization/85156
3824 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3825 evaluating the argument multiple times.
3827 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3829 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3831 (_mm_cvtpd_ps): Likewise.
3832 (_mm_cvttpd_epi32): Likewise.
3833 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3834 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3835 vector, pixel, and bool following altivec.h include.
3837 2018-04-03 Martin Sebor <msebor@redhat.com>
3839 * doc/extend.texi (Common Function Attributes): Clarify.
3840 (const attribute): Likewise.
3841 (pure attribute): Likewise.
3843 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3846 * config/i386/i386.c (ix86_expand_vector_set): Use
3847 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3849 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3851 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3852 instructions when changing rounding bits to preserve precision bits
3853 in the x87 control word.
3855 2018-04-03 Martin Liska <mliska@suse.cz>
3857 PR tree-optimization/82491
3858 * rtl.h (strip_offset_and_add): Replace += suboffset with
3859 poly_uint64 () + suboffset.
3861 2018-03-29 Martin Liska <mliska@suse.cz>
3862 Martin Jambor <mjambor@suse.cz>
3865 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3866 param_type is not an integral or pointer type.
3868 2018-04-03 Richard Biener <rguenther@suse.de>
3870 * sese.h (recompute_all_dominators): Remove.
3872 2018-04-02 Martin Sebor <msebor@redhat.com>
3874 * doc/invoke.texi (-Wrestrict): Fix typos.
3876 2018-04-02 Jim Wilson <jimw@sifive.com>
3878 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3879 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3880 (<optab>di3, <optab>si3_extend): Likewise.
3881 (<optab>si3_mask, <optab>si3_mask_1): New.
3882 (<optab>di3_mask, <optab>di3_mask_1): New.
3883 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3884 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3885 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3887 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3889 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3892 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3894 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3895 (nds32_canonicalize_comparison): New function.
3897 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3898 Kito Cheng <kito.cheng@gmail.com>
3899 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3901 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3902 * config/nds32/constants.md (unspec_volatile_element): Add
3903 UNSPEC_VOLATILE_RELAX_GROUP.
3904 * config/nds32/nds32-relax-opt.c: New file.
3905 * config/nds32/nds32-predicates.c
3906 (nds32_symbol_load_store_p): New function.
3907 * config/nds32/nds32-protos.h
3908 (nds32_symbol_load_store_p): Declare function.
3909 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3910 * config/nds32/nds32.c
3911 (nds32_register_pass): New function to register pass.
3912 (nds32_register_passes): New function to register passes.
3913 * config/nds32/nds32.md (relax_group): New pattern.
3914 * config/nds32/nds32.opt (mrelax-hint): New option.
3915 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3917 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3919 * config/nds32/t-nds32: Modify files dependency.
3921 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3923 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3924 (PROFILE_HOOK): Define its implementation.
3926 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3928 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3929 type and 32-bit size.
3931 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3934 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3935 (V_128_256): New mode iterator.
3936 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3937 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3938 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3940 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3941 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3943 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3946 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3947 NaN inputs correctly.
3949 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3952 * config/rs6000/vsx.md (??r): New mode attribute.
3953 (*vsx_mov<mode>_64bit): Use it.
3954 (*vsx_mov<mode>_32bit): Likewise.
3956 2018-03-30 Martin Sebor <msebor@redhat.com>
3958 PR tree-optimization/84818
3959 * builtins.c (check_access): Use warning_n.
3961 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3964 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3966 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3969 2018-03-30 Julia Koval <julia.koval@intel.com>
3972 * x86-tune.def (movx, partial_reg_dependency): Enable for
3975 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3978 * lra-constraints.c (process_alt_operands): Move setting
3979 this_alternative_matches below.
3981 2018-03-29 Martin Liska <mliska@suse.cz>
3984 * doc/invoke.texi: Document how LTO works with debug info.
3985 Describe auto-load support of binutils. Mention 'x86-64'
3986 as valid option value of -march option.
3988 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3990 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3993 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3994 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3995 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3998 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4001 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4002 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4003 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4004 for RS6000_BTM_POWERPC64.
4005 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4006 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4007 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4012 2018-03-28 Carl Love <cel@us.ibm.com>
4015 2017-09-27 Carl Love <cel@us.ibm.com>
4017 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4018 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4019 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4022 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4024 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4025 instead of __vector bool.
4026 (_mm_max_pu8): Likewise.
4027 (_mm_min_pi16): Likewise.
4029 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4032 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4033 (DIVWEUO): Likewise.
4035 (DIVDEUO): Likewise.
4036 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4037 DIVWEUO and DIVDEUO.
4038 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4039 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4040 (div_extend): Likewise.
4041 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4043 (__builtin_divweuo): Likewise.
4044 (__builtin_divdeo): Likewise.
4045 (__builtin_divdeuo): Likewise.
4047 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4050 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4051 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4053 PR tree-optimization/82004
4054 * gimple-match-head.c (optimize_pow_to_exp): New function.
4055 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4056 Don't fold to exp if optimize_pow_to_exp is false.
4058 2018-03-28 Martin Liska <mliska@suse.cz>
4061 * calls.c (initialize_argument_information): Fix trailing space.
4062 * common.opt: Fix typo and provide better explanation for
4063 -fsanitize-coverage option.
4064 * config/i386/i386.opt: Fix typo.
4066 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4067 Martin Liska <mliska@suse.cz>
4070 * gimplify.c (asan_poison_variable): Don't do the check for
4071 gimplify_omp_ctxp here.
4072 (gimplify_decl_expr): Do it here.
4073 (gimplify_target_expr): Likewise.
4075 2018-03-28 Martin Liska <mliska@suse.cz>
4078 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4079 chkp_type_bounds_count if MPX is not enabled.
4081 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4083 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4085 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4088 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4089 function to create the function decl for complex long double
4090 multiply and divide for -mabi=ieeelongdouble.
4091 (init_float128_ieee): Call it.
4093 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4096 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4097 -fcf-protection=branch -mibt.
4098 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4100 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4103 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4105 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4108 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4109 extern array declarations.
4111 2018-03-27 Richard Biener <rguenther@suse.de>
4114 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4115 explicit single_use checks.
4117 2018-03-27 Richard Biener <rguenther@suse.de>
4119 PR tree-optimization/85082
4120 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4123 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4125 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4126 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4127 Turn on fasynchronous-unwind-tables and funwind-tables.
4129 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4132 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4133 (*bmi_blsr_<mode>_ccz): Ditto.
4135 2018-03-26 Tom de Vries <tom@codesourcery.com>
4137 PR tree-optimization/85063
4138 * omp-general.c (offloading_function_p): New function. Factor out
4140 * omp-offload.c (pass_omp_target_link::gate): ... here.
4141 * omp-general.h (offloading_function_p): Declare.
4142 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4143 with attribute omp declare target for offloading functions.
4145 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4147 PR tree-optimization/84005
4148 * tree-data-ref.h (get_base_for_alignment): Declare.
4149 * tree-data-ref.c (get_base_for_alignment_1): New function.
4150 (get_base_for_alignment): Likewise.
4151 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4152 get_base_for_alignment to find a suitable base object, instead
4153 of always using drb->base_address.
4155 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4158 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4159 known size by default.
4161 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4164 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4165 and non BLKmode operands.
4167 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4170 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4171 Clean up attributes.
4173 2018-03-23 Richard Biener <rguenther@suse.de>
4176 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4177 we are going to emit early debug for LTO.
4179 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4182 * function.c (match_asm_constraints_1): Don't optimize if input
4183 doesn't satisfy general_operand predicate for output's mode.
4186 * alias.c (write_dependence_p): Don't require for x_canonicalized
4187 non-VOIDmode if x has VOIDmode.
4190 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4191 just don't try to optimize it rather than assert it never happens.
4193 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4195 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4196 macro expansions for definition of ST_INTERNAL_<mode> and
4197 LD_INTERNAL_<mode> builtins.
4198 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4200 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4202 (altivec_expand_st_builtin): Likewise.
4203 (altivec_expand_builtin): Remove calls to deleted functions.
4204 (rs6000_address_for_altivec): Delete this function.
4205 * config/rs6000/vector.md: Remove expands for
4206 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4208 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4211 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4212 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4213 re-computing once computed.
4214 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4215 (arm_init_machine_status): Initialize
4216 machine->static_chain_stack_bytes.
4218 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4221 * doc/extend.texi: Add four new prototypes for vec_ld.
4222 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4223 definitions for more logical presentation.
4224 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4225 entries for V1TI variants of __builtin_altivec_ld builtin.
4226 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4227 handling of V1TI variant of LVX icode pattern.
4228 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4229 (rs6000_gimple_fold_builtin): Likewise.
4230 (altivec_init_builtins): Add code to define
4231 __builtin_altivec_lvx_v1ti function.
4233 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4236 * function.c (match_asm_constraints_1): Don't do the optimization
4237 if input isn't a REG, SUBREG, MEM or constant.
4239 2018-03-22 Tom de Vries <tom@codesourcery.com>
4241 PR tree-optimization/84956
4242 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4243 bb_has_abnormal_pred.
4245 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4248 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4249 DECL_INITIAL (decl) to decl at the end.
4250 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4253 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4255 * doc/extend.texi (__builtin_tgmath): Document when complex
4256 integer types are treated as _Complex _Float64.
4258 2018-03-21 Tom de Vries <tom@codesourcery.com>
4260 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4262 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4264 PR tree-optimization/84960
4265 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4266 if it is ENTRY block, move them into single succ of ENTRY in that case.
4268 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4270 PR tree-optimization/84811
4271 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4272 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4273 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4274 (known_subrange_p): Update accordingly. Cast each value involved
4275 in the size comparison, rather than casting the result of the
4278 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4280 PR tree-optimization/84982
4281 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4282 by flipping the least significant bit rather than all bits from
4283 bitpos to bitpos + bitsize - 1.
4285 2018-03-21 Nathan Sidwell <nathan@acm.org>
4287 * doc/extend.texi (Deprecated Features): Remove mention of
4288 long-deleted deprecations.
4290 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4293 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4294 * configure: Regenerate.
4296 2018-03-21 Tom de Vries <tom@codesourcery.com>
4298 PR tree-optimization/83126
4299 * tree-parloops.c (num_phis): New function.
4300 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4302 2018-03-21 Nathan Sidwell <nathan@acm.org>
4304 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4305 mention anon-struct/union members and trailing attributes.
4307 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4309 PR tree-optimization/84969
4310 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4311 builtin memset partitions if they set different rhs values.
4313 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4315 PR rtl-optimization/84989
4316 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4317 VEC_DUPLICATE with scalar result mode.
4319 2018-03-21 Martin Liska <mliska@suse.cz>
4322 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4323 not intended return statement.
4325 2018-03-21 Martin Liska <mliska@suse.cz>
4328 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4329 (chkp_find_bound_slots_1): Limit number of iterations.
4331 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4334 * Minor grammar fixes for x86 options.
4336 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4339 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4340 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4342 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4345 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4346 (altivec_lvx_<mode>_1op): Likewise.
4347 (altivec_stvx_<mode>_2op): Likewise.
4348 (altivec_stvx_<mode>_1op): Likewise.
4349 (altivec_lvx_<VM2:mode>): New define_expand.
4350 (altivec_stvx_<VM2:mode>): Likewise.
4351 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4352 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4353 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4354 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4355 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4356 (rs6000_gen_lvx): Likewise.
4357 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4358 (altivec_expand_stv_builtin): Likewise.
4359 (altivec_expand_builtin): Likewise.
4360 * config/rs6000/vector.md: Likewise.
4362 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4365 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4368 2018-03-20 Richard Biener <rguenther@suse.de>
4371 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4372 sign-conversions as zero, fall back to standard scalar_stmt
4375 2018-03-20 Martin Liska <mliska@suse.cz>
4378 * predict.c (rebuild_frequencies): Handle case when we have
4379 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4381 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4384 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4385 flag_section_anchors.
4386 * varasm.c (use_blocks_for_decl_p): Remove hack for
4387 dw2_force_const_mem.
4390 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4392 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4393 be created, use lowpart_subreg of operands[0] rather than operands[0]
4395 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4396 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4397 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4398 and n constraint instead of aarch64_shift_imm_di and Usd.
4399 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4400 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4402 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4405 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4406 to favor GPR over NEON registers.
4407 (<shift>di3_neon): Likewise.
4409 2018-03-20 Tom de Vries <tom@codesourcery.com>
4412 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4413 (nvptx_process_pars): Emit bar.sync asap and alap.
4415 2018-03-20 Tom de Vries <tom@codesourcery.com>
4418 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4419 seen_label if seen_label is already set.
4421 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4424 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4425 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4426 Use 1U instead of 1. Formatting fixes.
4429 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4430 instead of TREE_TYPE (s1) for the return value.
4432 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4434 PR tree-optimization/84946
4435 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4436 bitsize + bitsize in poly_uint64 rather than poly_int64.
4439 * dwarf2asm.c: Include fold-const.c.
4440 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4441 of decl rather than decl itself.
4443 PR rtl-optimization/84643
4444 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4446 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4449 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4450 calling assemble_variable.
4452 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4455 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4456 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4458 2018-03-19 Jim Wilson <jimw@sifive.com>
4461 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4462 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4463 (riscv_first_stack_step): Likewise.
4464 (riscv_option_override): Use STACK_BOUNDARY instead of
4466 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4468 (BIGGEST_ALIGNMENT): Set to 128.
4469 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4470 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4473 2018-03-19 Richard Biener <rguenther@suse.de>
4475 PR tree-optimization/84933
4476 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4477 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4479 2018-03-19 Richard Biener <rguenther@suse.de>
4481 PR tree-optimization/84859
4482 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4483 (cond_if_else_store_replacement): Perform sinking operation on
4484 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4485 Generalize what a BB with a single eligible store is.
4487 2018-03-19 Richard Biener <rguenther@suse.de>
4489 PR tree-optimization/84929
4490 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4491 chrec_is_positive against non-chrec arg.
4493 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4496 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4498 2018-03-18 Martin Liska <mliska@suse.cz>
4500 PR rtl-optimization/84635
4501 * regrename.c (build_def_use): Use matches_mode only when
4504 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4506 PR tree-optimization/84913
4507 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4508 vectorize chains of COND_EXPRs.
4510 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4512 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4514 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4516 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4518 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4520 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4522 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4523 Kito Cheng <kito.cheng@gmail.com>
4525 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4526 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4527 (nds32_adjust_reg_alloc_order): New function.
4528 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4530 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4532 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4533 nds32_print_operand, nds32_print_operand_address): Use
4534 HOST_WIDE_INT_PRINT_DEC instead.
4536 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4538 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4540 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4543 * config/i386/i386.c (initial_ix86_tune_features,
4544 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4546 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4547 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4548 rather than 1u << ix86_tune. Formatting fix.
4549 (ix86_option_override_internal): Change ix86_arch_mask from
4550 unsigned int to unsigned HOST_WIDE_INT, initialize to
4551 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4552 (ix86_function_specific_restore): Likewise.
4554 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4557 * postreload.c (reload_combine_recognize_pattern): Perform
4558 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4559 truncate_int_for_mode the result for the destination's mode.
4562 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4563 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4565 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4568 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4571 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4572 regno_allocno_class_array and sorted_pseudos.
4573 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4574 insns where regno is used.
4576 2018-03-16 Martin Liska <mliska@suse.cz>
4579 * multiple_target.c (create_dispatcher_calls): Redirect
4580 reference in the symbol table.
4582 2018-03-16 Martin Liska <mliska@suse.cz>
4585 * multiple_target.c (create_dispatcher_calls): Redirect also
4588 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4592 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4593 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4594 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4596 2018-03-16 Julia Koval <julia.koval@intel.com>
4598 * doc/invoke.texi (Skylake Server): Add CLWB.
4599 Cannonlake): Remove CLWB.
4601 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4603 PR tree-optimization/84841
4604 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4606 (FLOAT_ONE_CONST_TYPE): Define.
4607 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4608 (sort_by_operand_rank): Put entries with higher constant_type last
4609 rather than first to match comments.
4611 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4613 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4616 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4619 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4620 check and use address of target always.
4622 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4625 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4626 (indirect_thunk_bnd_needed): Likewise.
4627 (indirect_thunks_used): Likewise.
4628 (indirect_thunks_bnd_used): Likewise.
4629 (indirect_return_needed): New.
4630 (indirect_return_bnd_needed): Likewise.
4631 (output_indirect_thunk_function): Add a bool argument for
4633 (output_indirect_thunk_function): Don't generate alias for
4634 function return thunk.
4635 (ix86_code_end): Call output_indirect_thunk_function to generate
4636 function return thunks.
4637 (ix86_output_function_return): Set indirect_return_bnd_needed
4638 and indirect_return_needed instead of indirect_thunk_bnd_needed
4639 and indirect_thunk_needed.
4641 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4643 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4644 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4645 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4647 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4648 Paul Hua <paul.hua.gm@gmail.com>
4651 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4653 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4655 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4656 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4659 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4662 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4663 instead of GET_MODE_SIZE when comparing Units.
4665 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4668 * varasm.c (hash_section): Return an unchangeble hash value
4669 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4670 Return !aarch64_can_use_per_function_literal_pools_p ().
4672 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4675 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4676 rather than address of cmode as last argument to prepare_cmp_insn.
4678 2018-03-15 Julia Koval <julia.koval@intel.com>
4680 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4681 F_AVX512VNNI, F_AVX512BITALG): New.
4683 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4686 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4687 insn for floating-point loads and stores.
4689 2018-03-14 Carl Love <cel@us.ibm.com>
4691 * config/rs6000/rs6000-c.c: Add macro definitions for
4692 ALTIVEC_BUILTIN_VEC_PERMXOR.
4693 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4694 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4695 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4696 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4698 * config/doc/extend.texi: Add prototypes for vec_permxor.
4700 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4703 * diagnostic-show-locus.c (class layout_point): Convert m_line
4704 from int to linenum_type.
4705 (line_span::comparator): Use linenum "compare" function when
4706 comparing line numbers.
4707 (test_line_span): New function.
4708 (layout_range::contains_point): Convert param "row" from int to
4710 (layout_range::intersects_line_p): Likewise.
4711 (layout::will_show_line_p): Likewise.
4712 (layout::print_source_line): Likewise.
4713 (layout::should_print_annotation_line_p): Likewise.
4714 (layout::print_annotation_line): Likewise.
4715 (layout::print_leading_fixits): Likewise.
4716 (layout::annotation_line_showed_range_p): Likewise.
4717 (struct line_corrections): Likewise for field m_row.
4718 (line_corrections::line_corrections): Likewise for param "row".
4719 (layout::print_trailing_fixits): Likewise.
4720 (layout::get_state_at_point): Likewise.
4721 (layout::get_x_bound_for_row): Likewise.
4722 (layout::print_line): Likewise.
4723 (diagnostic_show_locus): Likewise for locals "last_line" and
4725 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4726 * input.c (selftest::test_linenum_comparisons): New function.
4727 (selftest::input_c_tests): Call it.
4728 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4729 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4730 * selftest.h (ASSERT_GT): New macro.
4731 (ASSERT_GT_AT): New macro.
4732 (ASSERT_LT): New macro.
4733 (ASSERT_LT_AT): New macro.
4735 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4737 PR rtl-optimization/84780
4738 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4740 2018-03-14 Martin Liska <mliska@suse.cz>
4742 * tree.c (record_node_allocation_statistics): Use
4743 get_stats_node_kind.
4744 (get_stats_node_kind): New function extracted from
4745 record_node_allocation_statistics.
4746 (free_node): Use get_stats_node_kind.
4748 2018-03-14 Richard Biener <rguenther@suse.de>
4750 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4751 that the value-set of ANTIC_IN doesn't grow.
4754 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4756 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4757 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4759 2018-03-14 Julia Koval <julia.koval@intel.com>
4761 * config.gcc (icelake-client, icelake-server): New.
4763 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4764 (initial_ix86_arch_features): Ditto.
4765 (PTA_SKYLAKE): Add SGX.
4766 (PTA_ICELAKE): Remove.
4767 (PTA_ICELAKE_CLIENT): New.
4768 (PTA_ICELAKE_SERVER): New.
4769 (ix86_option_override_internal): Split up icelake on icelake client and
4771 (get_builtin_code_for_version): Ditto.
4772 (fold_builtin_cpu): Ditto.
4773 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4774 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4775 * config/i386/i386.h (processor_type): Ditto.
4776 * doc/invoke.texi: Ditto.
4778 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4781 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4782 INTEGER_CST offset, add it together with bitpos / 8 and
4783 sign extend based on POINTER_SIZE.
4787 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4790 * config/i386/constraints.md (Yc): New register constraint.
4791 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4792 Use Yc constraint for alternative 2 of operand 0. Remove
4793 preferred_for_speed attribute.
4795 2018-03-14 Richard Biener <rguenther@suse.de>
4797 PR tree-optimization/84830
4798 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4799 with the old one to avoid oscillations.
4801 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4804 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4806 (assign_by_spills): Return a flag of reload assignment failure.
4807 Do not process the reload assignment failures. Do not spill other
4808 reload pseudos if they has the same reg class. Update n if
4810 (lra_assign): Add a return arg. Set up from the result of
4811 assign_by_spills call.
4812 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4813 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4814 usage_insns if it is not NULL.
4815 (spill_hard_reg_in_range): New function.
4816 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4817 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4818 function prototypes.
4819 (lra_assign): Change prototype.
4820 * lra.c (lra): Add code to deal with fails by splitting hard reg
4823 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4825 * config/riscv/riscv.opt (mrelax): New option.
4826 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4827 "norelax" when riscv_mrelax is disabled.
4828 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4830 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4833 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4834 reassociation for int modes.
4836 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4838 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4839 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4841 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4842 * config/aarch64/aarch64-sve.md
4843 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4844 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4845 (*extend<mode><Vwide>2): Rename to...
4846 (aarch64_sve_extend<mode><Vwide>2): ...this.
4847 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4848 renaming the old pattern to...
4849 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4851 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4852 define_expand, renaming the old pattern to...
4853 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4854 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4855 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4856 account when deciding which SVE instruction the optab should use.
4857 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4859 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4861 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4862 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4863 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4864 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4865 (tlsdesc_small_<mode>): Turn a define_expand and use
4866 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4867 (tlsdesc_small_advsimd_<mode>): ...this.
4868 (tlsdesc_small_sve_<mode>): New pattern.
4870 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4872 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4873 (UNSPEC_UMUL_HIGHPART): New constants.
4874 (MUL_HIGHPART): New int iteraor.
4875 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4876 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4878 (*<su>mul<mode>3_highpart): New define_insn.
4880 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4883 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4886 2018-03-13 Martin Liska <mliska@suse.cz>
4889 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4891 (sem_item_optimizer::~sem_item_optimizer): Release it.
4892 (sem_item_optimizer::merge_classes): Register variable aliases.
4893 (sem_item_optimizer::fixup_pt_set): New function.
4894 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4895 * ipa-icf.h: Declare new variables and functions.
4897 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4900 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4901 integer_pow2p@2 and test integer_pow2p in condition.
4902 (A < 0 ? C : 0): Similarly for @1.
4905 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4906 characters starting at p contain '\0' character, don't look beyond
4910 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4911 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4914 * reg-stack.c (change_stack): Change update_end var from int to
4915 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4916 also call set_block_for_insn on the newly added insns and rescan.
4919 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4920 on the last operand.
4923 * tree.c (stabilize_reference_1): Return save_expr (e) for
4924 STATEMENT_LIST even if it doesn't have side-effects.
4926 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4928 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4930 2018-03-12 Renlin Li <renlin.li@arm.com>
4932 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4933 aarch64_output_scalar_simd_mov_immediate.
4935 2018-03-12 Martin Sebor <msebor@redhat.com>
4937 PR tree-optimization/83456
4938 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4939 for perfectly overlapping calls to memcpy.
4940 (gimple_fold_builtin_memory_chk): Same.
4941 (gimple_fold_builtin_strcpy): Handle no-warning.
4942 (gimple_fold_builtin_stxcpy_chk): Same.
4943 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4945 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4947 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4948 parameter. Use it for SFmode.
4949 (rs6000_function_arg_advance_1): Adjust.
4950 (rs6000_function_arg): Adjust.
4951 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4953 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4955 PR rtl-optimization/84169
4956 PR rtl-optimization/84780
4957 * combine.c (can_combine_p): Check for a 2-insn combination whether
4958 the destination register is used between the two insns, too.
4960 2018-03-12 Richard Biener <rguenther@suse.de>
4962 PR tree-optimization/84803
4963 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4964 for refs DR analysis didn't process.
4966 2018-03-12 Richard Biener <rguenther@suse.de>
4968 PR tree-optimization/84777
4969 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4970 force-vectorize loops ignore whether we are optimizing for size.
4972 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4974 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4975 (TARGET_MD_ASM_ADJUST): Define.
4977 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4978 Kito Cheng <kito.cheng@gmail.com>
4979 Chung-Ju Wu <jasonwucj@gmail.com>
4981 * config/nds32/nds32.c (nds32_compute_stack_frame,
4982 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4983 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4984 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4985 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4986 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4987 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4988 * config/nds32/nds32.md (prologue, epilogue): Use macro
4989 NDS32_V3PUSH_AVAILABLE_P to do checking.
4991 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4994 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4995 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4996 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4997 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4998 addition of most attributes on !orig_type_die or the attribute not
4999 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5001 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5002 Chung-Ju Wu <jasonwucj@gmail.com>
5004 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5006 * config/nds32/nds32.opt (mvh): New option.
5008 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5009 Chung-Ju Wu <jasonwucj@gmail.com>
5011 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5013 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5014 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5017 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5018 Chung-Ju Wu <jasonwucj@gmail.com>
5020 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5022 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5023 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5025 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5026 Kito Cheng <kito.cheng@gmail.com>
5027 Chung-Ju Wu <jasonwucj@gmail.com>
5029 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5030 UNSPEC_FFMISM and UNSPEC_FLMISM.
5031 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5032 for ffb, ffmism and flmism.
5033 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5034 (unspec_ffmism): Ditto.
5035 (unspec_flmism): Ditto.
5036 (nds32_expand_builtin_impl): Check if string extension is available.
5037 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5038 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5040 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5043 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5046 * lra-assigns.c (assign_by_spills): Return a flag of reload
5047 assignment failure. Do not process the reload assignment
5048 failures. Do not spill other reload pseudos if they has the same
5050 (lra_assign): Add a return arg. Set up from the result of
5051 assign_by_spills call.
5052 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5053 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5054 usage_insns if it is not NULL.
5055 (spill_hard_reg_in_range): New function.
5056 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5057 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5058 function prototypes.
5059 (lra_assign): Change prototype.
5060 * lra.c (lra): Add code to deal with fails by splitting hard reg
5063 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5066 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5068 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5071 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5072 (dw_val_node): Add val_symbolic_view.
5073 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5074 (symview_upper_bound): New.
5075 (new_line_info_table): Initialize symviews_since_reset.
5076 (dwarf2out_source_line): Count symviews_since_reset and set
5077 symview_upper_bound.
5078 (dw_val_equal_p): Handle symview.
5079 (add_AT_symview): New.
5080 (print_dw_val): Handle symview.
5081 (attr_checksum, attr_checksum_ordered): Likewise.
5082 (same_dw_val_p, size_of_die): Likewise.
5083 (value_format, output_die): Likewise.
5084 (add_high_low_attributes): Use add_AT_symview for entry_view.
5085 (dwarf2out_finish): Reset symview_upper_bound, clear
5088 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5091 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5092 Add strict argument and use it.
5093 (rs6000_split_multireg_move): Update for new strict argument.
5094 (mem_operand_gpr): Disallow all non-offsettable addresses.
5095 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5097 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5100 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5101 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5102 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5105 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5106 decl, use remap_type if we want to use the type.
5108 2018-03-09 Martin Sebor <msebor@redhat.com>
5110 PR tree-optimization/84526
5111 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5113 (builtin_access::generic_overlap): Be prepared to handle non-array
5116 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5118 PR rtl-optimization/84682
5119 * lra-constraints.c (process_address_1): Check is_address flag
5120 for address constraints.
5121 (process_alt_operands): Likewise.
5122 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5123 preprocess_constraints.
5124 * recog.h (preprocess_constraints): Add oploc parameter.
5126 * recog.c (preprocess_constraints): Test address_operand for
5127 CT_ADDRESS constraints.
5129 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5132 * lra-assigns.c (assign_by_spills): Return a flag of reload
5133 assignment failure. Do not process the reload assignment
5134 failures. Do not spill other reload pseudos if they has the same
5136 (lra_assign): Add a return arg. Set up from the result of
5137 assign_by_spills call.
5138 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5139 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5140 usage_insns if it is not NULL.
5141 (spill_hard_reg_in_range): New function.
5142 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5143 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5144 function prototypes.
5145 (lra_assign): Change prototype.
5146 * lra.c (lra): Add code to deal with fails by splitting hard reg
5149 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5152 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5153 Accept complain bool parameter. Only emit errors if it is true.
5154 (arm_parse_cpu_option_name): Likewise.
5155 (arm_target_thumb_only): Adjust callers of the above.
5156 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5157 prototype to take a default true bool parameter.
5158 (arm_parse_arch_option_name): Likewise.
5160 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5161 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5165 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5166 * configure: Regenerate.
5167 * configure.ac ("linker --version-script option"): New.
5168 ("linker soname option"): New.
5170 2018-03-09 Richard Biener <rguenther@suse.de>
5172 PR tree-optimization/84775
5173 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5174 immediate uses of predicate stmts and mark them modified.
5177 PR tree-optimization/84178
5178 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5180 (version_loop_for_if_conversion): Delay update_ssa call.
5181 (tree_if_conversion): Delay update_ssa until after predicate
5184 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5187 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5188 when the function accesses prior frames.
5190 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5193 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5194 gen_llsym, otherwise call maybe_gen_llsym.
5197 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5198 has ',' character inside of it.
5200 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5203 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5204 as clobbering CC_REGNUM.
5206 2018-03-08 Richard Biener <rguenther@suse.de>
5209 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5210 (follow_copies_to_constant): Do not follow SSA names registered
5213 2018-03-08 Richard Biener <rguenther@suse.de>
5215 PR tree-optimization/84178
5216 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5218 (version_loop_for_if_conversion): Delay update_ssa call.
5219 (tree_if_conversion): Delay update_ssa until after predicate
5222 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5224 PR tree-optimization/84178
5225 * tree-if-conv.c (release_bb_predicate): Remove the
5226 the assertion that the stmts have NULL use_ops.
5227 Discard the statements, asserting that they haven't
5228 yet been added to a BB.
5230 2018-03-08 Richard Biener <rguenther@suse.de>
5232 PR tree-optimization/84746
5233 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5234 (phi_translate): Pass in destination ANTIC_OUT set.
5235 (phi_translate_1): Likewise. For a simplified result lookup
5236 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5237 (phi_translate_set): Adjust.
5238 (do_pre_regular_insertion): Likewise.
5239 (do_pre_partial_partial_insertion): Likewise.
5241 2018-03-08 Martin Liska <mliska@suse.cz>
5243 PR gcov-profile/84735
5244 * doc/gcov.texi: Document usage of profile files.
5245 * gcov-io.h: Document changes in the format.
5247 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5251 * dwarf2out.c (struct dw_line_info_table): Update comments for
5253 (FORCE_RESET_NEXT_VIEW): New.
5254 (FORCE_RESETTING_VIEW_P): New.
5255 (RESETTING_VIEW_P): Check for -1 too.
5256 (ZERO_VIEW_P): Likewise.
5257 (new_line_info_table): Force-reset next view.
5258 (dwarf2out_begin_function): Likewise.
5259 (dwarf2out_source_line): Simplify zero_view_p initialization.
5260 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5261 view directly. Omit view when omitting .loc at line 0.
5263 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5265 PR tree-optimization/84740
5266 * tree-switch-conversion.c (process_switch): Call build_constructors
5267 only if info.phi_count is non-zero.
5269 PR tree-optimization/84739
5270 * tree-tailcall.c (find_tail_calls): Check call arguments against
5271 DECL_ARGUMENTS (current_function_decl) rather than
5272 DECL_ARGUMENTS (func) when checking for tail recursion.
5274 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5276 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5277 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5278 Volker Reichelt's entry and add entries for people that perform
5279 GCC fuzzy testing and report numerous bugs.
5281 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5284 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5285 readonly data in sdata, if that is disabled.
5286 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5287 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5288 -mreadonly-in-sdata option.
5290 2018-03-07 Martin Sebor <msebor@redhat.com>
5292 PR tree-optimization/84468
5293 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5294 basic block when looking for nul assignment.
5296 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5299 * except.h (output_function_exception_table): Adjust prototype.
5300 * except.c (output_function_exception_table): Remove FNNAME parameter
5301 and add SECTION parameter. Ouput one part of the table at a time.
5302 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5303 the first part of the exception table and emit unwind directives.
5304 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5305 (i386_pe_seh_cold_init): Likewise.
5306 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5307 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5308 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5309 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5310 * config/i386/winnt.c: Include except.h.
5311 (struct seh_frame_state): Add reg_offset, after_prologue and
5312 in_cold_section fields.
5313 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5314 (i386_pe_seh_cold_init): New function.
5315 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5316 to seh->in_cold_section.
5317 (seh_emit_push): Record the offset of the push.
5318 (seh_emit_save): Record the offet of the save.
5319 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5320 Test seh->after_prologue to disregard the epilogue.
5321 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5322 (i386_pe_end_cold_function): New function.
5324 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5327 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5328 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5331 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5333 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5334 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5337 * multiple_target.c: Include tree-inline.h and intl.h.
5338 (expand_target_clones): Diagnose and fail if node->definition and
5339 !tree_versionable_function_p (node->decl).
5341 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5343 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5345 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5346 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5347 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5349 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5352 * combine.c (try_combine): Use reg_or_subregno instead of handling
5353 just paradoxical SUBREGs and REGs.
5355 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5357 * config/arc/arc.c (arc_finalize_pic): Remove function.
5358 (arc_must_save_register): We use single base PIC register, remove
5359 checks to save/restore the PIC register.
5360 (arc_expand_prologue): Likewise.
5361 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5363 (arc_verify_short): Likewise.
5364 (arc_attr_type): Likewise.
5365 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5366 (walk_stores): Likewise.
5367 (arc_address_cost): Make it static.
5368 (arc_verify_short): Likewise.
5369 (branch_dest): Likewise.
5370 (arc_attr_type): Likewise.
5371 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5372 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5373 (arc_final_prescan_insn): Remove inserting the nops due to
5374 hardware hazards. It is done in reorg step.
5375 (insn_length_variant_t): Remove.
5376 (insn_length_parameters_t): Likewise.
5377 (arc_insn_length_parameters): Likewise.
5378 (arc_get_insn_variants): Likewise.
5379 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5381 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5384 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5387 PR tree-optimization/84687
5388 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5390 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5392 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5394 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5395 Rename to ppc_speculation_barrier.
5396 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5397 __builtin_ppc_speculation_barrier.
5399 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5402 * combine.c (combine_simplify_rtx): Don't try to simplify if
5403 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5406 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5408 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5409 to 32 bytes when compiling for POWER9.
5411 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5414 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5415 regparm >= 3 with no arg reg available also for calls with
5416 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5419 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5421 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5423 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5426 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5428 2018-03-05 Richard Biener <rguenther@suse.de>
5430 PR tree-optimization/84486
5431 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5432 When inserting a __builtin_assume_aligned call set the LHS
5433 SSA name alignment info accordingly.
5435 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5437 PR tree-optimization/84114
5438 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5439 Avoid reassociation of FLOAT_MODE addition.
5441 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5443 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5444 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5445 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5446 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5447 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5448 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5449 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5451 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5452 __builtin_ia32_wbinvd): New builtins.
5453 (SPECIAL_ARGS2): New.
5454 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5455 (SPECIAL_ARGS2): New.
5456 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5457 (ix86_valid_target_attribute_inner_p): Ditto.
5458 (ix86_init_mmx_sse_builtins): Add special_args2.
5459 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5460 TARGET_WBNOINVD_P): New.
5461 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5462 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5463 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5464 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5465 * config/i386/pconfigintrin.h: New file.
5466 * config/i386/wbnoinvdintrin.h: Ditto.
5467 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5468 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5470 2018-03-05 Richard Biener <rguenther@suse.de>
5472 PR tree-optimization/84670
5473 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5475 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5476 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5477 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5478 doesn't grow if all successors (recursively) were visited at least
5481 2018-03-05 Richard Biener <rguenther@suse.de>
5483 PR tree-optimization/84650
5484 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5485 if executed in the loop pipeline.
5487 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5489 * doc/configfiles.texi (Configuration Files): Move info about
5490 conditionalizing $target-protos.h to...
5491 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5492 differs from $target-protos.h.
5494 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5495 Chung-Ju Wu <jasonwucj@gmail.com>
5497 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5498 * config/nds32/nds32-multiple.md (setmemsi): Define.
5499 * config/nds32/nds32-memory-manipulation.c
5500 (nds32_gen_dup_4_byte_to_word_value): New.
5501 (emit_setmem_word_loop): New.
5502 (emit_setmem_byte_loop): New.
5503 (nds32_expand_setmem_loop): New.
5504 (nds32_expand_setmem_loop_v3m): New.
5505 (nds32_expand_setmem_unroll): New.
5506 (nds32_expand_setmem): New.
5508 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5509 Chung-Ju Wu <jasonwucj@gmail.com>
5511 * config/nds32/nds32-memory-manipulation.c
5512 (nds32_emit_load_store): New.
5513 (nds32_emit_post_inc_load_store): New.
5514 (nds32_emit_mem_move): New.
5515 (nds32_emit_mem_move_block): New.
5516 (nds32_expand_movmemsi_loop_unknown_size): New.
5517 (nds32_expand_movmemsi_loop_known_size): New.
5518 (nds32_expand_movmemsi_loop): New.
5519 (nds32_expand_movmemsi_unroll): New.
5520 (nds32_expand_movmemqi): Rename ...
5521 (nds32_expand_movmemsi): ... to this.
5522 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5523 (movmemsi): ... to this.
5524 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5525 (nds32_expand_movmemsi): ... to this.
5527 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5528 Monk Chiang <sh.chiang04@gmail.com>
5529 Chung-Ju Wu <jasonwucj@gmail.com>
5531 * config/nds32/nds32-protos.h
5532 (nds32_expand_load_multiple): New arguments.
5533 (nds32_expand_store_multiple): Ditto.
5534 (nds32_valid_multiple_load_store): Rename ...
5535 (nds32_valid_multiple_load_store_p): ... to this.
5536 * config/nds32/nds32-memory-manipulation.c
5537 (nds32_expand_load_multiple): Refine implementation.
5538 (nds32_expand_store_multiple): Ditto.
5539 * config/nds32/nds32-multiple.md
5540 (load_multiple): Update nds32_expand_load_multiple interface.
5541 (store_multiple): Update nds32_expand_store_multiple interface.
5542 * config/nds32/nds32-predicates.c
5543 (nds32_valid_multiple_load_store): Rename ...
5544 (nds32_valid_multiple_load_store_p): ... to this and refine
5546 * config/nds32/predicates.md
5547 (nds32_load_multiple_and_update_address_operation): New predicate.
5548 (nds32_store_multiple_and_update_address_operation): New predicate.
5550 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5551 Chung-Ju Wu <jasonwucj@gmail.com>
5553 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5554 (combo): New attribute.
5555 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5557 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5559 * config/nds32/nds32.opt: Change -mcmodel= default value.
5561 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5562 Monk Chiang <sh.chiang04@gmail.com>
5563 Chung-Ju Wu <jasonwucj@gmail.com>
5565 * config/nds32/constants.md (unspec_element): New enum.
5566 * config/nds32/constraints.md (Umw): New constraint.
5567 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5568 * config/nds32/nds32-intrinsic.md: Likewise.
5569 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5570 (nds32_valid_smw_lwm_base_p): New.
5571 (nds32_output_smw_single_word): New.
5572 (nds32_output_lmw_single_word): New.
5573 (nds32_expand_unaligned_load): New.
5574 (nds32_expand_unaligned_store): New.
5575 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5576 (nds32_output_smw_single_word): Declare.
5577 (nds32_output_lmw_single_word): Declare.
5578 (nds32_expand_unaligned_load): Declare.
5579 (nds32_expand_unaligned_store): Declare.
5580 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5581 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5582 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5583 NDS32_BUILTIN_UASTORE_DW.
5584 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5587 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5588 Kito Cheng <kito.cheng@gmail.com>
5589 Chung-Ju Wu <jasonwucj@gmail.com>
5591 * config/nds32/nds32-intrinsic.c
5592 (nds32_expand_builtin_null_ftype_reg): Delete.
5593 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5594 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5595 (nds32_read_argument): New.
5596 (nds32_legitimize_target): Ditto.
5597 (nds32_legitimize_argument): Ditto.
5598 (nds32_check_constant_argument): Ditto.
5599 (nds32_expand_unop_builtin): Ditto.
5600 (nds32_expand_unopimm_builtin): Ditto.
5601 (nds32_expand_binop_builtin): Ditto.
5602 (nds32_builtin_decl_impl): Ditto.
5603 (builtin_description): Ditto.
5604 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5605 (nds32_init_builtins_impl): Ditto.
5606 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5607 (nds32_builtin_decl): New.
5608 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5609 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5611 2018-03-02 Jeff Law <law@redhat.com>
5613 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5614 (redundant_insn, fill_simple_delay_slots): Likewise.
5615 (fill_slots_from_thread): Likewise.
5616 * resource.c (mark_referenced_resources): Likewise.
5617 (mark_set_resources, find_dead_or_set_registers): Likewise.
5619 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5621 * substring-locations.h (format_warning_va): Formatting fix for
5623 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5625 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5626 * substring-locations.c: Include intl.h.
5627 (format_warning_va): Turned into small wrapper around
5628 format_warning_n_va, renamed to ...
5629 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5630 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5632 (format_warning_at_substring_n): New function.
5633 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5634 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5635 format_warning_at_substring with just a shorter name instead of
5636 const function pointer.
5637 (fmtwarn_n): New function.
5638 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5639 appropriate, get rid of all the fmtstr temporaries, move conditionals
5640 with G_() wrapped string literals directly into fmtwarn arguments,
5641 cast dir.len to (int), formatting fixes.
5643 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5645 * doc/invoke.texi: Remove "Cilk Plus" references.
5647 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5648 Richard Biener <rguenther@suse.de>
5651 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5652 for error or warning attributes if CALL_FROM_THUNK_P is set.
5655 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5658 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5659 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5662 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5663 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5665 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5666 instead of -1U in last predictors element's probability member.
5668 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5671 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5672 arguments if they are comparable.
5674 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5676 PR tree-optimization/84634
5677 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5678 masks and masked_loop_p with a single loop_masks, making sure it's
5679 null for bb vectorization.
5681 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5683 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5684 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5685 loop->force_vectorize to check whether there is no alias.
5687 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5690 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5692 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5694 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5695 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5696 instead of a loop around prev_real_insn.
5697 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5701 * config/i386/i386.c (ix86_print_operand): Use conditional
5702 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5705 2018-03-02 Richard Biener <rguenther@suse.de>
5707 PR tree-optimization/84427
5708 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5709 (bitmap_set_subtract_values): Rewrite to handle multiple
5712 (prune_clobbered_mems): Likewise.
5713 (phi_translate): Take edge instead of pred/phiblock.
5714 (phi_translate_1): Likewise.
5715 (phi_translate_set): Likewise. Insert all translated
5716 exprs for a value into the set, keeping possibly multiple
5717 expressions per value.
5718 (compute_antic_aux): Adjust for phi_translate changes.
5719 When intersecting union the expressions and prune those
5720 not in the final value set, keeping possibly multiple
5721 expressions per value. Do not use value-insertion
5722 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5723 all expressions. Add verification that the value-sets
5724 only shrink during iteration.
5725 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5726 (do_pre_regular_insertion): Likewise.
5727 (do_pre_partial_partial_insertion): Likewise.
5729 2018-03-02 Richard Biener <rguenther@suse.de>
5732 * config/darwin.c (saved_debug_info_level): New static global.
5733 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5734 (darwin_asm_lto_end): Restore debug info generation settings.
5736 2018-03-01 Martin Liska <mliska@suse.cz>
5739 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5742 2018-03-01 Richard Biener <rguenther@suse.de>
5745 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5746 type annotation with LTO when debug was disabled at compile-time.
5748 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5750 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5752 (mips_final_postscan_insn): Likewise.
5754 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5756 PR rtl-optimization/84528
5757 * alias.c (init_alias_target): Add commentary.
5758 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5759 a unique base value if the frame pointer is not eliminated
5760 to the stack pointer.
5762 2018-03-01 Tom de Vries <tom@codesourcery.com>
5764 PR rtl-optimization/83327
5765 * lra-int.h (hard_regs_spilled_into): Declare.
5766 * lra.c (hard_regs_spilled_into): Define.
5767 (init_reg_info): Init hard_regs_spilled_into.
5768 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5769 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5770 (process_bb_lives): Handle hard_regs_spilled_into.
5771 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5772 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5774 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5776 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5777 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5778 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5779 * config/rs6000/aix72.h: New file.
5781 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5783 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5784 instead of warning_at with conditional singular and plural messages
5788 * stor-layout.c (update_alignment_for_field): For
5789 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5790 && !DECL_PACKED (field), do the alignment update, just use
5791 only desired_align instead of MAX (type_align, desired_align)
5793 (place_field): Don't do known_align < desired_align handling
5794 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5795 is non-NULL, instead do it after rli->prev_field handling and
5796 only if not within a bitfield word. For DECL_PACKED (field)
5797 use type_align of BITS_PER_UNIT.
5799 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5801 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5802 superfluous parentheses and trailing spaces.
5804 2018-02-28 Richard Biener <rguenther@suse.de>
5806 PR tree-optimization/84584
5807 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5808 SCoPs with fake exit edge.
5810 2018-02-28 Martin Liska <mliska@suse.cz>
5813 * timevar.c (timer::print): Fix format to properly print 100%
5816 2018-02-28 Richard Biener <rguenther@suse.de>
5819 * genmatch.c (capture_info::walk_match): Do not mark
5820 captured expressions without operands as expr_p given
5821 they act more like predicates and should be subject to
5822 "lost tail" side-effect preserving.
5824 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5826 PR rtl-optimization/81611
5827 * auto-inc-dec.c (attempt_change): Move dead note from
5828 mem_insn if it's the next use of regno
5829 (find_address): Take address use of reg holding
5830 non-incremented value. Add parm to limit search to the named
5832 (merge_in_block): Attempt to use a mem insn that is the next
5833 use of the original regno.
5835 2018-02-27 Martin Sebor <msebor@redhat.com>
5838 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5839 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5841 2018-02-27 Martin Sebor <msebor@redhat.com>
5843 PR translation/84207
5844 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5845 n argument to unsigned HOST_WIDE_INT.
5846 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5847 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5848 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5849 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5851 2018-02-27 Richard Biener <rguenther@suse.de>
5853 PR tree-optimization/84512
5854 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5855 Do not use the estimate returned from record_stmt_cost for
5856 the scalar iteration cost but sum properly using add_stmt_cost.
5858 2018-02-27 Richard Biener <rguenther@suse.de>
5860 PR tree-optimization/84466
5861 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5862 Adjust last change to less strictly validate use operands.
5864 2018-02-27 Martin Liska <mliska@suse.cz>
5866 PR gcov-profile/84548
5867 * gcov.c (process_file): Allow partial overlap and consider it
5868 also as group functions.
5869 (output_lines): Properly calculate range of lines for a group.
5871 2018-02-27 Martin Liska <mliska@suse.cz>
5873 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5874 'ggc' suffixes. Change first column width.
5875 (timer::print): Fix formatting of the column.
5877 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5879 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5880 preserve inline entry blocks for the sake of debug inline
5881 entry point markers alone.
5882 (remove_unused_locals): Suggest in comments a better place to
5883 force the preservation of inline entry blocks that are
5884 otherwise unused, but do not preserve them.
5886 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5888 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5890 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5893 * config/i386/constraints.md (Bs): Replace
5894 ix86_indirect_branch_register with
5895 TARGET_INDIRECT_BRANCH_REGISTER.
5897 * config/i386/i386.md (indirect_jump): Likewise.
5898 (tablejump): Likewise.
5899 (*sibcall_memory): Likewise.
5900 (*sibcall_value_memory): Likewise.
5901 Peepholes of indirect call and jump via memory: Likewise.
5902 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5903 (*sibcall_value_GOT_32): Likewise.
5904 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5905 (GOT_memory_operand): Likewise.
5906 (call_insn_operand): Likewise.
5907 (sibcall_insn_operand): Likewise.
5908 (GOT32_symbol_operand): Likewise.
5909 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5911 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5913 PR rtl-optimization/83496
5914 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5915 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5916 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5917 redundant insn, if any.
5918 (relax_delay_slots): Likewise.
5919 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5921 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5923 PR tree-optimization/83965
5924 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5925 that grouped statements are part of a reduction chain. Return
5926 true if the statement is not marked as a reduction itself but
5928 (vect_recog_dot_prod_pattern): Don't check whether the statement
5929 is part of a group here.
5930 (vect_recog_sad_pattern): Likewise.
5931 (vect_recog_widen_sum_pattern): Likewise.
5933 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5936 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5938 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5941 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5943 (ix86_output_indirect_function_return): New prototype.
5944 (ix86_split_simple_return_pop_internal): Likewise.
5945 * config/i386/i386.c (indirect_return_via_cx): New.
5946 (indirect_return_via_cx_bnd): Likewise.
5947 (indirect_thunk_name): Handle return va CX_REG.
5948 (output_indirect_thunk_function): Create alias for
5949 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5950 (ix86_output_indirect_jmp): Remove the bool argument.
5951 (ix86_output_indirect_function_return): New function.
5952 (ix86_split_simple_return_pop_internal): Likewise.
5953 * config/i386/i386.md (*indirect_jump): Don't pass false
5954 to ix86_output_indirect_jmp.
5955 (*tablejump_1): Likewise.
5956 (simple_return_pop_internal): Change it to define_insn_and_split.
5957 Call ix86_split_simple_return_pop_internal to split it for
5959 (simple_return_indirect_internal): Call
5960 ix86_output_indirect_function_return instead of
5961 ix86_output_indirect_jmp.
5963 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5966 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5967 memset and value initialization afterwards.
5969 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5971 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5973 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5976 * common/config/aarch64/aarch64-common.c
5977 (aarch_option_optimization_table[]): Switch
5978 off fomit-frame-pointer
5980 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5981 Chung-Ju Wu <jasonwucj@gmail.com>
5983 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5985 (store_multiple): Ditto.
5987 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5989 * config.gcc: Add --with-cpu support for nds32 target.
5990 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5991 * config/nds32/nds32.opt: Add -mcpu= option.
5993 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5995 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5996 isel=yes): Warn for these deprecated options.
5998 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6000 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6001 ISA_2_5_MASKS_EMBEDDED.
6003 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6005 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6006 p->max as pointers rather than using iterative_hash_expr.
6008 2018-02-23 Carl Love <cel@us.ibm.com>
6010 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6011 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6013 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6014 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6015 P8V_BUILTIN_VEC_VUNSIGNED2.
6017 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6020 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6021 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6023 * lra-constraints.c (curr_insn_transform): Set up
6024 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6026 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6027 * lra-eliminations.c (spill_pseudos): Ditto.
6028 (process_insn_for_elimination): Ditto.
6029 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6030 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6031 LRA_NON_CLOBBERED_ALT.
6033 2018-02-22 Martin Sebor <msebor@redhat.com>
6035 PR tree-optimization/84480
6036 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6037 to maybe_diag_stxncpy_trunc. Call it.
6038 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6039 from gimple_fold_builtin_strcpy. Print inlining stack.
6040 (handle_builtin_stxncpy): Print inlining stack.
6041 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6043 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6046 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6047 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6048 and -fcheck-pointer-bounds are used together.
6049 (indirect_thunk_prefix): New enum.
6050 (indirect_thunk_need_prefix): New function.
6051 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6052 "_nt" instead of "_bnd" for NOTRACK prefix.
6053 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6054 (output_indirect_thunk_function): Likewise.
6056 (ix86_code_end): Update output_indirect_thunk_function calls.
6057 (ix86_output_indirect_branch_via_reg): Replace
6058 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6059 (ix86_output_indirect_branch_via_push): Likewise.
6060 (ix86_output_function_return): Likewise.
6061 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6062 incompatible with -fcf-protection=branch and
6063 -fcheck-pointer-bounds.
6065 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6068 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6069 Change gcc_assert call to output_operand_lossage.
6071 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6073 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6075 2018-02-22 DJ Delorie <dj@redhat.com>
6076 Sebastian Perta <sebastian.perta@renesas.com>
6077 Oleg Endo <olegendo@gcc.gnu.org>
6079 * config/rx/rx.c (rx_rtx_costs): New function.
6080 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6082 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6084 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6086 2018-02-22 Martin Liska <mliska@suse.cz>
6089 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6090 Add "native" as a possible value.
6092 2018-02-22 Martin Liska <mliska@suse.cz>
6095 * config/i386/i386.c (ix86_option_override_internal):
6096 Add "native" as a possible value for -march and -mtune.
6098 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6101 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6102 to all type variants.
6104 PR tree-optimization/84503
6105 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6106 width as info->bitpos + info->bitsize - start.
6107 (merged_store_group::merge_overlapping): Simplify width computation.
6108 (check_no_overlap): New function.
6109 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6110 start + width and last_order of the group, fail if check_no_overlap
6112 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6113 to group if check_no_overlap fails.
6115 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6117 * config/rs6000/altivec.md: Delete contraint arguments to
6118 define_expand, define_split, and define_peephole2, and in
6119 define_insn_and_split if always unused.
6120 * config/rs6000/darwin.md: Ditto.
6121 * config/rs6000/dfp.md: Ditto.
6122 * config/rs6000/rs6000.md: Ditto.
6123 * config/rs6000/sync.md: Ditto.
6124 * config/rs6000/vector.md: Ditto.
6125 * config/rs6000/vsx.md: Ditto.
6127 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6129 * config/rs6000/altivec.md: Write output control strings as braced
6130 blocks instead of double-quoted strings.
6131 * config/rs6000/darwin.md: Ditto.
6132 * config/rs6000/rs6000.md: Ditto.
6133 * config/rs6000/vector.md: Ditto.
6134 * config/rs6000/vsx.md: Ditto.
6136 2018-02-21 Jason Merrill <jason@redhat.com>
6138 PR c++/84314 - ICE with templates and fastcall attribute.
6139 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6141 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6143 * ipa-cp.c (determine_versionability): Fix comment typos.
6145 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6148 * ipa-cp.c (determine_versionability): Do not version functions caling
6151 2018-02-21 Martin Liska <mliska@suse.cz>
6154 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6155 Add "native" as a possible value.
6156 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6157 the macro when native cpu detection is available.
6159 2018-02-21 Martin Liska <mliska@suse.cz>
6162 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6163 Add "native" as a possible value.
6164 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6165 when native cpu detection is available.
6167 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6168 Martin Sebor <msebor@redhat.com>
6170 PR tree-optimization/84478
6171 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6173 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6174 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6175 support which is conservatively correct, for 2 only stay conservative
6176 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6177 argument to the 2 argument get_range_strlen, adjust 6 arg
6178 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6180 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6181 (gimple_fold_builtin_strlen): Pass true as last argument to
6184 2018-02-20 Martin Sebor <msebor@redhat.com>
6187 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6188 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6189 (builtin_memref::builtin_memref): Factor out parts into
6190 set_base_and_offset and call it.
6192 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6195 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6196 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6197 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6198 search at the associated MODE_INT.
6200 2018-02-20 Jeff Law <law@redhat.com>
6203 PR tree-optimization/81592
6205 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6206 for range data rather than using global data.
6207 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6208 range data rather than using global data.
6209 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6210 pass it to children as needed.
6211 (struct directive::fmtresult): Similarly.
6212 (struct directive::set_width): Similarly.
6213 (struct directive::set_precision): Similarly.
6214 (format_integer, format_directive, parse_directive): Similarly.
6215 (format_none): Accept unnamed vr_values parameter.
6216 (format_percent, format_floating, format_character): Similarly.
6217 (format_string, format_plain): Similarly.
6218 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6219 the EVRP range analyzer for range data rather than using global data.
6220 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6221 gimple-ssa-evrp-analyze.h
6222 (class sprintf_dom_walker): Add after_dom_children member function.
6223 Add evrp_range_analyzer member.
6224 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6225 range analyzer as needed.
6226 (sprintf_dom_walker::after_dom_children): New member function.
6227 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6229 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6230 (evrp_range_analyzer::pop_to_marker): Likewise.
6232 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6234 PR tree-optimization/84419
6235 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6236 with the required type if its current type is compatible but
6239 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6242 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6243 after vectorization.
6245 2018-02-20 Martin Liska <mliska@suse.cz>
6248 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6249 possible values if we don't have a hint.
6251 2018-02-20 Martin Liska <mliska@suse.cz>
6255 * final.c (shorten_branches): Build align_tab array with one
6257 * opts.c (finish_options): Add alignment option limit check.
6258 (MAX_CODE_ALIGN): Likewise.
6259 (MAX_CODE_ALIGN_VALUE): Likewise.
6260 * doc/invoke.texi: Document maximum allowed option value for
6261 all -falign-* options.
6263 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6266 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6267 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6268 * var-tracking.c (emit_note_insn_var_location): Remove all references
6269 to NOTE_INSN_CALL_ARG_LOCATION.
6270 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6271 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6272 Use copy_rtx_if_shared.
6273 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6274 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6275 (dwarf2out_var_location): Remove handling of
6276 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6278 * final.c (final_scan_insn): Remove all references to
6279 NOTE_INSN_CALL_ARG_LOCATION.
6280 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6281 before dumping final insns.
6282 * except.c (emit_note_eh_region_end): Remove all references to
6283 NOTE_INSN_CALL_ARG_LOCATION.
6284 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6285 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6286 * config/arc/arc.c (hwloop_optimize): Likewise.
6287 * config/arm/arm.c (create_fix_barrier): Likewise.
6288 * config/s390/s390.c (s390_chunkify_start): Likewise.
6289 * config/sh/sh.c (find_barrier): Likewise.
6290 * config/i386/i386.c (rest_of_insert_endbranch,
6291 ix86_seh_fixup_eh_fallthru): Likewise.
6292 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6293 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6294 * config/frv/frv.c (frv_function_prologue): Likewise.
6295 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6297 (note_outside_basic_block_p): Remove all references to
6298 NOTE_INSN_CALL_ARG_LOCATION.
6299 * gengtype.c (adjust_field_rtx_def): Likewise.
6300 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6302 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6303 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6306 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6309 PR tree-optimization/84452
6310 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6311 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6314 2018-02-19 Martin Liska <mliska@suse.cz>
6317 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6319 2018-02-19 Martin Liska <mliska@suse.cz>
6320 Richard Sandiford <richard.sandiford@linaro.org>
6322 PR tree-optimization/82491
6323 * gimple-fold.c (get_base_constructor): Make earlier bail out
6326 2018-02-19 Carl Love <cel@us.ibm.com>
6328 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6329 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6331 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6332 P8V_BUILTIN_VEC_NEG.
6334 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6336 * config/rl78/rl78.md (movdf): New define expand.
6338 2018-02-19 Martin Liska <mliska@suse.cz>
6341 * doc/invoke.texi: Fix typo.
6342 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6344 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6346 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6347 handle rs6000_single_float and rs6000_double_float specially for
6350 2018-02-16 Jeff Law <law@redhat.com>
6352 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6355 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6358 * ipa-inline.c (inline_small_functions): Fix a typo.
6360 2018-02-16 Nathan Sidwell <nathan@acm.org>
6362 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6364 2018-02-16 Carl Love <cel@us.ibm.com>
6366 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6367 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6368 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6369 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6370 expansion to P8V_BUILTIN_VEC_FLOAT2.
6372 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6374 PR rtl-optimization/70023
6375 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6376 src_regno into account.
6378 2018-02-16 Carl Love <cel@us.ibm.com>
6380 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6381 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6382 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6383 * config/rs6000/rs6000.c: Remove case statements for
6384 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6385 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6386 and P9V_BUILTIN_VEC_VINSERT4B.
6387 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6388 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6389 * config/rs6000/vsx.md:
6390 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6393 2018-02-16 Carl Love <cel@us.ibm.com>
6395 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6397 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6399 * config/rs6000/rs6000-c.c: Add the definitions for
6400 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6401 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6402 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6403 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6404 definition for insert4b and define insn *insert3b_internal.
6405 * doc/extend.texi: Add documentation for vec_extract4b.
6407 2018-02-16 Nathan Sidwell <nathan@acm.org>
6409 * doc/extend.texi (Backwards Compatibility): Mention friend
6410 injection. Note for-scope is deprecated.
6411 * doc/invoke.texi (-ffriend-injection): Deprecate.
6413 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6415 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6416 that moved to I2, also allow destinations that are a paradoxical
6417 subreg (instead of a normal reg).
6419 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6422 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6425 2018-02-16 Richard Biener <rguenther@suse.de>
6427 PR tree-optimization/84037
6428 PR tree-optimization/84016
6430 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6431 Adjust vec_construct for the fact we need additional higher latency
6432 128bit inserts for AVX256 and AVX512 vector builds.
6433 (ix86_add_stmt_cost): Scale vector construction cost for
6436 2018-02-16 Richard Biener <rguenther@suse.de>
6438 PR tree-optimization/84417
6439 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6440 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6441 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6443 2018-02-16 Martin Liska <mliska@suse.cz>
6446 * internal-fn.def (ASAN_CHECK): Set proper flags.
6447 (ASAN_MARK): Likewise.
6449 2018-02-16 Julia Koval <julia.koval@intel.com>
6451 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6452 from PTA_CANNONLAKE.
6454 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6457 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6458 Use ++iter rather than iter++ for std::list iterators.
6459 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6460 defer deleting them until all nodes in the forest are processed. Do
6461 free even leaf nodes. Change to_process into auto_vec.
6464 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6465 * vec.h (vec_default_construct): Use memset instead of placement new
6466 if BROKEN_VALUE_INITIALIZATION is defined.
6467 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6468 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6471 PR rtl-optimization/83723
6472 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6473 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6474 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6476 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6478 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6480 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6482 PR rtl-optimization/81443
6483 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6484 from inner REGs to paradoxical SUBREGs.
6486 2018-02-16 Richard Biener <rguenther@suse.de>
6488 PR tree-optimization/84399
6489 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6490 For operands we can analyze at their definition make sure we can
6491 analyze them at each use as well.
6493 2018-02-16 Richard Biener <rguenther@suse.de>
6495 PR tree-optimization/84190
6496 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6497 volatile accesses if the decl isn't volatile.
6499 2018-02-15 Jason Merrill <jason@redhat.com>
6501 PR c++/84314 - ICE with templates and fastcall attribute.
6502 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6503 TYPE_CANONICAL on an existing type.
6505 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6507 PR tree-optimization/84383
6508 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6509 dstoff nor call operand_equal_p if dstbase is NULL.
6511 PR tree-optimization/84334
6512 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6513 also a CONSTANT_CLASS_P, punt.
6515 2018-02-14 Jim Wilson <jimw@sifive.com>
6517 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6518 first SMALL_OPERAND check. New local min_second_step. Move assert
6519 to where locals are set. Add TARGET_RVC support.
6520 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6522 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6524 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6526 2018-02-14 Martin Sebor <msebor@redhat.com>
6528 PR tree-optimization/83698
6529 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6530 arrays constrain the offset range to their bounds.
6531 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6532 (builtin_access::overlap): Avoid setting the size of overlap if it's
6534 (maybe_diag_overlap): Also consider arrays when deciding what values
6535 of offsets to include in diagnostics.
6537 2018-02-14 Martin Sebor <msebor@redhat.com>
6540 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6541 that correspond to the kind of a declaration.
6543 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6546 * config/pa/pa.md: Load address of PIC label using the linkage table
6547 if the label is nonlocal.
6549 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6551 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6552 warning message if user requests -maltivec=be.
6553 * doc/invoke.texi: Document deprecation of -maltivec=be.
6555 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6558 * config/rs6000/rs6000-c.c: Update definitions for
6559 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6560 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6562 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6565 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6566 add _get_ssp intrinsics. Remove argument from
6567 __builtin_ia32_rdssp[d|q].
6568 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6569 * config/i386/i386-builtin.def: Remove argument from
6570 __builtin_ia32_rdssp[d|q].
6571 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6572 ix86_expand_special_args_builtin for _rdssp[d|q].
6573 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6574 Clear register before usage.
6575 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6576 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6578 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6580 PR tree-optimization/84357
6581 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6582 operand 1 of an ARRAY_REF too.
6584 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6587 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6588 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6590 (set_of_reg): New struct.
6591 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6592 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6593 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6595 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6596 Split into bitclr, bitset, bitinvert patterns if appropriate.
6597 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6598 use rx_fuse_in_memory_bitop.
6599 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6600 to named insn, correct maximum insn length.
6602 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6605 * machmode.def: Define a complex mode for PARTIAL_INT.
6606 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6608 * doc/rtl.texi: Document CSPImode.
6609 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6611 (msp430_hard_regno_nregs_with_padding): Likewise.
6613 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6616 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6618 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6620 PR rtl-optimization/84169
6621 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6622 we generated a parallel as new i3 and we split that to new i2 and i3
6623 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6624 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6625 those to i2, not i1. Partially rewrite this scan code.
6627 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6630 * stor-layout.c (place_field): For variable length fields, adjust
6631 offset_align afterwards not just based on the field's alignment,
6632 but also on the size.
6635 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6636 of exps and logs in the use_exp2 case.
6638 2018-02-13 Jeff Law <law@redhat.com>
6640 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6643 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6646 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6650 * common.opt (gas-loc-support, gas-locview-support): New.
6651 (ginline-points, ginternal-reset-location-views): New.
6652 * doc/invoke.texi: Document them. Use @itemx where intended.
6653 (gvariable-location-views): Adjust.
6654 * target.def (reset_location_view): New.
6655 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6656 (TARGET_RESET_LOCATION_VIEW): New.
6657 * doc/tm.texi: Rebuilt.
6658 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6659 (dwarf2out_default_as_locview_support): New.
6660 (output_asm_line_debug_info): Use option variables.
6661 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6662 (output_loc_list): Likewise.
6663 (add_high_low_attributes): Check option variables.
6664 Don't output entry view attribute in strict mode.
6665 (gen_inlined_subroutine_die): Check option variables.
6666 (dwarf2out_inline_entry): Likewise.
6667 (init_sections_and_labels): Likewise.
6668 (dwarf2out_early_finish): Likewise.
6669 (maybe_reset_location_view): New, from...
6670 (dwarf2out_var_location): ... here. Call it.
6671 * debug.h (dwarf2out_default_as_loc_support): Declare.
6672 (dwarf2out_default_as_locview_support): Declare.
6673 * hooks.c (hook_int_rtx_insn_0): New.
6674 * hooks.h (hook_int_rtx_insn_0): Declare.
6675 * toplev.c (process_options): Take -gas-loc-support and
6676 -gas-locview-support from dwarf2out. Enable
6677 -gvariable-location-views by default only with locview
6678 assembler support. Enable -ginternal-reset-location-views by
6679 default only if the target defines the corresponding hook.
6680 Enable -ginline-points by default if location views are
6681 enabled; force it disabled if statement frontiers are
6683 * tree-inline.c (expand_call_inline): Check option variables.
6684 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6686 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6688 PR tree-optimization/84321
6689 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6690 handling. Also check whether the anti-range contains any values
6691 that satisfy the mask; switch to a VR_RANGE if not.
6693 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6696 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6698 2018-02-13 Martin Jambor <mjambor@suse.cz>
6701 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6702 of call statements, also set location of a load to a temporary.
6704 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6706 * config/rl78/rl78.c (add_vector_labels): New function.
6707 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6708 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6709 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6710 which checks that no arguments are passed.
6711 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6712 * doc/extend.texi: Documentation for the new attribute.
6714 2018-02-13 Andreas Schwab <schwab@suse.de>
6716 * config/riscv/linux.h (CPP_SPEC): Define.
6718 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6721 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6722 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6723 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6724 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6725 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6726 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6727 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6728 temporarily for AES and PCLMUL builtins.
6730 PR tree-optimization/84339
6731 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6732 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6736 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6737 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6738 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6740 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6742 * omp-simd-clone.h: New file.
6743 * omp-simd-clone.c: Include omp-simd-clone.h.
6744 (expand_simd_clones): No longer static.
6745 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6746 cgraph.h and omp-simd-clone.h.
6747 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6748 (vect_recog_widen_shift_pattern): Formatting fix.
6749 (vect_pattern_recog_1): Don't check optab for calls.
6752 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6753 operands[2] into a REG before using gen_lowpart on it.
6755 2018-02-12 Jeff Law <law@redhat.com>
6758 * config/sh/sh.c (find_barrier): Consider a sibling call
6761 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6762 successfully back substituting a reg.
6764 2018-02-12 Richard Biener <rguenther@suse.de>
6766 PR tree-optimization/84037
6767 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6768 parameter, move visited init to caller.
6769 (vect_slp_analyze_operations): Separate cost from validity
6770 check, initialize visited once for all instances.
6771 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6773 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6775 (vect_model_promotion_demotion_cost): Likewise.
6776 (vectorizable_bswap): Guard cost modeling with !slp_node
6777 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6779 (vectorizable_call): Likewise.
6780 (vectorizable_conversion): Likewise.
6781 (vectorizable_assignment): Likewise.
6782 (vectorizable_shift): Likewise.
6783 (vectorizable_operation): Likewise.
6784 (vectorizable_store): Likewise.
6785 (vectorizable_load): Likewise.
6786 (vectorizable_condition): Likewise.
6787 (vectorizable_comparison): Likewise.
6789 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6792 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6793 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6794 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6796 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6799 * params.def (inline-min-speedup): Increase from 8 to 15.
6800 (max-inline-insns-auto): Decrease from 40 to 30.
6801 * ipa-split.c (consider_split): Add some buffer for function to
6802 be considered inlining candidate.
6803 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6806 2018-02-12 Richard Biener <rguenther@suse.de>
6808 PR tree-optimization/84037
6809 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6810 matched stmts if we cannot swap the non-matched ones.
6812 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6814 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6815 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6816 _mm_maskz_scalef_round_ss): New intrinsics.
6817 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6818 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6819 __builtin_ia32_scalefss_round): Remove.
6820 (__builtin_ia32_scalefsd_mask_round,
6821 __builtin_ia32_scalefss_mask_round): New intrinsics.
6822 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6823 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6824 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6825 "<round_constraint>")): Changed to ...
6826 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6827 "<round_scalar_constraint>")): ... this.
6828 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6829 %0, %1, %2<round_op3>}"): Changed to ...
6830 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6831 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6832 %2<round_scalar_mask_op3>}"): ... this.
6833 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6835 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6837 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6838 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6839 (_mm_maskz_sqrt_round_ss): New intrinsics.
6840 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6841 (__builtin_ia32_sqrtsd_mask_round)
6842 (__builtin_ia32_sqrtss_mask_round): New builtins.
6843 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6844 (__builtin_ia32_sqrtss_round): Remove.
6845 (__builtin_ia32_sqrtsd_mask_round)
6846 (__builtin_ia32_sqrtss_mask_round): New builtins.
6847 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6848 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6849 ((match_operand:VF_128 1 "vector_operand"
6850 "xBm,<round_constraint>")): Changed to ...
6851 ((match_operand:VF_128 1 "vector_operand"
6852 "xBm,<round_scalar_constraint>")): ... this.
6853 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6854 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6855 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6856 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6857 %<iptr>1<round_scalar_mask_op3>}): ... this.
6858 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6859 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6861 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6864 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6865 Cast vec_cmpeq result to correct type.
6866 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6867 Cast vec_cmpgt result to correct type.
6869 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6871 * final.c (final_scan_insn_1): Renamed from...
6872 (final_scan_insn): ... this. New wrapper, to recover
6873 seen from the outermost call in recursive ones.
6874 * config/sparc/sparc.c (output_return): Drop seen from call.
6875 (output_sibcall): Likewise.
6876 * config/visium/visium.c (output_branch): Likewise.
6878 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6880 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6883 2018-02-10 Alan Modra <amodra@gmail.com>
6886 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6887 Specify LR as an input.
6889 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6892 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6893 remove_member_access_dummy_vars): New functions.
6894 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6895 lower_omp_1, execute_lower_omp): Use them.
6897 PR rtl-optimization/84308
6898 * shrink-wrap.c (spread_components): Release todo vector.
6900 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6902 PR rtl-optimization/57193
6903 * ira-color.c (struct allocno_color_data): Add member
6904 conflict_allocno_hard_prefs.
6905 (update_conflict_allocno_hard_prefs): New.
6906 (bucket_allocno_compare_func): Add a preference based on
6907 conflict_allocno_hard_prefs.
6908 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6909 (color_allocnos): Remove a dead code. Initiate
6910 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6912 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6915 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6916 constraint from =wa to wa. Avoid a subreg on the output operand,
6917 instead use a pseudo and subreg it in a move.
6918 (p9_xxbrd_<mode>): Changed to ...
6919 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6920 (p9_xxbrd_v2df): New expander.
6921 (p9_xxbrw_<mode>): Changed to ...
6922 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6923 (p9_xxbrw_v4sf): New expander.
6925 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6927 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6929 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6932 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6933 multiply in 32-bit mode.
6934 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6935 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6938 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6940 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6941 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6942 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6943 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6945 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6948 * dwarf2out.c (is_trivial_indirect_ref): New function.
6949 (dwarf2out_late_global_decl): Do not generate a location
6950 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6951 and that are not defined in the current unit.
6953 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6955 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6956 instead of a libcall for UNORDERED.
6958 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6961 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6962 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6964 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6967 * config/s390/s390.c (s390_set_current_function): Invoke
6968 s390_indirect_branch_settings also if fndecl didn't change.
6970 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6972 * config/rs6000/rs6000.md (blockage): Set length to zero.
6974 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6976 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6978 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6981 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6982 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6986 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6987 PARALLEL incoming that failed vt_get_decl_and_offset check.
6990 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6991 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6993 (get_variable_section): For decls in named .bss* sections pass true as
6994 second argument to bss_initializer_p.
6996 2018-02-09 Marek Polacek <polacek@redhat.com>
6997 Jakub Jelinek <jakub@redhat.com>
7000 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7001 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7002 Sync some changes from cxx_fold_indirect_ref.
7004 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7006 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7008 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7009 (BLOCK_INLINE_ENTRY_LABEL): New.
7010 (dwarf2out_var_location): Disregard inline entry markers.
7011 (inline_entry_data): New struct.
7012 (inline_entry_data_hasher): New hashtable type.
7013 (inline_entry_data_hasher::hash): New.
7014 (inline_entry_data_hasher::equal): New.
7015 (inline_entry_data_table): New variable.
7016 (add_high_low_attributes): Add DW_AT_entry_pc and
7017 DW_AT_GNU_entry_view attributes if a pending entry is found
7018 in inline_entry_data_table. Add old entry_pc attribute only
7019 if debug nonbinding markers are disabled.
7020 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7021 markers are enabled.
7022 (block_within_block_p, dwarf2out_inline_entry): New.
7023 (dwarf2out_finish): Check that no entries remained in
7024 inline_entry_data_table.
7025 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7026 (final_scan_insn, notice_source_line): Likewise.
7027 (rest_of_clean_state): Skip inline entry markers.
7028 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7030 * gimple.c (gimple_build_debug_inline_entry): New.
7031 * gimple.h (enum gimple_debug_subcode): Add
7032 GIMPLE_DEBUG_INLINE_ENTRY.
7033 (gimple_build_debug_inline_entry): Declare.
7034 (gimple_debug_inline_entry_p): New.
7035 (gimple_debug_nonbind_marker_p): Adjust.
7036 * insn-notes.def (INLINE_ENTRY): New.
7037 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7038 inline entry marker notes.
7039 (print_insn): Likewise.
7040 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7041 (INSN_DEBUG_MARKER_KIND): Likewise.
7042 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7043 * tree-inline.c (expand_call_inline): Build and insert
7044 debug_inline_entry stmt.
7045 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7046 inline entry blocks early, if nonbind markers are enabled.
7047 (dump_scope_block): Dump fragment info.
7048 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7049 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7050 (gimple_build_debug_inline_entry): New.
7051 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7052 Enable/disable inline entry points too.
7053 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7054 (DEBUG_INSN): Describe inline entry markers.
7056 * common.opt (gvariable-location-views): New.
7057 (gvariable-location-views=incompat5): New.
7058 * config.in: Rebuilt.
7059 * configure: Rebuilt.
7060 * configure.ac: Test assembler for view support.
7061 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7062 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7063 * dwarf2out.c (var_loc_view): New typedef.
7064 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7065 (dwarf2out_locviews_in_attribute): New.
7066 (dwarf2out_locviews_in_loclist): New.
7067 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7068 (enum dw_line_info_opcode): Add LI_adv_address.
7069 (struct dw_line_info_table): Add view.
7070 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7071 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7072 (zero_view_p): New variable.
7073 (ZERO_VIEW_P): New macro.
7074 (output_asm_line_debug_info): New.
7075 (struct var_loc_node): Add view.
7076 (add_AT_view_list, AT_loc_list): New.
7077 (add_var_loc_to_decl): Add view param. Test it against last.
7078 (new_loc_list): Add view params. Record them.
7079 (AT_loc_list_ptr): Handle loc and view lists.
7080 (view_list_to_loc_list_val_node): New.
7081 (print_dw_val): Handle dw_val_class_view_list.
7082 (size_of_die): Likewise.
7083 (value_format): Likewise.
7084 (loc_list_has_views): New.
7085 (gen_llsym): Set vl_symbol too.
7086 (maybe_gen_llsym, skip_loc_list_entry): New.
7087 (dwarf2out_maybe_output_loclist_view_pair): New.
7088 (output_loc_list): Output view list or entries too.
7089 (output_view_list_offset): New.
7090 (output_die): Handle dw_val_class_view_list.
7091 (output_dwarf_version): New.
7092 (output_compilation_unit_header): Use it.
7093 (output_skeleton_debug_sections): Likewise.
7094 (output_rnglists, output_line_info): Likewise.
7095 (output_pubnames, output_aranges): Update version comments.
7096 (output_one_line_info_table): Output view numbers in asm comments.
7097 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7098 Call maybe_gen_llsym.
7099 (loc_list_from_tree_1): Adjust.
7100 (add_AT_location_description): Create view list attribute if
7101 needed, check it's absent otherwise.
7102 (convert_cfa_to_fb_loc_list): Adjust.
7103 (maybe_emit_file): Call output_asm_line_debug_info for test.
7104 (dwarf2out_var_location): Reset views as needed. Precompute
7105 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7106 attribute. Set view.
7107 (new_line_info_table): Reset next view.
7108 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7109 (dwarf2out_source_line): Likewise. Output view resets and labels to
7110 the assembler, or select appropriate line info opcodes.
7111 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7112 (optimize_string_length): Catch it. Adjust.
7113 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7114 dw_val_class_view_list, and remove it if no longer needed.
7115 (hash_loc_list): Hash view numbers.
7116 (loc_list_hasher::equal): Compare them.
7117 (optimize_location_lists): Check whether a view list symbol is
7118 needed, and whether the locview attribute is present, and
7119 whether they match. Remove the locview attribute if no longer
7121 (index_location_lists): Call skip_loc_list_entry for test.
7122 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7123 Use output_dwarf_version.
7124 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7125 (struct dw_val_node): Add val_view_list.
7126 * final.c (SEEN_NEXT_VIEW): New.
7127 (set_next_view_needed): New.
7128 (clear_next_view_needed): New.
7129 (maybe_output_next_view): New.
7130 (final_start_function): Rename to...
7131 (final_start_function_1): ... this. Take pointer to FIRST,
7132 add SEEN parameter. Emit param bindings in the initial view.
7133 (final_start_function): Reintroduce SEEN-less interface.
7134 (final): Rename to...
7135 (final_1): ... this. Take SEEN parameter. Output final pending
7136 next view at the end.
7137 (final): Reintroduce seen-less interface.
7138 (final_scan_insn): Output pending next view before switching
7139 sections or ending a block. Mark the next view as needed when
7140 outputting variable locations. Notify debug backend of section
7141 changes, and of location view changes.
7142 (rest_of_handle_final): Adjust.
7143 * toplev.c (process_options): Autodetect value for debug variable
7144 location views option. Warn on incompat5 without -gdwarf-5.
7145 * doc/invoke.texi (gvariable-location-views): New.
7146 (gvariable-location-views=incompat5): New.
7147 (gno-variable-location-views): New.
7149 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7151 PR tree-optimization/84136
7152 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7153 that the result of find_edge is non-NULL.
7155 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7158 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7159 storing integer register in SImode. Fix cost of 256 and 512
7160 byte aligned SSE register store.
7162 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7164 * config/i386/i386.c (ix86_multiplication_cost): Fix
7165 multiplication cost for TARGET_AVX512DQ.
7167 2018-02-08 Marek Polacek <polacek@redhat.com>
7169 PR tree-optimization/84238
7170 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7173 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7175 PR tree-optimization/84265
7176 * tree-vect-stmts.c (vectorizable_store): Don't treat
7177 VMAT_CONTIGUOUS accesses as grouped.
7178 (vectorizable_load): Likewise.
7180 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7182 PR tree-optimization/81635
7183 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7184 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7185 (test_round_for_mask): New functions.
7186 (wide_int_cc_tests): Call test_round_for_mask.
7187 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7188 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7189 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7190 range returned by get_range_info.
7192 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7195 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7196 * symtab.c: Include builtins.h
7197 (symtab_node::output_to_lto_symbol_table_p): Move here
7198 from lto-streamer-out.c:output_symbol_p.
7199 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7200 (output_symbol_p): Move all logic to symtab.c
7201 (produce_symtab): Update.
7203 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7205 * config/s390/s390-opts.h (enum indirect_branch): Define.
7206 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7207 (s390_indirect_branch_via_thunk)
7208 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7209 (enum s390_indirect_branch_type): Define.
7210 * config/s390/s390.c (struct s390_frame_layout, struct
7211 machine_function): Remove.
7212 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7213 (indirect_branch_table_label_no, indirect_branch_table_name):
7215 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7216 (enum s390_indirect_branch_option): Define.
7217 (s390_return_addr_from_memory): New function.
7218 (s390_handle_string_attribute): New function.
7219 (s390_attribute_table): Add new attribute handler.
7220 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7221 (s390_indirect_branch_via_thunk): New function.
7222 (s390_indirect_branch_via_inline_thunk): New function.
7223 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7224 sibling call optimization for non z10 compiles.
7225 (s390_emit_call): Force indirect branch target to be a single
7226 register. Add r1 clobber for non-z10 compiles.
7227 (s390_emit_epilogue): Emit return jump via return_use expander.
7228 (s390_reorg): Handle JUMP_INSNs as execute targets.
7229 (s390_option_override_internal): Perform validity checks for the
7230 new command line options.
7231 (s390_indirect_branch_attrvalue): New function.
7232 (s390_indirect_branch_settings): New function.
7233 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7234 (s390_output_indirect_thunk_function): New function.
7235 (s390_code_end): Implement target hook.
7236 (s390_case_values_threshold): Implement target hook.
7237 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7239 * config/s390/s390.h (struct s390_frame_layout)
7240 (struct machine_function): Move here from s390.c.
7241 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7242 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7243 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7244 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7245 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7246 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7247 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7248 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7249 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7250 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7251 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7252 (mnemonic attribute): Add values which aren't recognized
7254 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7255 pattern for branch conversion. Fix mnemonic attribute.
7256 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7257 indirect branch via thunk if requested.
7258 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7259 ("*indirect_jump"): Disable for branch conversion using out of
7261 ("indirect_jump_via_thunk<mode>_z10")
7262 ("indirect_jump_via_thunk<mode>")
7263 ("indirect_jump_via_inlinethunk<mode>_z10")
7264 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7265 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7266 ("casesi_jump_via_inlinethunk<mode>_z10")
7267 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7268 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7269 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7270 ("*indirect2_jump"): Disable for branch conversion.
7271 ("casesi_jump"): Turn into expander and expand patterns for branch
7273 ("return_use"): New expander.
7274 ("*return"): Emit return via thunk and rename it to ...
7275 ("*return<mode>"): ... this one.
7276 * config/s390/s390.opt: Add new options and and enum for the
7279 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7281 * lra-constraints.c (match_reload): Unconditionally use
7282 gen_lowpart_SUBREG, rather than selecting between that
7283 and equivalent gen_rtx_SUBREG code.
7285 2018-02-08 Richard Biener <rguenther@suse.de>
7287 PR tree-optimization/84233
7288 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7289 changed flag instead of boguously re-using phi_inserted.
7291 2018-02-08 Martin Jambor <mjambor@suse.cz>
7293 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7294 static local variables.
7296 2018-02-08 Richard Biener <rguenther@suse.de>
7298 PR tree-optimization/84278
7299 * tree-vect-stmts.c (vectorizable_store): When looking for
7300 smaller vector types to perform grouped strided loads/stores
7301 make sure the mode is supported by the target.
7302 (vectorizable_load): Likewise.
7304 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7306 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7307 Increase LDP/STP opportunities by adding adjacent callee-saves.
7309 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7311 PR rtl-optimization/84068
7312 PR rtl-optimization/83459
7313 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7315 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7317 PR tree-optimization/84224
7318 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7319 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7322 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7325 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7326 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7327 count, remove one USE.
7329 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7331 * doc/install.texi (Configuration): Document the
7332 --with-long-double-format={ibm,ieee} PowerPC configuration
7336 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7337 Convert from define_expand to be define_insn_and_split. Rework
7338 float/double/_Float128 conversions to QI/HI/SImode to work with
7339 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7340 conversions to QI/HImode types did a store and then a load to
7341 truncate the value. For conversions to VSX registers, don't split
7342 the insn, instead emit the code directly. Use the code iterator
7343 any_fix to combine signed and unsigned conversions.
7344 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7345 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7346 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7347 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7348 (fix_<mode>di2_hw): Likewise.
7349 (fixuns_<mode>di2_hw): Likewise.
7350 (fix_<mode>si2_hw): Likewise.
7351 (fixuns_<mode>si2_hw): Likewise.
7352 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7353 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7354 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7355 fix<uns>_trunc<SFDF:mode>si2_p8.
7356 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7358 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7359 (fix<uns>_<mode>_mem): Likewise.
7360 (fctiw<u>z_<mode>_mem): Likewise.
7361 (fix<uns>_<mode>_mem): Likewise.
7362 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7363 the register allocator from doing a direct move to the GPRs to do
7364 a store, and instead use the ISA 3.0 store byte/half-word from
7365 vector register instruction. For IEEE 128-bit floating point,
7366 also optimize stores of 32-bit ints.
7367 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7369 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7371 * genextract.c (push_pathstr_operand): New function to support
7373 (walk_rtx): Call push_pathstr_operand.
7374 (print_path): Support [a-zA-Z].
7376 2018-02-07 Richard Biener <rguenther@suse.de>
7378 PR tree-optimization/84037
7379 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7380 (cse_and_gimplify_to_preheader): Declare.
7381 (vect_get_place_in_interleaving_chain): Likewise.
7382 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7384 (_loop_vec_info::~_loop_vec_info): Delete it.
7385 (cse_and_gimplify_to_preheader): New function.
7386 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7387 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7388 (vectorizable_load): Likewise. For grouped stores always base
7389 the IV on the first element.
7390 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7391 condition before gimplifying.
7393 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7395 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7396 *DIV_EXPR and *MOD_EXPR.
7398 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7401 * config/i386/i386.c (ix86_option_override_internal): Mask out
7402 the CF_SET bit when checking -fcf-protection.
7404 2018-02-07 Tom de Vries <tom@codesourcery.com>
7407 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7410 2018-02-07 Richard Biener <rguenther@suse.de>
7412 PR tree-optimization/84204
7413 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7416 PR tree-optimization/84205
7417 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7418 special-case isl_ast_op_zdiv_r.
7420 PR tree-optimization/84223
7421 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7422 Only add conditions from within the region.
7423 (gather_bbs::after_dom_children): Adjust.
7425 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7428 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7429 * config/avr/avr.md: Only post-reload split REG-REG moves if
7430 either register is GENERAL_REG_P.
7432 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7434 PR tree-optimization/84235
7435 * tree-ssa-scopedtables.c
7436 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7437 if the subtraction is performed in floating point type where NaNs are
7438 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7439 build 1. Formatting fix.
7441 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7444 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7445 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7446 and skip it regardless of bb boundaries. Use CALL_P macro,
7447 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7448 unnecessarily, formatting fix.
7450 2018-02-06 Michael Collison <michael.collison@arm.com>
7452 * config/arm/thumb2.md:
7453 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7454 (*thumb_mov_notscc): Ditto.
7456 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7459 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7460 unsigned_fix, not "s".
7462 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7464 * configure.ac (gcc_fn_eh_frame_ro): New function.
7465 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7466 correct .eh_frame permissions.
7467 * configure: Regenerate.
7469 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7471 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7474 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7476 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7477 Display warning message for -mno-speculate-indirect-jumps.
7479 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7481 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7483 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7485 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7487 PR tree-optimization/84225
7488 * tree-eh.c (find_trapping_overflow): Only call
7489 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7491 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7494 * config/i386/i386.c: Reimplement the check of possible options
7495 -mibt/-mshstk conbination. Change error messages.
7496 * doc/invoke.texi: Fix a typo: remove extra '='.
7498 2018-02-06 Marek Polacek <polacek@redhat.com>
7500 PR tree-optimization/84228
7501 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7503 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7506 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7507 emitted arch directives.
7508 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7509 __ARM_FEATURE_COPROC before changing architectures.
7511 2018-02-06 Richard Biener <rguenther@suse.de>
7513 * config/i386/i386.c (print_reg): Fix typo.
7514 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7516 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7518 * configure: Regenerate.
7520 2018-02-05 Martin Sebor <msebor@redhat.com>
7522 PR tree-optimization/83369
7523 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7526 2018-02-05 Martin Liska <mliska@suse.cz>
7528 * doc/invoke.texi: Cherry-pick upstream r323995.
7530 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7532 * ira.c (ira_init_register_move_cost): Adjust comment.
7534 2018-02-05 Martin Liska <mliska@suse.cz>
7536 PR gcov-profile/84137
7537 * doc/gcov.texi: Fix typo in documentation.
7539 2018-02-05 Martin Liska <mliska@suse.cz>
7541 PR gcov-profile/83879
7542 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7543 using dlopen functionality.
7545 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7547 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7548 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7549 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7550 _mm_maskz_range_round_ss): New intrinsics.
7551 (__builtin_ia32_rangesd128_round)
7552 (__builtin_ia32_rangess128_round): Remove.
7553 (__builtin_ia32_rangesd128_mask_round,
7554 __builtin_ia32_rangess128_mask_round): New builtins.
7555 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7556 __builtin_ia32_rangess128_round): Remove.
7557 (__builtin_ia32_rangesd128_mask_round,
7558 __builtin_ia32_rangess128_mask_round): New builtins.
7559 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7560 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7561 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7562 "<round_saeonly_constraint>")): Changed to ...
7563 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7564 "<round_saeonly_scalar_constraint>")): ... this.
7565 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7566 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7567 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7568 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7569 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7571 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7573 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7575 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7576 Remove all values except native, 8540 and 8548.
7578 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7580 * config/i386/i386.c (ix86_output_function_return): Pass
7581 INVALID_REGNUM, instead of -1, as invalid register number to
7582 indirect_thunk_name and output_indirect_thunk.
7584 2018-02-02 Julia Koval <julia.koval@intel.com>
7586 * config.gcc: Add -march=icelake.
7587 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7588 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7589 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7590 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7591 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7592 (processor_target_table): Add icelake.
7593 (ix86_option_override_internal): Handle new PTAs.
7594 (get_builtin_code_for_version): Handle icelake.
7595 (M_INTEL_COREI7_ICELAKE): New.
7596 (fold_builtin_cpu): Handle icelake.
7597 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7598 * doc/invoke.texi: Add -march=icelake.
7600 2018-02-02 Julia Koval <julia.koval@intel.com>
7602 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7603 to wide_int_bitmask.
7604 * wide-int-bitmask.h: New.
7606 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7609 * config/i386/i386.md: Replace Pmode with word_mode in
7610 builtin_setjmp_setup and builtin_longjmp to support x32.
7612 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7616 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7618 (rs6000_supported_cpu_names): New static variable.
7619 (linux_cpu_translation_table): Likewise.
7620 (elf_platform) <cpu>: Define new static variable and use it.
7621 Translate kernel AT_PLATFORM name to canonical name if needed.
7622 Error if platform name is unknown.
7624 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7627 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7629 2018-02-01 Jeff Law <law@redhat.com>
7632 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7633 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7634 the scratch if RELEASE_VIA_POP is false.
7635 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7636 If we have to save a temporary register, decrement SIZE appropriately.
7637 Pass new arguments to release_scratch_register_on_entry.
7638 (ix86_adjust_stack_and_probe): Likewise.
7639 (ix86_emit_probe_stack_range): Pass new arguments to
7640 release_scratch_register_on_entry.
7642 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7644 PR rtl-optimization/84157
7645 * combine.c (change_zero_ext): Use REG_P predicate in
7646 front of HARD_REGISTER_P predicate.
7648 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7650 * config/avr/avr.c (avr_option_override): Move disabling of
7651 -fdelete-null-pointer-checks to...
7652 * common/config/avr/avr-common.c (avr_option_optimization_table):
7655 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7657 PR tree-optimization/81635
7658 * tree-data-ref.c (split_constant_offset_1): For types that
7659 wrap on overflow, try to use range info to prove that wrapping
7662 2018-02-01 Renlin Li <renlin.li@arm.com>
7665 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7667 (aarch64_register_move_cost): Likewise.
7668 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7670 (REG_CLASS_NAMES): Likewise.
7671 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7672 TAILCALL_ADDR_REGS. Remove IP registers.
7673 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7675 2018-02-01 Richard Biener <rguenther@suse.de>
7677 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7678 for specifying RPO order and allow NULL for that.
7679 * domwalk.c (dom_walker::dom_walker): Likewise.
7680 (dom_walker::walk): Handle NULL RPO order.
7681 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7683 (rewrite_update_dom_walker): Likewise.
7684 (mark_def_dom_walker): Likewise.
7686 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7688 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7689 (aarch64_maybe_expand_sve_subreg_move): Declare.
7690 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7691 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7693 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7694 that are semantically a reverse operation.
7695 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7696 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7697 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7699 (aarch64_can_change_mode_class): For big-endian, forbid changes
7700 between two SVE modes if they have different element sizes.
7702 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7704 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7705 the TImode handling for big-endian targets.
7707 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7709 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7710 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7712 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7713 Remove BSWAP handing for big-endian targets and use the form of
7714 LD1RQ appropariate for the mode.
7716 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7718 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7719 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7722 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7725 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7726 check for operands that need to go through aarch64_sve_reload_be.
7728 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7730 PR tree-optimization/81661
7731 PR tree-optimization/84117
7732 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7733 * tree-eh.c: Include gimplify.h.
7734 (find_trapping_overflow, replace_trapping_overflow,
7735 rewrite_to_non_trapping_overflow): New functions.
7736 * tree-vect-loop.c: Include tree-eh.h.
7737 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7738 * tree-data-ref.c: Include tree-eh.h.
7739 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7741 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7743 PR rtl-optimization/84123
7744 * combine.c (change_zero_ext): Check if hard register satisfies
7745 can_change_dest_mode before calling gen_lowpart_SUBREG.
7747 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7750 * ira.c (ira_init_register_move_cost): Remove assert.
7752 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7754 PR rtl-optimization/84071
7755 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7756 * doc/tm.texi: Regenerate.
7758 2018-01-31 Richard Biener <rguenther@suse.de>
7760 PR tree-optimization/84132
7761 * tree-data-ref.c (analyze_miv_subscript): Properly
7762 check whether evolution_function_is_affine_multivariate_p
7763 before calling gcd_of_steps_may_divide_p.
7765 2018-01-31 Julia Koval <julia.koval@intel.com>
7768 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7769 * config/i386/i386.md (rdpid_rex64) New.
7770 (rdpid): Make 32bit only.
7772 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7775 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7776 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7778 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7781 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7783 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7785 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7787 PR rtl-optimization/84071
7788 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7789 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7791 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7793 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7794 (arc_attribute_table): Add 'aux' attribute.
7795 (arc_in_small_data_p): Consider aux like variables.
7796 (arc_is_aux_reg_p): New function.
7797 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7798 (arc_get_aux_arg): New function.
7799 (prepare_move_operands): Handle aux-register access.
7800 (arc_handle_aux_attribute): New function.
7801 * doc/extend.texi (ARC Variable attributes): Add subsection.
7803 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7805 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7806 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7807 (arc_attribute_table): Add 'uncached' attribute.
7808 (arc_print_operand): Print '.di' flag for uncached memory
7810 (arc_in_small_data_p): Do not consider for small data the uncached
7812 (arc_is_uncached_mem_p): New function.
7813 * config/arc/predicates.md (compact_store_memory_operand): Check
7814 for uncached memory accesses.
7815 (nonvol_nonimm_operand): Likewise.
7816 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7818 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7821 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7822 falign-loops=): Add Optimization flag.
7824 2018-01-30 Jeff Law <law@redhat.com>
7827 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7828 INT_REGISTERS_SAVED. Check it prior to calling
7829 get_scratch_register_on_entry.
7830 (ix86_adjust_stack_and_probe): Similarly.
7831 (ix86_emit_probe_stack_range): Similarly.
7832 (ix86_expand_prologue): Corresponding changes.
7834 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7837 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7838 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7840 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7843 * lra-constraints.c (curr_insn_transform): Process AND in the
7846 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7848 PR rtl-optimization/83986
7849 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7850 dependence against last_pending_memory_flush in addition to
7853 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7855 PR tree-optimization/81611
7856 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7859 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7862 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7865 2018-01-30 Richard Biener <rguenther@suse.de>
7866 Jakub Jelinek <jakub@redhat.com>
7868 PR tree-optimization/84111
7869 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7870 inner loops added during recursion, as they don't have up-to-date
7873 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7876 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7877 (can_inline_edge_by_limits_p): ... here.
7878 (can_early_inline_edge_p, check_callers,
7879 update_caller_keys, update_callee_keys, recursive_inlining,
7880 add_new_edges_to_heap, speculation_useful_p,
7881 inline_small_functions,
7882 inline_small_functions, flatten_function,
7883 inline_to_all_callers_1): Update.
7885 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7887 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7890 2018-01-30 Richard Biener <rguenther@suse.de>
7892 PR tree-optimization/83008
7893 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7894 invariant and constant vector uses in stmts when they need
7897 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7900 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7901 * configure: Regenerate.
7903 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7905 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7907 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7908 Use gen_rtx_REG rather than gen_lowpart.
7910 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7912 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7913 rather than 0 when creating partial subregs.
7915 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7917 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7920 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7923 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7924 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7925 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7926 flags. This restores the settings used before the 2017-07-24.
7927 Turning off pre increment/decrement/modify allows IVOPTS to
7928 optimize DF/SF loops where the index is an int.
7930 2018-01-29 Richard Biener <rguenther@suse.de>
7931 Kelvin Nilsen <kelvin@gcc.gnu.org>
7934 * tree-vect-stmts.c (vectorizable_call): Don't call
7935 targetm.vectorize_builtin_md_vectorized_function if callee is
7938 2018-01-22 Carl Love <cel@us.ibm.com>
7940 * doc/extend.tex: Fix typo in second arg in
7941 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7943 2018-01-29 Richard Biener <rguenther@suse.de>
7945 PR tree-optimization/84086
7946 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7947 (flush_ssaname_freelist): When SSA names were released reset
7948 the SCEV hash table.
7950 2018-01-29 Richard Biener <rguenther@suse.de>
7952 PR tree-optimization/84057
7953 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7954 removed paths when removing edges.
7956 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7958 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7959 -mfunction-return=@var{choice}.
7961 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7964 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7966 (layout::print_source_line): Likewise.
7967 (test_get_line_width_without_trailing_whitespace): Add test cases.
7969 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7972 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7975 2018-01-26 Jim Wilson <jimw@sifive.com>
7977 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7979 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7982 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7984 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7985 and CMP + SUB-immediate -> SUBS.
7987 2018-01-26 Martin Sebor <msebor@redhat.com>
7989 PR tree-optimization/83896
7990 * tree-ssa-strlen.c (get_string_len): Rename...
7991 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7992 Avoid assuming length is constant.
7993 (handle_char_store): Use HOST_WIDE_INT for string length.
7995 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7998 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7999 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8001 2018-01-26 Richard Biener <rguenther@suse.de>
8003 PR rtl-optimization/84003
8004 * dse.c (record_store): Only record redundant stores when
8005 the earlier store aliases at least all accesses the later one does.
8007 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8009 PR rtl-optimization/83985
8010 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8011 REG_CFA_RESTORE insns.
8012 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8013 REG_CFA_RESTORE insns here.
8016 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8017 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8019 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8021 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8022 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8023 (arc_init): Likewise.
8024 (arc_override_options): Likewise.
8025 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8027 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8029 * config/arc/arc.h (TARGET_DBNZ): Define.
8030 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8031 properly set the tune attribute.
8032 (dbnz): Use TARGET_DBNZ guard.
8033 * config/arc/arc.opt (mtune): Add core3 option.
8035 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8037 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8038 recognize new pic like addresses.
8039 (arc_delegitimize_address): Clean up.
8041 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8043 * config/arc/arc-arches.def: Option mrf16 valid for all
8045 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8046 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8047 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8048 * config/arc/arc-tables.opt: Regenerate.
8049 * config/arc/arc.c (arc_conditional_register_usage): Handle
8050 reduced register file case.
8051 (arc_file_start): Set must have build attributes.
8052 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8054 * config/arc/arc.opt (mrf16): Add new option.
8055 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8056 * config/arc/genmultilib.awk: Handle new mrf16 option.
8057 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8058 * config/arc/t-multilib: Regenerate.
8059 * doc/invoke.texi (ARC Options): Document mrf16 option.
8061 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8063 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8064 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8065 (arc_attribute_table): Add 'secure_call' attribute.
8066 (arc_print_operand): Print secure call operand.
8067 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8069 (arc_is_secure_call_p): New function. * config/arc/arc.md
8070 (call_i): Add support for sjli instruction.
8071 (call_value_i): Likewise.
8072 * config/arc/constraints.md (Csc): New constraint.
8074 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8075 John Eric Martin <John.Martin@emmicro-us.com>
8077 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8078 * config/arc/arc.c (_arc_jli_section): New struct.
8079 (arc_jli_section): New type.
8080 (rc_jli_sections): New static variable.
8081 (arc_handle_jli_attribute): New function.
8082 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8083 (arc_file_end): New function.
8084 (TARGET_ASM_FILE_END): Define.
8085 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8086 (arc_add_jli_section): New function.
8087 (jli_call_scan): Likewise.
8088 (arc_reorg): Call jli_call_scan.
8089 (arc_output_addsi): Remove 'S' from printing asm operand.
8090 (arc_is_jli_call_p): New function.
8091 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8093 (movhi_insn): Likewise.
8094 (movsi_insn): Likewise.
8095 (movsi_set_cc_insn): Likewise.
8096 (loadqi_update): Likewise.
8097 (load_zeroextendqisi_update): Likewise.
8098 (load_signextendqisi_update): Likewise.
8099 (loadhi_update): Likewise.
8100 (load_zeroextendhisi_update): Likewise.
8101 (load_signextendhisi_update): Likewise.
8102 (loadsi_update): Likewise.
8103 (loadsf_update): Likewise.
8104 (movsicc_insn): Likewise.
8105 (bset_insn): Likewise.
8106 (bxor_insn): Likewise.
8107 (bclr_insn): Likewise.
8108 (bmsk_insn): Likewise.
8109 (bicsi3_insn): Likewise.
8110 (cmpsi_cc_c_insn): Likewise.
8111 (movsi_ne): Likewise.
8112 (movsi_cond_exec): Likewise.
8113 (clrsbsi2): Likewise.
8124 (call_i): Remove 'S' asm letter, add jli instruction.
8125 (call_value_i): Likewise.
8126 * config/arc/arc.op (mjli-always): New option.
8127 * config/arc/constraints.md (Cji): New constraint.
8128 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8130 (subsf3_fpx): Likewise.
8131 (mulsf3_fpx): Likewise.
8132 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8134 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8136 * doc/invoke.texi (ARC): Document mjli-always option.
8138 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8140 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8141 avoid addition with 0 and use incw and decw where possible.
8143 2018-01-26 Richard Biener <rguenther@suse.de>
8145 PR tree-optimization/81082
8146 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8147 association if it requires casting to unsigned.
8148 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8149 from fold_plusminus_mult_expr to catch important cases late when
8150 range info is available.
8152 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8154 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8155 * configure.ac (hidden_linkonce): New test.
8156 * configure: Regenerate.
8157 * config.in: Regenerate.
8159 2018-01-26 Julia Koval <julia.koval@intel.com>
8161 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8162 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8163 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8164 _mm_mask_bitshuffle_epi64_mask): Fix type.
8165 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8166 USI_FTYPE_V4DI_V4DI_USI): Remove.
8167 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8168 __builtin_ia32_vpshufbitqmb256_mask,
8169 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8170 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8171 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8173 2018-01-26 Alan Modra <amodra@gmail.com>
8176 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8177 UNSPEC_VBPERMQ. Sort other unspecs.
8179 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8181 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8183 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8186 * predict.c (drop_profile): Do not push/pop cfun; update also
8188 (handle_missing_profiles): Fix logic looking for zero profiles.
8190 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8193 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8194 on functions with #pragma omp declare simd or functions with simd
8196 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8197 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8198 Remove trailing \n from warning_at calls.
8200 2018-01-25 Tom de Vries <tom@codesourcery.com>
8203 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8204 for neutered workers.
8206 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8209 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8210 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8212 2018-01-24 Jeff Law <law@redhat.com>
8215 * i386.c (get_probe_interval): Move to earlier point.
8216 (ix86_compute_frame_layout): If -fstack-clash-protection and
8217 the frame is larger than the probe interval, then use pushes
8218 to save registers rather than reg->mem moves.
8219 (ix86_expand_prologue): Remove conditional for int_registers_saved
8222 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8225 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8226 min/max for never referenced object.
8228 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8231 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8233 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8234 attributes from DECL_ATTRIBUTES (decl) without affecting
8235 DECL_ATTRIBUTES (current_function_decl).
8236 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8237 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8239 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8241 PR tree-optimization/83979
8242 * fold-const.c (fold_comparison): Use constant_boolean_node
8243 instead of boolean_{true,false}_node.
8245 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8247 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8250 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8252 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8253 Simplify the clause that sets the length attribute.
8254 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8255 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8256 clause that sets the length attribute.
8257 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8259 2018-01-24 Tom de Vries <tom@codesourcery.com>
8262 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8263 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8264 Add strict parameter.
8265 (prevent_branch_around_nothing): Insert dummy insn between branch to
8266 label and label with no ptx insn inbetween.
8267 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8269 2018-01-24 Tom de Vries <tom@codesourcery.com>
8272 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8273 for neutered threads in warp.
8274 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8276 2018-01-24 Richard Biener <rguenther@suse.de>
8278 PR tree-optimization/83176
8279 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8282 2018-01-24 Richard Biener <rguenther@suse.de>
8284 PR tree-optimization/82819
8285 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8286 code generating pluses that are no-ops in the target precision.
8288 2018-01-24 Richard Biener <rguenther@suse.de>
8291 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8293 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8295 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8296 to merge probabilities.
8297 * predict.c (probably_never_executed): Also mark as cold functions
8298 with global 0 profile and guessed local profile.
8299 * profile-count.c (profile_probability::combine_with_count): New
8301 * profile-count.h (profile_probability::operator*,
8302 profile_probability::operator*=, profile_probability::operator/,
8303 profile_probability::operator/=): Reduce precision to adjusted
8304 and set value to guessed on contradictory divisions.
8305 (profile_probability::combine_with_freq): Remove.
8306 (profile_probability::combine_wiht_count): Declare.
8307 (profile_count::force_nonzero):: Set to adjusted.
8308 (profile_count::probability_in):: Set quality to adjusted.
8309 * tree-ssa-tail-merge.c (replace_block_by): Use
8312 2018-01-23 Andrew Waterman <andrew@sifive.com>
8313 Jim Wilson <jimw@sifive.com>
8315 * config/riscv/riscv.c (riscv_stack_boundary): New.
8316 (riscv_option_override): Set riscv_stack_boundary. Handle
8317 riscv_preferred_stack_boundary_arg.
8318 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8319 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8320 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8321 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8322 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8323 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8325 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8328 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8329 of struct ix86_frame.
8330 (ix86_expand_epilogue): Likewise. Add a local variable for
8331 the reg_save_offset field in struct ix86_frame.
8333 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8335 PR tree-optimization/82604
8336 * tree-loop-distribution.c (enum partition_kind): New enum item
8337 PKIND_PARTIAL_MEMSET.
8338 (partition_builtin_p): Support above new enum item.
8339 (generate_code_for_partition): Ditto.
8340 (compute_access_range): Differentiate cases that equality can be
8341 proven at all loops, the innermost loops or no loops.
8342 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8343 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8344 (finalize_partitions, distribute_loop): Don't fuse partition of
8345 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8346 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8349 2018-01-23 Martin Liska <mliska@suse.cz>
8351 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8352 order to ignore the predictor.
8353 (PRED_POLYMORPHIC_CALL): Likewise.
8354 (PRED_RECURSIVE_CALL): Likewise.
8356 2018-01-23 Martin Liska <mliska@suse.cz>
8358 * tree-profile.c (tree_profiling): Print function header to
8359 aware reader which function we are working on.
8360 * value-prof.c (gimple_find_values_to_profile): Do not print
8361 not interesting value histograms.
8363 2018-01-23 Martin Liska <mliska@suse.cz>
8365 * profile-count.h (enum profile_quality): Add
8366 profile_uninitialized as the first value. Do not number values
8367 as they are zero based.
8368 (profile_count::verify): Update sanity check.
8369 (profile_probability::verify): Likewise.
8371 2018-01-23 Nathan Sidwell <nathan@acm.org>
8373 * doc/invoke.texi (ffor-scope): Deprecate.
8375 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8377 PR tree-optimization/83510
8378 * domwalk.c (set_all_edges_as_executable): New function.
8379 (dom_walker::dom_walker): Convert bool param
8380 "skip_unreachable_blocks" to enum reachability. Move setup of
8381 edge flags to set_all_edges_as_executable and only do it when
8382 reachability is REACHABLE_BLOCKS.
8383 * domwalk.h (enum dom_walker::reachability): New enum.
8384 (dom_walker::dom_walker): Convert bool param
8385 "skip_unreachable_blocks" to enum reachability.
8386 (set_all_edges_as_executable): New decl.
8387 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8388 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8390 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8391 but converting true to REACHABLE_BLOCKS.
8392 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8394 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8395 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8396 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8398 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8399 if check_all_array_refs will be called.
8401 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8403 * tree.c (selftest::test_location_wrappers): Add more test
8406 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8408 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8409 (selftest::test_bit_in_range): Likewise.
8411 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8414 * doc/sourcebuild.texi (vect_float): Say that the selector
8415 only describes the situation when -funsafe-math-optimizations is on.
8416 (vect_float_strict): Document.
8418 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8420 PR tree-optimization/83965
8421 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8422 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8423 instead of checking only for a reduction.
8424 (vect_recog_widen_sum_pattern): Likewise.
8426 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8428 * predict.c (probably_never_executed): Only use precise profile info.
8429 (compute_function_frequency): Skip after inlining hack since we now
8430 have quality checking.
8432 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8434 * profile-count.h (profile_probability::very_unlikely,
8435 profile_probability::unlikely, profile_probability::even): Set
8436 precision to guessed.
8438 2018-01-23 Richard Biener <rguenther@suse.de>
8440 PR tree-optimization/83963
8441 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8442 Properly terminate dominator walk when crossing the exit edge not
8443 when visiting its source block.
8445 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8448 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8449 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8451 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8453 PR tree-optimization/83957
8454 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8455 semicolon after for body surrounded by braces.
8457 PR tree-optimization/83081
8458 * profile-count.h (profile_probability::split): New method.
8459 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8460 Use profile_probability::split.
8461 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8462 when splitting a single conditional jump into 2.
8464 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8466 PR tree-optimization/69452
8467 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8470 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8472 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8473 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8474 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8476 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8478 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8479 * config/rl78/rl78.md (movdi): New define_expand.
8480 * config/rl78/rl78.c (rl78_split_movdi): New function.
8482 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8485 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8487 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8488 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8489 128-bit to produce an UNSPEC move to get the double word with the
8490 signbit and then a shift directly to do signbit.
8491 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8492 implementation with a new version that just does either a direct
8493 move or a regular move. Move memory interface to separate insns.
8494 Move insns so they are next to the expander.
8495 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8496 with signbit move. Split big and little endian case.
8497 (signbit<mode>2_dm_mem_le): Likewise.
8498 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8499 (signbit<mode>2_dm2): Likewise.
8501 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8503 * config/rl78/rl78.md (anddi3): New define_expand.
8505 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8507 * config/rl78/rl78.md (umindi3): New define_expand.
8509 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8511 * config/rl78/rl78.md (smindi3): New define_expand.
8513 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8515 * config/rl78/rl78.md (smaxdi3): New define_expand.
8517 2018-01-22 Carl Love <cel@us.ibm.com>
8519 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8520 LVX_V1TI): Add macro expansion.
8521 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8522 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8523 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8524 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8525 Change check to determine if the instruction is a byte reversing
8526 entry. Fix typo in comment.
8527 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8528 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8529 Add def_builtin calls for new builtins.
8530 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8531 Add define_insn expansion.
8533 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8535 * config/rl78/rl78.md (umaxdi3): New define_expand.
8537 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8539 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8540 for non-QImode registers.
8542 2018-01-22 Richard Biener <rguenther@suse.de>
8544 PR tree-optimization/83963
8545 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8546 including the loop exit block.
8547 (scop_detection::merge_sese): Likewise.
8548 (scop_detection::add_scop): Do it here instead.
8550 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8552 * doc/sourcebuild.texi (arm_softfloat): Document.
8554 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8557 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8558 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8559 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8561 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8562 David Edelsohn <dje.gcc@gmail.com>
8565 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8566 Change "crset eq" to "crset 2".
8567 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8568 (*call_indirect_aix<mode>_nospec): Likewise.
8569 (*call_value_indirect_aix<mode>_nospec): Likewise.
8570 (*call_indirect_elfv2<mode>_nospec): Likewise.
8571 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8572 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8573 change assembly output from . to $.
8574 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8575 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8576 (*tablejump<mode>_internal1_nospec): Likewise.
8578 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8581 * config/sh/sh_optimize_sett_clrt.cc:
8582 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8584 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8586 PR tree-optimization/83940
8587 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8588 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8589 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8590 use it to pass back the definition type.
8591 (vect_check_store_rhs): Likewise rhs_dt_out.
8592 (vect_build_gather_load_calls): Add a mask_dt argument and use
8593 it instead of a call to vect_is_simple_use.
8594 (vectorizable_store): Update calls to vect_check_load_store_mask
8595 and vect_check_store_rhs. Use the dt returned by the latter instead
8596 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8597 instead of calls to vect_is_simple_use. Pass the scalar rather
8598 than the vector operand to vect_is_simple_use when handling
8599 second and subsequent copies of an rhs value.
8600 (vectorizable_load): Update calls to vect_check_load_store_mask
8601 and vect_build_gather_load_calls. Use the cached mask_dt and
8602 gs_info.offset_dt instead of calls to vect_is_simple_use.
8604 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8607 * tree-emutls.c: Include gimplify.h.
8608 (lower_emutls_2): New function.
8609 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8610 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8611 it before further processing.
8614 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8615 UINTVAL (trueop1) instead of INTVAL (op1).
8617 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8621 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8622 INCOMING_FRAME_SP_OFFSET if not defined.
8623 (scan_trace): Add ENTRY argument. If true and
8624 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8625 emit a note to adjust the CFA offset.
8626 (create_cfi_notes): Adjust scan_trace callers.
8627 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8628 INCOMING_FRAME_SP_OFFSET in the CIE.
8629 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8630 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8632 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8633 * doc/tm.texi: Regenerated.
8635 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8637 PR rtl-optimization/83147
8638 * lra-constraints.c (remove_inheritance_pseudos): Use
8639 lra_substitute_pseudo_within_insn.
8641 2018-01-19 Tom de Vries <tom@codesourcery.com>
8642 Cesar Philippidis <cesar@codesourcery.com>
8645 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8647 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8650 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8651 spaces for function labels.
8653 2018-01-19 Martin Liska <mliska@suse.cz>
8655 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8656 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8657 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8658 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8659 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8660 (PRED_CONST_RETURN): Change from 69 to 65.
8661 (PRED_NULL_RETURN): Change from 91 to 71.
8662 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8663 (PRED_LOOP_GUARD): Change from 66 to 73.
8665 2018-01-19 Martin Liska <mliska@suse.cz>
8667 * predict.c (predict_insn_def): Add new assert.
8668 (struct branch_predictor): Change type to signed integer.
8669 (test_prediction_value_range): Amend test to cover
8671 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8672 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8673 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8674 (PRED_LOOP_IV_COMPARE): Likewise.
8675 * predict.h (PROB_UNINITIALIZED): Define new constant.
8677 2018-01-19 Martin Liska <mliska@suse.cz>
8679 * predict.c (dump_prediction): Add new format for
8680 analyze_brprob.py script which is enabled with -details
8682 * profile-count.h (precise_p): New function.
8684 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8686 PR tree-optimization/83922
8687 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8688 there are no statements that need masking.
8689 (vect_active_double_reduction_p): New function.
8690 (vect_analyze_loop_operations): Use it when handling phis that
8691 are not in the loop header.
8693 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8695 PR tree-optimization/83914
8696 * tree-vect-loop.c (vectorizable_induction): Don't convert
8697 init_expr or apply the peeling adjustment for inductions
8698 that are nested within the vectorized loop.
8700 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8702 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8705 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8709 * function.h (gimplify_parameters): Add gimple_seq * argument.
8710 * function.c: Include gimple.h and options.h.
8711 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8712 for the added local temporaries if needed.
8713 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8714 if there are any parameter cleanups, wrap whole body into a
8715 try/finally with the cleanups.
8717 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8720 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8721 Use GET_MODE_CLASS for scalar floating point.
8723 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8727 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8728 Fix call of call_cgraph_insertion_hooks.
8730 2018-01-18 Martin Sebor <msebor@redhat.com>
8732 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8734 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8737 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8740 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8743 * common.opt: (-ffile-prefix-map): New option.
8744 * opts.c (common_handle_option): Defer it.
8745 * opts-global.c (handle_common_deferred_options): Handle it.
8746 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8747 * file-prefix-map.h: New file.
8748 (remap_debug_filename, add_debug_prefix_map): ...here.
8749 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8750 * final.c (debug_prefix_map, add_debug_prefix_map
8751 remap_debug_filename): Move to...
8752 * file-prefix-map.c: New file.
8753 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8754 generalize, get rid of alloca(), use strrchr() instead of strchr().
8755 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8756 Implement in terms of add_prefix_map().
8757 (remap_macro_filename, remap_debug_filename): Implement in term of
8759 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8760 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8761 * dbxout.c: Include file-prefix-map.h.
8762 * varasm.c: Likewise.
8763 * vmsdbgout.c: Likewise.
8764 * xcoffout.c: Likewise.
8765 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8766 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8767 * doc/invoke.texi (-ffile-prefix-map): Document.
8768 (-fdebug-prefix-map): Update description.
8770 2018-01-18 Martin Liska <mliska@suse.cz>
8772 * config/i386/i386.c (indirect_thunk_name): Document that also
8774 (output_indirect_thunk): Document why both instructions
8775 (pause and lfence) are generated.
8777 2018-01-18 Richard Biener <rguenther@suse.de>
8779 PR tree-optimization/83887
8780 * graphite-scop-detection.c
8781 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8782 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8783 (scop_detection::merge_sese): Re-implement with a flood-fill
8784 algorithm that properly finds a SESE region if it exists.
8786 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8789 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8790 pointer_diff optimizations use view_convert instead of convert.
8792 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8794 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8795 Generate different code for -mno-speculate-indirect-jumps.
8796 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8797 (*call_indirect_aix<mode>): Disable for
8798 -mno-speculate-indirect-jumps.
8799 (*call_indirect_aix<mode>_nospec): New define_insn.
8800 (*call_value_indirect_aix<mode>): Disable for
8801 -mno-speculate-indirect-jumps.
8802 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8803 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8804 -mno-speculate-indirect-jumps.
8805 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8807 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8809 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8810 long double type, set the flags for noting the default long double
8811 type, even if we don't pass or return a long double type.
8813 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8816 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8819 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8821 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8822 support for merge[hl].
8823 (fold_mergehl_helper): New helper function.
8824 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8825 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8826 (altivec_vmrglw_direct): Add xxmrglw insn.
8828 2018-01-17 Andrew Waterman <andrew@sifive.com>
8830 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8831 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8833 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8836 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8837 call the lto_location_cache before reading the
8838 DECL_SOURCE_LOCATION of the types.
8840 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8841 Richard Sandiford <richard.sandiford@linaro.org>
8843 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8844 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8845 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8846 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8847 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8849 * config/aarch64/constraints.md (aarch64_movti_operand):
8851 * config/aarch64/predicates.md (Uti): Add new constraint.
8853 2018-01-17 Carl Love <cel@us.ibm.com>
8855 * config/rs6000/vsx.md (define_expand xl_len_r,
8856 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8857 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8859 (define_expand, define_insn): Move the shift left from the
8860 define_insn to the define_expand for lxvl and stxvl instructions.
8861 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8862 and XL_LEN_R definitions to PURE.
8864 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8866 * config/i386/i386.c (indirect_thunk_name): Declare regno
8867 as unsigned int. Compare regno with INVALID_REGNUM.
8868 (output_indirect_thunk): Ditto.
8869 (output_indirect_thunk_function): Ditto.
8870 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8871 in the call to output_indirect_thunk_function.
8873 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8876 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8877 rather than the size of inner_type to determine the stack slot size
8878 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8880 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8883 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8886 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8888 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8889 endian Linux systems to optionally enable multilibs for selecting
8890 the long double type if the user configured an explicit type.
8891 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8892 have no long double multilibs if not defined.
8893 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8894 warn if the user used -mabi={ieee,ibm}longdouble and we built
8895 multilibs for long double.
8896 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8897 appropriate multilib option.
8898 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8900 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8901 for building long double multilibs.
8902 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8904 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8906 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8909 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8911 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8914 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8917 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8920 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8922 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8923 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8925 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8927 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8928 different rtl trees depending on TARGET_64BIT.
8929 (rs6000_gen_lvx): Likewise.
8931 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8933 * config/visium/visium.md (nop): Tweak comment.
8934 (hazard_nop): Likewise.
8936 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8938 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8939 -mspeculate-indirect-jumps.
8940 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8941 for -mno-speculate-indirect-jumps.
8942 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8943 (*call_value_indirect_elfv2<mode>): Disable for
8944 -mno-speculate-indirect-jumps.
8945 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8946 (indirect_jump): Emit different RTL for
8947 -mno-speculate-indirect-jumps.
8948 (*indirect_jump<mode>): Disable for
8949 -mno-speculate-indirect-jumps.
8950 (*indirect_jump<mode>_nospec): New define_insn.
8951 (tablejump): Emit different RTL for
8952 -mno-speculate-indirect-jumps.
8953 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8954 (tablejumpsi_nospec): New define_expand.
8955 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8956 (tablejumpdi_nospec): New define_expand.
8957 (*tablejump<mode>_internal1): Disable for
8958 -mno-speculate-indirect-jumps.
8959 (*tablejump<mode>_internal1_nospec): New define_insn.
8960 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8963 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8965 * caller-save.c (insert_save): Drop unnecessary parameter. All
8968 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8969 Richard Biener <rguenth@suse.de>
8972 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8973 return early, inline manually is_gimple_sizepos. Make sure if we
8974 call gimplify_expr we don't end up with a gimple constant.
8975 * tree.c (variably_modified_type_p): Don't return true for
8976 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8977 * gimplify.h (is_gimple_sizepos): Remove.
8979 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8981 PR tree-optimization/83857
8982 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8983 vectorizable_live_operation for pure SLP statements.
8984 (vectorizable_live_operation): Handle PHIs.
8986 2018-01-16 Richard Biener <rguenther@suse.de>
8988 PR tree-optimization/83867
8989 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8990 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8992 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8995 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8996 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8997 If off is not INTEGER_CST, issue a may not be aligned warning
8998 rather than isn't aligned. Use isn%'t rather than isn't.
8999 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9001 <case MULT_EXPR>: Improve the case when bottom and one of the
9002 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9003 operand, in that case check if the other operand is multiple of
9004 bottom divided by the INTEGER_CST operand.
9006 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9009 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9010 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9011 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9012 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9013 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9014 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9015 (pa_function_arg_size): New function.
9017 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9019 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9020 in a separate statement.
9022 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9024 PR tree-optimization/83847
9025 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9026 group gathers and scatters.
9028 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9030 PR rtl-optimization/86620
9031 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9033 PR rtl-optimization/83213
9034 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9035 to last if both are JUMP_INSNs.
9037 PR tree-optimization/83843
9038 * gimple-ssa-store-merging.c
9039 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9040 store_immediate_info for bswap/nop orig_stores.
9042 2018-01-15 Andrew Waterman <andrew@sifive.com>
9044 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9046 <UDIV>: Increase cost if !TARGET_DIV.
9048 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9050 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9051 (define_attr "cr_logical_3op"): New.
9052 (cceq_ior_compare): Adjust.
9053 (cceq_ior_compare_complement): Adjust.
9054 (*cceq_rev_compare): Adjust.
9055 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9056 (is_cracked_insn): Adjust.
9057 (insn_must_be_first_in_group): Adjust.
9058 * config/rs6000/40x.md: Adjust.
9059 * config/rs6000/440.md: Adjust.
9060 * config/rs6000/476.md: Adjust.
9061 * config/rs6000/601.md: Adjust.
9062 * config/rs6000/603.md: Adjust.
9063 * config/rs6000/6xx.md: Adjust.
9064 * config/rs6000/7450.md: Adjust.
9065 * config/rs6000/7xx.md: Adjust.
9066 * config/rs6000/8540.md: Adjust.
9067 * config/rs6000/cell.md: Adjust.
9068 * config/rs6000/e300c2c3.md: Adjust.
9069 * config/rs6000/e500mc.md: Adjust.
9070 * config/rs6000/e500mc64.md: Adjust.
9071 * config/rs6000/e5500.md: Adjust.
9072 * config/rs6000/e6500.md: Adjust.
9073 * config/rs6000/mpc.md: Adjust.
9074 * config/rs6000/power4.md: Adjust.
9075 * config/rs6000/power5.md: Adjust.
9076 * config/rs6000/power6.md: Adjust.
9077 * config/rs6000/power7.md: Adjust.
9078 * config/rs6000/power8.md: Adjust.
9079 * config/rs6000/power9.md: Adjust.
9080 * config/rs6000/rs64.md: Adjust.
9081 * config/rs6000/titan.md: Adjust.
9083 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9085 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9086 ix86_indirect_branch_register logic.
9088 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9090 * config/i386/constraints.md (Bs): Update
9091 ix86_indirect_branch_register check. Don't check
9092 ix86_indirect_branch_register with GOT_memory_operand.
9094 * config/i386/predicates.md (GOT_memory_operand): Don't check
9095 ix86_indirect_branch_register here.
9096 (GOT32_symbol_operand): Likewise.
9098 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9100 * config/i386/predicates.md (constant_call_address_operand):
9101 Rewrite ix86_indirect_branch_register logic.
9102 (sibcall_insn_operand): Likewise.
9104 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9106 * config/i386/constraints.md (Bs): Replace
9107 ix86_indirect_branch_thunk_register with
9108 ix86_indirect_branch_register.
9110 * config/i386/i386.md (indirect_jump): Likewise.
9111 (tablejump): Likewise.
9112 (*sibcall_memory): Likewise.
9113 (*sibcall_value_memory): Likewise.
9114 Peepholes of indirect call and jump via memory: Likewise.
9115 * config/i386/i386.opt: Likewise.
9116 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9117 (GOT_memory_operand): Likewise.
9118 (call_insn_operand): Likewise.
9119 (sibcall_insn_operand): Likewise.
9120 (GOT32_symbol_operand): Likewise.
9122 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9125 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9126 type rather than type addr's type points to.
9127 (expand_omp_atomic_mutex): Likewise.
9128 (expand_omp_atomic): Likewise.
9130 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9133 * config/i386/i386.c (output_indirect_thunk_function): Use
9134 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9135 for __x86_return_thunk.
9137 2018-01-15 Richard Biener <rguenther@suse.de>
9140 * expmed.c (extract_bit_field_1): Fix typo.
9142 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9145 * config/arm/iterators.md (VF): New mode iterator.
9146 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9147 Remove integer-related logic from pattern.
9148 (neon_vabd<mode>_3): Likewise.
9150 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9153 * common.opt (fstrict-overflow): No longer an alias.
9154 (fwrapv-pointer): New option.
9155 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9156 also for pointer types based on flag_wrapv_pointer.
9157 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9158 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9159 opts->x_flag_wrapv got set.
9160 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9161 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9162 POINTER_TYPE_OVERFLOW_UNDEFINED.
9163 * match.pd: Likewise in address comparison pattern.
9164 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9166 2018-01-15 Richard Biener <rguenther@suse.de>
9169 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9170 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9171 Reset type names to their identifier if their TYPE_DECL doesn't
9172 have linkage (and thus is used for ODR and devirt).
9173 (save_debug_info_for_decl): Remove.
9174 (save_debug_info_for_type): Likewise.
9175 (add_tree_to_fld_list): Adjust.
9176 * tree-pretty-print.c (dump_generic_node): Make dumping of
9177 type names more robust.
9179 2018-01-15 Richard Biener <rguenther@suse.de>
9181 * BASE-VER: Bump to 8.0.1.
9183 2018-01-14 Martin Sebor <msebor@redhat.com>
9186 * builtins.c (check_access): Avoid warning when the no-warning bit
9189 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9191 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9192 * ira-color (allocno_hard_regs_compare): Likewise.
9194 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9197 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9198 Use .pushsection/.popsection.
9200 2018-01-14 Martin Sebor <msebor@redhat.com>
9203 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9205 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9207 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9208 entry from extra_headers.
9209 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9210 extra_headers, make the list bitwise identical to the i?86-*-* one.
9212 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9214 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9215 -mcmodel=large with -mindirect-branch=thunk,
9216 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9217 -mfunction-return=thunk-extern.
9218 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9219 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9220 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9222 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9224 * config/i386/i386.c (print_reg): Print the name of the full
9225 integer register without '%'.
9226 (ix86_print_operand): Handle 'V'.
9227 * doc/extend.texi: Document 'V' modifier.
9229 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9231 * config/i386/constraints.md (Bs): Disallow memory operand for
9232 -mindirect-branch-register.
9234 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9235 (GOT_memory_operand): Likewise.
9236 (call_insn_operand): Likewise.
9237 (sibcall_insn_operand): Likewise.
9238 (GOT32_symbol_operand): Likewise.
9239 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9240 for -mindirect-branch-register.
9241 (tablejump): Likewise.
9242 (*sibcall_memory): Likewise.
9243 (*sibcall_value_memory): Likewise.
9244 Disallow peepholes of indirect call and jump via memory for
9245 -mindirect-branch-register.
9246 (*call_pop): Replace m with Bw.
9247 (*call_value_pop): Likewise.
9248 (*sibcall_pop_memory): Replace m with Bs.
9249 * config/i386/i386.opt (mindirect-branch-register): New option.
9250 * doc/invoke.texi: Document -mindirect-branch-register option.
9252 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9254 * config/i386/i386-protos.h (ix86_output_function_return): New.
9255 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9256 set function_return_type.
9257 (indirect_thunk_name): Add ret_p to indicate thunk for function
9259 (output_indirect_thunk_function): Pass false to
9260 indirect_thunk_name.
9261 (ix86_output_indirect_branch_via_reg): Likewise.
9262 (ix86_output_indirect_branch_via_push): Likewise.
9263 (output_indirect_thunk_function): Create alias for function
9264 return thunk if regno < 0.
9265 (ix86_output_function_return): New function.
9266 (ix86_handle_fndecl_attribute): Handle function_return.
9267 (ix86_attribute_table): Add function_return.
9268 * config/i386/i386.h (machine_function): Add
9269 function_return_type.
9270 * config/i386/i386.md (simple_return_internal): Use
9271 ix86_output_function_return.
9272 (simple_return_internal_long): Likewise.
9273 * config/i386/i386.opt (mfunction-return=): New option.
9274 (indirect_branch): Mention -mfunction-return=.
9275 * doc/extend.texi: Document function_return function attribute.
9276 * doc/invoke.texi: Document -mfunction-return= option.
9278 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9280 * config/i386/i386-opts.h (indirect_branch): New.
9281 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9282 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9283 with local indirect jump when converting indirect call and jump.
9284 (ix86_set_indirect_branch_type): New.
9285 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9286 (indirectlabelno): New.
9287 (indirect_thunk_needed): Likewise.
9288 (indirect_thunk_bnd_needed): Likewise.
9289 (indirect_thunks_used): Likewise.
9290 (indirect_thunks_bnd_used): Likewise.
9291 (INDIRECT_LABEL): Likewise.
9292 (indirect_thunk_name): Likewise.
9293 (output_indirect_thunk): Likewise.
9294 (output_indirect_thunk_function): Likewise.
9295 (ix86_output_indirect_branch_via_reg): Likewise.
9296 (ix86_output_indirect_branch_via_push): Likewise.
9297 (ix86_output_indirect_branch): Likewise.
9298 (ix86_output_indirect_jmp): Likewise.
9299 (ix86_code_end): Call output_indirect_thunk_function if needed.
9300 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9302 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9303 (ix86_attribute_table): Add indirect_branch.
9304 * config/i386/i386.h (machine_function): Add indirect_branch_type
9305 and has_local_indirect_jump.
9306 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9308 (tablejump): Likewise.
9309 (*indirect_jump): Use ix86_output_indirect_jmp.
9310 (*tablejump_1): Likewise.
9311 (simple_return_indirect_internal): Likewise.
9312 * config/i386/i386.opt (mindirect-branch=): New option.
9313 (indirect_branch): New.
9316 (thunk-inline): Likewise.
9317 (thunk-extern): Likewise.
9318 * doc/extend.texi: Document indirect_branch function attribute.
9319 * doc/invoke.texi: Document -mindirect-branch= option.
9321 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9324 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9326 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9328 * ipa-inline.c (want_inline_small_function_p): Return false if
9329 inlining has already failed with CIF_FINAL_ERROR.
9330 (update_caller_keys): Call want_inline_small_function_p before
9332 (update_callee_keys): Likewise.
9334 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9336 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9338 (rs6000_quadword_masked_address_p): Likewise.
9339 (quad_aligned_load_p): Likewise.
9340 (quad_aligned_store_p): Likewise.
9341 (const_load_sequence_p): Add comment to describe the outer-most loop.
9342 (mimic_memory_attributes_and_flags): New function.
9343 (rs6000_gen_stvx): Likewise.
9344 (replace_swapped_aligned_store): Likewise.
9345 (rs6000_gen_lvx): Likewise.
9346 (replace_swapped_aligned_load): Likewise.
9347 (replace_swapped_load_constant): Capitalize argument name in
9348 comment describing this function.
9349 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9350 and stores that access quad-word aligned addresses and replace
9351 with stvx or lvx instructions when appropriate.
9352 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9353 New function prototype.
9354 (rs6000_quadword_masked_address_p): Likewise.
9355 (rs6000_gen_lvx): Likewise.
9356 (rs6000_gen_stvx): Likewise.
9357 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9358 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9359 when memory address is aligned.
9360 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9361 this split to select lvx instruction when memory address is aligned.
9362 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9363 instruction when memory address is aligned.
9364 (*vsx_le_perm_load_v16qi): Likewise.
9365 (four unnamed splitters): Modify to select the stvx instruction
9366 when memory is aligned.
9368 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9370 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9371 which appears in the queue multiple times.
9373 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9374 Alan Hayward <alan.hayward@arm.com>
9375 David Sherwood <david.sherwood@arm.com>
9377 * tree-vectorizer.h (vec_lower_bound): New structure.
9378 (_loop_vec_info): Add check_nonzero and lower_bounds.
9379 (LOOP_VINFO_CHECK_NONZERO): New macro.
9380 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9381 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9382 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9383 fields. Make seg_len the distance travelled, not including the
9385 (dr_direction_indicator): Declare.
9386 (dr_zero_step_indicator): Likewise.
9387 (dr_known_forward_stride_p): Likewise.
9388 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9390 (runtime_alias_check_p): Allow runtime alias checks with
9392 (operator ==): Compare access_size and align.
9393 (prune_runtime_alias_test_list): Rework for new distinction between
9394 the access_size and seg_len.
9395 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9397 (get_segment_min_max): New function.
9398 (create_intersect_range_checks): Use it.
9399 (dr_step_indicator): New function.
9400 (dr_direction_indicator): Likewise.
9401 (dr_zero_step_indicator): Likewise.
9402 (dr_known_forward_stride_p): Likewise.
9403 * tree-loop-distribution.c (data_ref_segment_size): Return
9404 DR_STEP * (niters - 1).
9405 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9407 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9408 (vect_preserves_scalar_order_p): New function, split out from...
9409 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9410 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9411 (vect_vfa_access_size): New function.
9412 (vect_vfa_align): Likewise.
9413 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9414 (dump_lower_bound): New function.
9415 (vect_check_lower_bound): Likewise.
9416 (vect_small_gap_p): Likewise.
9417 (vectorizable_with_step_bound_p): Likewise.
9418 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9419 depencies if the vectorization factor is 1. Convert the checks
9420 for nonzero steps into checks on the bounds of DR_STEP. Try using
9421 a bunds check for variable steps if the minimum required step is
9422 relatively small. Update calls to the dr_with_seg_len
9423 constructor and to vect_compile_time_alias.
9424 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9426 (vect_loop_versioning): Call it.
9427 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9429 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9431 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9432 Alan Hayward <alan.hayward@arm.com>
9433 David Sherwood <david.sherwood@arm.com>
9435 * doc/sourcebuild.texi (vect_scatter_store): Document.
9436 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9438 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9440 * genopinit.c (main): Add supports_vec_scatter_store and
9441 supports_vec_scatter_store_cached to target_optabs.
9442 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9443 IFN_MASK_SCATTER_STORE.
9444 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9446 * internal-fn.h (internal_store_fn_p): Declare.
9447 (internal_fn_stored_value_index): Likewise.
9448 * internal-fn.c (scatter_store_direct): New macro.
9449 (expand_scatter_store_optab_fn): New function.
9450 (direct_scatter_store_optab_supported_p): New macro.
9451 (internal_store_fn_p): New function.
9452 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9453 IFN_MASK_SCATTER_STORE.
9454 (internal_fn_mask_index): Likewise.
9455 (internal_fn_stored_value_index): New function.
9456 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9458 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9459 * optabs-query.c (supports_vec_scatter_store_p): New function.
9460 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9461 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9462 true for scatter stores.
9463 (vect_gather_scatter_fn_p): Handle scatter stores too.
9464 (vect_check_gather_scatter): Consider using scatter stores if
9465 supports_vec_scatter_store_p.
9466 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9468 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9469 internal_fn_stored_value_index.
9470 (check_load_store_masking): Handle scatter stores too.
9471 (vect_get_store_rhs): Make public.
9472 (vectorizable_call): Use internal_store_fn_p.
9473 (vectorizable_store): Handle scatter store internal functions.
9474 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9475 when deciding whether the end of the group has been reached.
9476 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9477 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9478 (mask_scatter_store<mode>): New insns.
9480 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9481 Alan Hayward <alan.hayward@arm.com>
9482 David Sherwood <david.sherwood@arm.com>
9484 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9485 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9486 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9488 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9489 Use vect_truncate_gather_scatter_offset if we can't treat the
9490 operation as a normal gather load or scatter store.
9491 (get_group_load_store_type): Take the gather_scatter_info
9492 as argument. Try using a gather load or scatter store for
9493 single-element groups.
9494 (get_load_store_type): Update calls to get_group_load_store_type
9495 and vect_use_strided_gather_scatters_p.
9497 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9498 Alan Hayward <alan.hayward@arm.com>
9499 David Sherwood <david.sherwood@arm.com>
9501 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9502 optional tree argument.
9503 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9505 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9506 but continue to use the current value as a fallback.
9507 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9508 to compare the updates.
9509 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9510 (get_load_store_type): Use it when handling a strided access.
9511 (vect_get_strided_load_store_ops): New function.
9512 (vect_get_data_ptr_increment): Likewise.
9513 (vectorizable_load): Handle strided gather loads. Always pass
9514 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9516 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9517 Alan Hayward <alan.hayward@arm.com>
9518 David Sherwood <david.sherwood@arm.com>
9520 * doc/md.texi (gather_load@var{m}): Document.
9521 (mask_gather_load@var{m}): Likewise.
9522 * genopinit.c (main): Add supports_vec_gather_load and
9523 supports_vec_gather_load_cached to target_optabs.
9524 * optabs-tree.c (init_tree_optimization_optabs): Use
9525 ggc_cleared_alloc to allocate target_optabs.
9526 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9527 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9529 * internal-fn.h (internal_load_fn_p): Declare.
9530 (internal_gather_scatter_fn_p): Likewise.
9531 (internal_fn_mask_index): Likewise.
9532 (internal_gather_scatter_fn_supported_p): Likewise.
9533 * internal-fn.c (gather_load_direct): New macro.
9534 (expand_gather_load_optab_fn): New function.
9535 (direct_gather_load_optab_supported_p): New macro.
9536 (direct_internal_fn_optab): New function.
9537 (internal_load_fn_p): Likewise.
9538 (internal_gather_scatter_fn_p): Likewise.
9539 (internal_fn_mask_index): Likewise.
9540 (internal_gather_scatter_fn_supported_p): Likewise.
9541 * optabs-query.c (supports_at_least_one_mode_p): New function.
9542 (supports_vec_gather_load_p): Likewise.
9543 * optabs-query.h (supports_vec_gather_load_p): Declare.
9544 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9545 and memory_type field.
9546 (NUM_PATTERNS): Bump to 15.
9547 * tree-vect-data-refs.c: Include internal-fn.h.
9548 (vect_gather_scatter_fn_p): New function.
9549 (vect_describe_gather_scatter_call): Likewise.
9550 (vect_check_gather_scatter): Try using internal functions for
9551 gather loads. Recognize existing calls to a gather load function.
9552 (vect_analyze_data_refs): Consider using gather loads if
9553 supports_vec_gather_load_p.
9554 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9555 (vect_get_gather_scatter_offset_type): Likewise.
9556 (vect_convert_mask_for_vectype): Likewise.
9557 (vect_add_conversion_to_patterm): Likewise.
9558 (vect_try_gather_scatter_pattern): Likewise.
9559 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9560 (vect_vect_recog_func_ptrs): Add it.
9561 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9562 internal_fn_mask_index and internal_gather_scatter_fn_p.
9563 (check_load_store_masking): Take the gather_scatter_info as an
9564 argument and handle gather loads.
9565 (vect_get_gather_scatter_ops): New function.
9566 (vectorizable_call): Check internal_load_fn_p.
9567 (vectorizable_load): Likewise. Handle gather load internal
9569 (vectorizable_store): Update call to check_load_store_masking.
9570 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9571 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9572 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9573 (aarch64_gather_scale_operand_d): New predicates.
9574 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9575 (mask_gather_load<mode>): New insns.
9577 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9578 Alan Hayward <alan.hayward@arm.com>
9579 David Sherwood <david.sherwood@arm.com>
9581 * optabs.def (fold_left_plus_optab): New optab.
9582 * doc/md.texi (fold_left_plus_@var{m}): Document.
9583 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9584 * internal-fn.c (fold_left_direct): Define.
9585 (expand_fold_left_optab_fn): Likewise.
9586 (direct_fold_left_optab_supported_p): Likewise.
9587 * fold-const-call.c (fold_const_fold_left): New function.
9588 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9589 * tree-parloops.c (valid_reduction_p): New function.
9590 (gather_scalar_reductions): Use it.
9591 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9592 (vect_finish_replace_stmt): Declare.
9593 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9594 (needs_fold_left_reduction_p): New function, split out from...
9595 (vect_is_simple_reduction): ...here. Accept reductions that
9596 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9597 (vect_force_simple_reduction): Also store the reduction type in
9598 the assignment's STMT_VINFO_REDUC_TYPE.
9599 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9600 (merge_with_identity): New function.
9601 (vect_expand_fold_left): Likewise.
9602 (vectorize_fold_left_reduction): Likewise.
9603 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9604 scalar phi in place for it. Check for target support and reject
9605 cases that would reassociate the operation. Defer the transform
9606 phase to vectorize_fold_left_reduction.
9607 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9608 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9609 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9611 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9613 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9614 call to ifc_temp_var.
9616 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9617 Alan Hayward <alan.hayward@arm.com>
9618 David Sherwood <david.sherwood@arm.com>
9620 * target.def (legitimize_address_displacement): Take the original
9621 offset as a poly_int.
9622 * targhooks.h (default_legitimize_address_displacement): Update
9624 * targhooks.c (default_legitimize_address_displacement): Likewise.
9625 * doc/tm.texi: Regenerate.
9626 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9627 as an argument, moving assert of ad->disp == ad->disp_term to...
9628 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9629 Try calling targetm.legitimize_address_displacement before expanding
9630 the address rather than afterwards, and adjust for the new interface.
9631 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9632 Match the new hook interface. Handle SVE addresses.
9633 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9636 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9638 * Makefile.in (OBJS): Add early-remat.o.
9639 * target.def (select_early_remat_modes): New hook.
9640 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9641 * doc/tm.texi: Regenerate.
9642 * targhooks.h (default_select_early_remat_modes): Declare.
9643 * targhooks.c (default_select_early_remat_modes): New function.
9644 * timevar.def (TV_EARLY_REMAT): New timevar.
9645 * passes.def (pass_early_remat): New pass.
9646 * tree-pass.h (make_pass_early_remat): Declare.
9647 * early-remat.c: New file.
9648 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9650 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9652 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9653 Alan Hayward <alan.hayward@arm.com>
9654 David Sherwood <david.sherwood@arm.com>
9656 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9657 vfm1 with a bound_epilog parameter.
9658 (vect_do_peeling): Update calls accordingly, and move the prologue
9659 call earlier in the function. Treat the base bound_epilog as 0 for
9660 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9661 this base when peeling for gaps.
9662 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9663 with fully-masked loops.
9664 (vect_estimate_min_profitable_iters): Handle the single peeled
9665 iteration in that case.
9667 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9668 Alan Hayward <alan.hayward@arm.com>
9669 David Sherwood <david.sherwood@arm.com>
9671 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9672 single-element interleaving even if the size is not a power of 2.
9673 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9674 accesses for single-element interleaving if the group size is
9677 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9678 Alan Hayward <alan.hayward@arm.com>
9679 David Sherwood <david.sherwood@arm.com>
9681 * doc/md.texi (fold_extract_last_@var{m}): Document.
9682 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9683 * optabs.def (fold_extract_last_optab): New optab.
9684 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9685 * internal-fn.c (fold_extract_direct): New macro.
9686 (expand_fold_extract_optab_fn): Likewise.
9687 (direct_fold_extract_optab_supported_p): Likewise.
9688 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9689 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9690 EXTRACT_LAST_REDUCTION.
9691 (get_initial_def_for_reduction): Do not create an initial vector
9692 for EXTRACT_LAST_REDUCTION reductions.
9693 (vectorizable_reduction): Leave the scalar phi in place for
9694 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9695 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9696 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9697 transform phase to vectorizable_condition.
9698 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9700 (vect_finish_stmt_generation): ...here.
9701 (vect_finish_replace_stmt): New function.
9702 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9703 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9705 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9707 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9708 Alan Hayward <alan.hayward@arm.com>
9709 David Sherwood <david.sherwood@arm.com>
9711 * doc/md.texi (extract_last_@var{m}): Document.
9712 * optabs.def (extract_last_optab): New optab.
9713 * internal-fn.def (EXTRACT_LAST): New internal function.
9714 * internal-fn.c (cond_unary_direct): New macro.
9715 (expand_cond_unary_optab_fn): Likewise.
9716 (direct_cond_unary_optab_supported_p): Likewise.
9717 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9718 loops using EXTRACT_LAST.
9719 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9720 (extract_last_<mode>): ...this optab.
9721 (vec_extract<mode><Vel>): Update accordingly.
9723 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9724 Alan Hayward <alan.hayward@arm.com>
9725 David Sherwood <david.sherwood@arm.com>
9727 * target.def (empty_mask_is_expensive): New hook.
9728 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9729 * doc/tm.texi: Regenerate.
9730 * targhooks.h (default_empty_mask_is_expensive): Declare.
9731 * targhooks.c (default_empty_mask_is_expensive): New function.
9732 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9733 if the target says that empty masks are expensive.
9734 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9736 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9738 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9739 Alan Hayward <alan.hayward@arm.com>
9740 David Sherwood <david.sherwood@arm.com>
9742 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9743 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9744 (vect_use_loop_mask_for_alignment_p): New function.
9745 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9746 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9747 niters_skip argument. Make sure that the first niters_skip elements
9748 of the first iteration are inactive.
9749 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9750 Update call to vect_set_loop_masks_directly.
9751 (get_misalign_in_elems): New function, split out from...
9752 (vect_gen_prolog_loop_niters): ...here.
9753 (vect_update_init_of_dr): Take a code argument that specifies whether
9754 the adjustment should be added or subtracted.
9755 (vect_update_init_of_drs): Likewise.
9756 (vect_prepare_for_masked_peels): New function.
9757 (vect_do_peeling): Skip prologue peeling if we're using a mask
9758 instead. Update call to vect_update_inits_of_drs.
9759 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9761 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9762 alignment. Do not include the number of peeled iterations in
9763 the minimum threshold in that case.
9764 (vectorizable_induction): Adjust the start value down by
9765 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9766 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9767 Take the number of skipped iterations into account when calculating
9769 * tree-vect-stmts.c (vect_gen_while_not): New function.
9771 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9772 Alan Hayward <alan.hayward@arm.com>
9773 David Sherwood <david.sherwood@arm.com>
9775 * doc/sourcebuild.texi (vect_fully_masked): Document.
9776 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9778 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9780 (vect_analyze_loop_2): ...here. Don't check the vectorization
9781 factor against the number of loop iterations if the loop is
9784 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9785 Alan Hayward <alan.hayward@arm.com>
9786 David Sherwood <david.sherwood@arm.com>
9788 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9789 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9790 (dump_groups): Update accordingly.
9791 (iv_use::mem_type): New member variable.
9792 (address_p): New function.
9793 (record_use): Add a mem_type argument and initialize the new
9795 (record_group_use): Add a mem_type argument. Use address_p.
9796 Remove obsolete null checks of base_object. Update call to record_use.
9797 (find_interesting_uses_op): Update call to record_group_use.
9798 (find_interesting_uses_cond): Likewise.
9799 (find_interesting_uses_address): Likewise.
9800 (get_mem_type_for_internal_fn): New function.
9801 (find_address_like_use): Likewise.
9802 (find_interesting_uses_stmt): Try find_address_like_use before
9803 calling find_interesting_uses_op.
9804 (addr_offset_valid_p): Use the iv mem_type field as the type
9805 of the addressed memory.
9806 (add_autoinc_candidates): Likewise.
9807 (get_address_cost): Likewise.
9808 (split_small_address_groups_p): Use address_p.
9809 (split_address_groups): Likewise.
9810 (add_iv_candidate_for_use): Likewise.
9811 (autoinc_possible_for_pair): Likewise.
9812 (rewrite_groups): Likewise.
9813 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9814 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9815 (get_alias_ptr_type_for_ptr_address): New function.
9816 (rewrite_use_address): Rewrite address uses in calls that were
9817 identified by find_address_like_use.
9819 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9820 Alan Hayward <alan.hayward@arm.com>
9821 David Sherwood <david.sherwood@arm.com>
9823 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9825 * gimple-expr.h (is_gimple_addressable: Likewise.
9826 * gimple-expr.c (is_gimple_address): Likewise.
9827 * internal-fn.c (expand_call_mem_ref): New function.
9828 (expand_mask_load_optab_fn): Use it.
9829 (expand_mask_store_optab_fn): Likewise.
9831 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9832 Alan Hayward <alan.hayward@arm.com>
9833 David Sherwood <david.sherwood@arm.com>
9835 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9836 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9837 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9838 (cond_umax@var{mode}): Document.
9839 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9840 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9841 (cond_umin_optab, cond_umax_optab): New optabs.
9842 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9843 (COND_IOR, COND_XOR): New internal functions.
9844 * internal-fn.h (get_conditional_internal_fn): Declare.
9845 * internal-fn.c (cond_binary_direct): New macro.
9846 (expand_cond_binary_optab_fn): Likewise.
9847 (direct_cond_binary_optab_supported_p): Likewise.
9848 (get_conditional_internal_fn): New function.
9849 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9850 Cope with reduction statements that are vectorized as calls rather
9852 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9853 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9854 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9855 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9856 (UNSPEC_COND_EOR): New unspecs.
9857 (optab): Add mappings for them.
9858 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9859 (sve_int_op, sve_fp_op): New int attributes.
9861 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9862 Alan Hayward <alan.hayward@arm.com>
9863 David Sherwood <david.sherwood@arm.com>
9865 * optabs.def (while_ult_optab): New optab.
9866 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9867 * internal-fn.def (WHILE_ULT): New internal function.
9868 * internal-fn.h (direct_internal_fn_supported_p): New override
9869 that takes two types as argument.
9870 * internal-fn.c (while_direct): New macro.
9871 (expand_while_optab_fn): New function.
9872 (convert_optab_supported_p): Likewise.
9873 (direct_while_optab_supported_p): New macro.
9874 * wide-int.h (wi::udiv_ceil): New function.
9875 * tree-vectorizer.h (rgroup_masks): New structure.
9876 (vec_loop_masks): New typedef.
9877 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9879 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9880 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9881 (vect_max_vf): New function.
9882 (slpeel_make_loop_iterate_ntimes): Delete.
9883 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9884 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9885 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9886 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9887 internal-fn.h, stor-layout.h and optabs-query.h.
9888 (vect_set_loop_mask): New function.
9889 (add_preheader_seq): Likewise.
9890 (add_header_seq): Likewise.
9891 (interleave_supported_p): Likewise.
9892 (vect_maybe_permute_loop_masks): Likewise.
9893 (vect_set_loop_masks_directly): Likewise.
9894 (vect_set_loop_condition_masked): Likewise.
9895 (vect_set_loop_condition_unmasked): New function, split out from
9896 slpeel_make_loop_iterate_ntimes.
9897 (slpeel_make_loop_iterate_ntimes): Rename to..
9898 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9899 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9900 (vect_do_peeling): Update call accordingly.
9901 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9903 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9904 mask_compare_type, can_fully_mask_p and fully_masked_p.
9905 (release_vec_loop_masks): New function.
9906 (_loop_vec_info): Use it to free the loop masks.
9907 (can_produce_all_loop_masks_p): New function.
9908 (vect_get_max_nscalars_per_iter): Likewise.
9909 (vect_verify_full_masking): Likewise.
9910 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9911 retries, and free the mask rgroups before retrying. Check loop-wide
9912 reasons for disallowing fully-masked loops. Make the final decision
9913 about whether use a fully-masked loop or not.
9914 (vect_estimate_min_profitable_iters): Do not assume that peeling
9915 for the number of iterations will be needed for fully-masked loops.
9916 (vectorizable_reduction): Disable fully-masked loops.
9917 (vectorizable_live_operation): Likewise.
9918 (vect_halve_mask_nunits): New function.
9919 (vect_double_mask_nunits): Likewise.
9920 (vect_record_loop_mask): Likewise.
9921 (vect_get_loop_mask): Likewise.
9922 (vect_transform_loop): Handle the case in which the final loop
9923 iteration might handle a partial vector. Call vect_set_loop_condition
9924 instead of slpeel_make_loop_iterate_ntimes.
9925 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9926 (check_load_store_masking): New function.
9927 (prepare_load_store_mask): Likewise.
9928 (vectorizable_store): Handle fully-masked loops.
9929 (vectorizable_load): Likewise.
9930 (supportable_widening_operation): Use vect_halve_mask_nunits for
9932 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9933 (vect_gen_while): New function.
9934 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9935 (aarch64_uqdec<mode>): New insn.
9937 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9938 Alan Hayward <alan.hayward@arm.com>
9939 David Sherwood <david.sherwood@arm.com>
9941 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9942 (reduc_xor_scal_optab): New optabs.
9943 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9944 (reduc_xor_scal_@var{m}): Document.
9945 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9946 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9948 * fold-const-call.c (fold_const_call): Handle them.
9949 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9950 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9951 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9952 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9953 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9954 (UNSPEC_XORV): New unspecs.
9955 (optab): Add entries for them.
9956 (BITWISEV): New int iterator.
9957 (bit_reduc_op): New int attributes.
9959 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9960 Alan Hayward <alan.hayward@arm.com>
9961 David Sherwood <david.sherwood@arm.com>
9963 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9964 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9965 * optabs.def (vec_shl_insert_optab): New optab.
9966 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9967 (duplicate_and_interleave): Likewise.
9968 * tree-vect-loop.c: Include internal-fn.h.
9969 (neutral_op_for_slp_reduction): New function, split out from
9970 get_initial_defs_for_reduction.
9971 (get_initial_def_for_reduction): Handle option 2 for variable-length
9972 vectors by loading the neutral value into a vector and then shifting
9973 the initial value into element 0.
9974 (get_initial_defs_for_reduction): Replace the code argument with
9975 the neutral value calculated by neutral_op_for_slp_reduction.
9976 Use gimple_build_vector for constant-length vectors.
9977 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9978 but the first group_size elements have a neutral value.
9979 Use duplicate_and_interleave otherwise.
9980 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9981 Update call to get_initial_defs_for_reduction. Handle SLP
9982 reductions for variable-length vectors by creating one vector
9983 result for each scalar result, with the elements associated
9984 with other scalar results stubbed out with the neutral value.
9985 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9986 Require IFN_VEC_SHL_INSERT for double reductions on
9987 variable-length vectors, or SLP reductions that have
9988 a neutral value. Require can_duplicate_and_interleave_p
9989 support for variable-length unchained SLP reductions if there
9990 is no neutral value, such as for MIN/MAX reductions. Also require
9991 the number of vector elements to be a multiple of the number of
9992 SLP statements when doing variable-length unchained SLP reductions.
9993 Update call to vect_create_epilog_for_reduction.
9994 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9995 and remove initial values.
9996 (duplicate_and_interleave): Make public.
9997 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9998 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10000 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10001 Alan Hayward <alan.hayward@arm.com>
10002 David Sherwood <david.sherwood@arm.com>
10004 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10005 (can_duplicate_and_interleave_p): New function.
10006 (vect_get_and_check_slp_defs): Take the vector of statements
10007 rather than just the current one. Remove excess parentheses.
10008 Restriction rejectinon of vect_constant_def and vect_external_def
10009 for variable-length vectors to boolean types, or types for which
10010 can_duplicate_and_interleave_p is false.
10011 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10012 (duplicate_and_interleave): New function.
10013 (vect_get_constant_vectors): Use gimple_build_vector for
10014 constant-length vectors and suitable variable-length constant
10015 vectors. Use duplicate_and_interleave for other variable-length
10016 vectors. Don't defer the update when inserting new statements.
10018 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10019 Alan Hayward <alan.hayward@arm.com>
10020 David Sherwood <david.sherwood@arm.com>
10022 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10023 min_profitable_iters doesn't go negative.
10025 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10026 Alan Hayward <alan.hayward@arm.com>
10027 David Sherwood <david.sherwood@arm.com>
10029 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10030 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10031 * optabs.def (vec_mask_load_lanes_optab): New optab.
10032 (vec_mask_store_lanes_optab): Likewise.
10033 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10034 (MASK_STORE_LANES): Likewise.
10035 * internal-fn.c (mask_load_lanes_direct): New macro.
10036 (mask_store_lanes_direct): Likewise.
10037 (expand_mask_load_optab_fn): Handle masked operations.
10038 (expand_mask_load_lanes_optab_fn): New macro.
10039 (expand_mask_store_optab_fn): Handle masked operations.
10040 (expand_mask_store_lanes_optab_fn): New macro.
10041 (direct_mask_load_lanes_optab_supported_p): Likewise.
10042 (direct_mask_store_lanes_optab_supported_p): Likewise.
10043 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10045 (vect_load_lanes_supported): Likewise.
10046 * tree-vect-data-refs.c (strip_conversion): New function.
10047 (can_group_stmts_p): Likewise.
10048 (vect_analyze_data_ref_accesses): Use it instead of checking
10049 for a pair of assignments.
10050 (vect_store_lanes_supported): Take a masked_p parameter.
10051 (vect_load_lanes_supported): Likewise.
10052 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10053 vect_store_lanes_supported and vect_load_lanes_supported.
10054 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10055 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10056 parameter. Don't allow gaps for masked accesses.
10057 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10058 and vect_load_lanes_supported.
10059 (get_load_store_type): Take a masked_p parameter and update
10060 call to get_group_load_store_type.
10061 (vectorizable_store): Update call to get_load_store_type.
10062 Handle IFN_MASK_STORE_LANES.
10063 (vectorizable_load): Update call to get_load_store_type.
10064 Handle IFN_MASK_LOAD_LANES.
10066 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10067 Alan Hayward <alan.hayward@arm.com>
10068 David Sherwood <david.sherwood@arm.com>
10070 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10072 * config/aarch64/aarch64-protos.h
10073 (aarch64_sve_struct_memory_operand_p): Declare.
10074 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10075 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10076 (VPRED, vpred): Handle SVE structure modes.
10077 * config/aarch64/constraints.md (Utx): New constraint.
10078 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10079 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10080 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10081 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10082 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10083 structure modes. Split into pieces after RA.
10084 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10085 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10087 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10088 SVE structure modes.
10089 (aarch64_classify_address): Likewise.
10090 (sizetochar): Move earlier in file.
10091 (aarch64_print_operand): Handle SVE register lists.
10092 (aarch64_array_mode): New function.
10093 (aarch64_sve_struct_memory_operand_p): Likewise.
10094 (TARGET_ARRAY_MODE): Redefine.
10096 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10097 Alan Hayward <alan.hayward@arm.com>
10098 David Sherwood <david.sherwood@arm.com>
10100 * target.def (array_mode): New target hook.
10101 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10102 * doc/tm.texi: Regenerate.
10103 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10104 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10105 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10106 targetm.array_mode.
10107 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10110 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10111 Alan Hayward <alan.hayward@arm.com>
10112 David Sherwood <david.sherwood@arm.com>
10114 * fold-const.c (fold_binary_loc): Check the argument types
10115 rather than the result type when testing for a vector operation.
10117 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10119 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10120 * doc/tm.texi: Regenerate.
10122 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10123 Alan Hayward <alan.hayward@arm.com>
10124 David Sherwood <david.sherwood@arm.com>
10126 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10127 (sve): Document new AArch64 extension.
10128 * doc/md.texi (w): Extend the description of the AArch64
10129 constraint to include SVE vectors.
10130 (Upl, Upa): Document new AArch64 predicate constraints.
10131 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10133 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10134 (msve-vector-bits=): New option.
10135 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10136 SVE when these are disabled.
10137 (sve): New extension.
10138 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10139 modes. Adjust their number of units based on aarch64_sve_vg.
10140 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10141 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10142 aarch64_addr_query_type.
10143 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10144 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10145 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10146 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10147 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10148 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10149 (aarch64_simd_imm_zero_p): Delete.
10150 (aarch64_check_zero_based_sve_index_immediate): Declare.
10151 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10152 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10153 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10154 (aarch64_sve_float_mul_immediate_p): Likewise.
10155 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10156 rather than an rtx.
10157 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10158 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10159 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10160 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10161 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10162 (aarch64_regmode_natural_size): Likewise.
10163 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10164 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10166 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10167 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10168 for VG and the SVE predicate registers.
10169 (V_ALIASES): Add a "z"-prefixed alias.
10170 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10171 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10172 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10173 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10174 (REG_CLASS_NAMES): Add entries for them.
10175 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10176 and the predicate registers.
10177 (aarch64_sve_vg): Declare.
10178 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10179 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10180 (REGMODE_NATURAL_SIZE): Define.
10181 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10183 * config/aarch64/aarch64.c: Include cfgrtl.h.
10184 (simd_immediate_info): Add a constructor for series vectors,
10185 and an associated step field.
10186 (aarch64_sve_vg): New variable.
10187 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10188 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10189 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10190 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10191 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10192 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10193 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10194 (aarch64_get_mask_mode): New functions.
10195 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10196 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10197 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10198 predicate modes and predicate registers. Explicitly restrict
10199 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10200 to store a vector mode if it is recognized by
10201 aarch64_classify_vector_mode.
10202 (aarch64_regmode_natural_size): New function.
10203 (aarch64_hard_regno_caller_save_mode): Return the original mode
10205 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10206 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10207 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10208 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10210 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10211 does not overlap dest if the function is frame-related. Handle
10213 (aarch64_split_add_offset): New function.
10214 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10215 them aarch64_add_offset.
10216 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10217 and update call to aarch64_sub_sp.
10218 (aarch64_add_cfa_expression): New function.
10219 (aarch64_expand_prologue): Pass extra temporary registers to the
10220 functions above. Handle the case in which we need to emit new
10221 DW_CFA_expressions for registers that were originally saved
10222 relative to the stack pointer, but now have to be expressed
10223 relative to the frame pointer.
10224 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10226 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10227 IP0 and IP1 values for SVE frames.
10228 (aarch64_expand_vec_series): New function.
10229 (aarch64_expand_sve_widened_duplicate): Likewise.
10230 (aarch64_expand_sve_const_vector): Likewise.
10231 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10232 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10233 into the register, rather than emitting a SET directly.
10234 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10235 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10236 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10237 (offset_9bit_signed_scaled_p): New functions.
10238 (aarch64_replicate_bitmask_imm): New function.
10239 (aarch64_bitmask_imm): Use it.
10240 (aarch64_cannot_force_const_mem): Reject expressions involving
10241 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10242 (aarch64_classify_index): Handle SVE indices, by requiring
10243 a plain register index with a scale that matches the element size.
10244 (aarch64_classify_address): Handle SVE addresses. Assert that
10245 the mode of the address is VOIDmode or an integer mode.
10246 Update call to aarch64_classify_symbol.
10247 (aarch64_classify_symbolic_expression): Update call to
10248 aarch64_classify_symbol.
10249 (aarch64_const_vec_all_in_range_p): New function.
10250 (aarch64_print_vector_float_operand): Likewise.
10251 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10252 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10253 and the FP immediates 1.0 and 0.5.
10254 (aarch64_print_address_internal): Handle SVE addresses.
10255 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10256 (aarch64_regno_regclass): Handle predicate registers.
10257 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10259 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10260 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10261 (aarch64_convert_sve_vector_bits): New function.
10262 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10263 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10264 rather than an rtx.
10265 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10266 Handle SVE vector and predicate modes. Accept VL-based constants
10267 that need only one temporary register, and VL offsets that require
10268 no temporary registers.
10269 (aarch64_conditional_register_usage): Mark the predicate registers
10270 as fixed if SVE isn't available.
10271 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10272 Return true for SVE vector and predicate modes.
10273 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10274 rather than an unsigned int. Handle SVE modes.
10275 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10277 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10279 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10280 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10281 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10282 (aarch64_sve_float_mul_immediate_p): New functions.
10283 (aarch64_sve_valid_immediate): New function.
10284 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10285 Explicitly reject structure modes. Check for INDEX constants.
10286 Handle PTRUE and PFALSE constants.
10287 (aarch64_check_zero_based_sve_index_immediate): New function.
10288 (aarch64_simd_imm_zero_p): Delete.
10289 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10290 vector modes. Accept constants in the range of CNT[BHWD].
10291 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10292 ask for an Advanced SIMD mode.
10293 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10294 (aarch64_simd_vector_alignment): Handle SVE predicates.
10295 (aarch64_vectorize_preferred_vector_alignment): New function.
10296 (aarch64_simd_vector_alignment_reachable): Use it instead of
10298 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10299 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10301 (MAX_VECT_LEN): Delete.
10302 (expand_vec_perm_d): Add a vec_flags field.
10303 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10304 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10305 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10307 (aarch64_evpc_rev): Rename to...
10308 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10309 (aarch64_evpc_rev_global): New function.
10310 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10311 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10313 (aarch64_evpc_sve_tbl): New function.
10314 (aarch64_expand_vec_perm_const_1): Update after rename of
10315 aarch64_evpc_rev. Handle SVE permutes too, trying
10316 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10317 than aarch64_evpc_tbl.
10318 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10319 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10320 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10321 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10322 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10323 (aarch64_expand_sve_vcond): New functions.
10324 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10325 of aarch64_vector_mode_p.
10326 (aarch64_dwarf_poly_indeterminate_value): New function.
10327 (aarch64_compute_pressure_classes): Likewise.
10328 (aarch64_can_change_mode_class): Likewise.
10329 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10330 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10331 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10332 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10333 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10334 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10335 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10336 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10338 (Dn, Dl, Dr): Accept const as well as const_vector.
10339 (Dz): Likewise. Compare against CONST0_RTX.
10340 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10341 of "vector" where appropriate.
10342 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10343 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10344 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10345 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10346 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10347 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10348 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10349 (v_int_equiv): Extend to SVE modes.
10350 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10352 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10353 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10354 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10355 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10356 (SVE_COND_FP_CMP): New int iterators.
10357 (perm_hilo): Handle the new unpack unspecs.
10358 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10360 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10361 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10362 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10363 (aarch64_equality_operator, aarch64_constant_vector_operand)
10364 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10365 (aarch64_sve_nonimmediate_operand): Likewise.
10366 (aarch64_sve_general_operand): Likewise.
10367 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10368 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10369 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10370 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10371 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10372 (aarch64_sve_float_arith_immediate): Likewise.
10373 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10374 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10375 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10376 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10377 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10378 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10379 (aarch64_sve_float_arith_operand): Likewise.
10380 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10381 (aarch64_sve_float_mul_operand): Likewise.
10382 (aarch64_sve_vec_perm_operand): Likewise.
10383 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10384 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10385 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10386 as well as const_vector.
10387 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10388 in file. Use CONST0_RTX and CONSTM1_RTX.
10389 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10390 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10391 Use aarch64_simd_imm_zero.
10392 * config/aarch64/aarch64-sve.md: New file.
10393 * config/aarch64/aarch64.md: Include it.
10394 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10395 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10396 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10397 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10398 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10399 (sve): New attribute.
10400 (enabled): Disable instructions with the sve attribute unless
10402 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10403 aarch64_expand_mov_immediate.
10404 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10405 CNT[BHSD] immediates.
10406 (movti): Split CONST_POLY_INT moves into two halves.
10407 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10408 Split additions that need a temporary here if the destination
10409 is the stack pointer.
10410 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10411 (*add<mode>3_poly_1): New instruction.
10412 (set_clobber_cc): New expander.
10414 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10416 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10417 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10418 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10419 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10420 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10421 Change innermode from fixed_mode_size to machine_mode.
10422 (simplify_subreg): Update call accordingly. Handle a constant-sized
10423 subreg of a variable-length CONST_VECTOR.
10425 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10426 Alan Hayward <alan.hayward@arm.com>
10427 David Sherwood <david.sherwood@arm.com>
10429 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10430 (add_offset_to_base): New function, split out from...
10431 (create_mem_ref): ...here. When handling a scale other than 1,
10432 check first whether the address is valid without the offset.
10433 Add it into the base if so, leaving the index and scale as-is.
10435 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10438 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10439 fold_for_warn before checking if arg2 is INTEGER_CST.
10441 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10443 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10444 (store_multiple_operation): Delete.
10445 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10446 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10447 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10448 guarded by TARGET_STRING.
10449 (rs6000_output_load_multiple): Delete.
10450 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10451 OPTION_MASK_STRING / TARGET_STRING handling.
10452 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10453 (const rs6000_opt_masks) <"string">: Change mask to 0.
10454 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10455 (MASK_STRING): Delete.
10456 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10458 (load_multiple): Delete.
10465 (store_multiple): Delete.
10472 (movmemsi_8reg): Delete.
10473 (corresponding unnamed define_insn): Delete.
10474 (movmemsi_6reg): Delete.
10475 (corresponding unnamed define_insn): Delete.
10476 (movmemsi_4reg): Delete.
10477 (corresponding unnamed define_insn): Delete.
10478 (movmemsi_2reg): Delete.
10479 (corresponding unnamed define_insn): Delete.
10480 (movmemsi_1reg): Delete.
10481 (corresponding unnamed define_insn): Delete.
10482 * config/rs6000/rs6000.opt (mno-string): New.
10483 (mstring): Replace by deprecation warning stub.
10484 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10486 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10488 * regrename.c (regrename_do_replace): If replacing the same
10489 reg multiple times, try to reuse last created gen_raw_REG.
10492 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10493 main to workaround a bug in GDB.
10495 2018-01-12 Tom de Vries <tom@codesourcery.com>
10498 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10500 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10502 PR rtl-optimization/80481
10503 * ira-color.c (get_cap_member): New function.
10504 (allocnos_conflict_by_live_ranges_p): Use it.
10505 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10506 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10508 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10511 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10512 (*saddl_se_1): Ditto.
10513 (*ssubsi_1): Ditto.
10514 (*ssubl_se_1): Ditto.
10516 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10518 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10519 rather than wi::to_widest for DR_INITs.
10520 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10521 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10522 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10524 (vect_analyze_group_access_1): Note that here.
10526 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10528 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10529 polynomial type sizes.
10531 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10533 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10534 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10535 (gimple_add_tmp_var): Likewise.
10537 2018-01-12 Martin Liska <mliska@suse.cz>
10539 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10540 (gimple_alloc_sizes): Likewise.
10541 (dump_gimple_statistics): Use PRIu64 in printf format.
10542 * gimple.h: Change uint64_t to int.
10544 2018-01-12 Martin Liska <mliska@suse.cz>
10546 * tree-core.h: Use uint64_t instead of int.
10547 * tree.c (tree_node_counts): Likewise.
10548 (tree_node_sizes): Likewise.
10549 (dump_tree_statistics): Use PRIu64 in printf format.
10551 2018-01-12 Martin Liska <mliska@suse.cz>
10553 * Makefile.in: As qsort_chk is implemented in vec.c, add
10554 vec.o to linkage of gencfn-macros.
10555 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10556 passing the info to record_node_allocation_statistics.
10557 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10559 * ggc-common.c (struct ggc_usage): Add operator== and use
10560 it in operator< and compare function.
10561 * mem-stats.h (struct mem_usage): Likewise.
10562 * vec.c (struct vec_usage): Remove operator< and compare
10563 function. Can be simply inherited.
10565 2018-01-12 Martin Jambor <mjambor@suse.cz>
10568 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10569 * tree-ssa-math-opts.c: Include domwalk.h.
10570 (convert_mult_to_fma_1): New function.
10571 (fma_transformation_info): New type.
10572 (fma_deferring_state): Likewise.
10573 (cancel_fma_deferring): New function.
10574 (result_of_phi): Likewise.
10575 (last_fma_candidate_feeds_initial_phi): Likewise.
10576 (convert_mult_to_fma): Added deferring logic, split actual
10577 transformation to convert_mult_to_fma_1.
10578 (math_opts_dom_walker): New type.
10579 (math_opts_dom_walker::after_dom_children): New method, body moved
10580 here from pass_optimize_widening_mul::execute, added deferring logic
10582 (pass_optimize_widening_mul::execute): Moved most of code to
10583 math_opts_dom_walker::after_dom_children.
10584 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10585 * config/i386/i386.c (ix86_option_override_internal): Added
10586 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10588 2018-01-12 Richard Biener <rguenther@suse.de>
10591 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10592 inline instance vars.
10594 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10597 * config/rx/rx.c (rx_is_restricted_memory_address):
10598 Handle SUBREG case.
10600 2018-01-12 Richard Biener <rguenther@suse.de>
10602 PR tree-optimization/80846
10603 * target.def (split_reduction): New target hook.
10604 * targhooks.c (default_split_reduction): New function.
10605 * targhooks.h (default_split_reduction): Declare.
10606 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10607 target requests first reduce vectors by combining low and high
10609 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10610 (get_vectype_for_scalar_type_and_size): Export.
10611 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10612 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10613 * doc/tm.texi: Regenerate.
10614 * config/i386/i386.c (ix86_split_reduction): Implement
10615 TARGET_VECTORIZE_SPLIT_REDUCTION.
10617 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10620 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10621 in PIC mode except for TARGET_VXWORKS_RTP.
10622 * config/sparc/sparc.c: Include cfgrtl.h.
10623 (TARGET_INIT_PIC_REG): Define.
10624 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10625 (sparc_pic_register_p): New predicate.
10626 (sparc_legitimate_address_p): Use it.
10627 (sparc_legitimize_pic_address): Likewise.
10628 (sparc_delegitimize_address): Likewise.
10629 (sparc_mode_dependent_address_p): Likewise.
10630 (gen_load_pcrel_sym): Remove 4th parameter.
10631 (load_got_register): Adjust call to above. Remove obsolete stuff.
10632 (sparc_expand_prologue): Do not call load_got_register here.
10633 (sparc_flat_expand_prologue): Likewise.
10634 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10635 (sparc_use_pseudo_pic_reg): New function.
10636 (sparc_init_pic_reg): Likewise.
10637 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10638 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10640 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10642 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10643 Add item for branch_cost.
10645 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10647 PR rtl-optimization/83565
10648 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10649 not extend the result to a larger mode for rotate operations.
10650 (num_sign_bit_copies1): Likewise.
10652 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10655 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10657 Use values-Xc.o for -pedantic.
10658 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10660 2018-01-12 Martin Liska <mliska@suse.cz>
10663 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10665 (possible_polymorphic_call_targets): Use it.
10666 (ipa_devirt): Likewise.
10668 2018-01-12 Martin Liska <mliska@suse.cz>
10670 * profile-count.h (enum profile_quality): Use 0 as invalid
10671 enum value of profile_quality.
10673 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10675 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10676 -mext-string options.
10678 2018-01-12 Richard Biener <rguenther@suse.de>
10680 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10681 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10682 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10684 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10686 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10688 * configure.ac (--with-long-double-format): Add support for the
10689 configuration option to change the default long double format on
10691 * config.gcc (powerpc*-linux*-*): Likewise.
10692 * configure: Regenerate.
10693 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10694 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10695 used without modification.
10697 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10699 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10700 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10701 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10702 MISC_BUILTIN_SPEC_BARRIER.
10703 (rs6000_init_builtins): Likewise.
10704 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10706 (speculation_barrier): New define_insn.
10707 * doc/extend.texi: Document __builtin_speculation_barrier.
10709 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10712 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10713 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10714 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10716 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10717 integral modes instead of "ss" and "sd".
10718 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10719 vectors with 32-bit and 64-bit elements.
10720 (vecdupssescalarmodesuffix): New mode attribute.
10721 (vec_dup<mode>): Use it.
10723 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10726 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10727 frame if argument is passed on stack.
10729 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10732 * ree.c (combine_reaching_defs): Optimize also
10733 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10734 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10736 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10738 PR middle-end/83189
10739 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10741 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10743 PR middle-end/83718
10744 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10745 after they are computed.
10747 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10749 PR tree-optimization/83695
10750 * gimple-loop-linterchange.cc
10751 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10752 reset cached scev information after interchange.
10753 (pass_linterchange::execute): Remove call to scev_reset_htab.
10755 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10757 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10758 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10759 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10760 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10761 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10762 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10763 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10764 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10765 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10766 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10767 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10768 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10769 (V_lane_reg): Likewise.
10770 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10772 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10773 (vfmal_lane_low<mode>_intrinsic,
10774 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10775 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10776 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10777 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10778 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10779 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10781 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10783 * config/arm/arm-cpus.in (fp16fml): New feature.
10784 (ALL_SIMD): Add fp16fml.
10785 (armv8.2-a): Add fp16fml as an option.
10786 (armv8.3-a): Likewise.
10787 (armv8.4-a): Add fp16fml as part of fp16.
10788 * config/arm/arm.h (TARGET_FP16FML): Define.
10789 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10791 * config/arm/arm-modes.def (V2HF): Define.
10792 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10793 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10794 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10795 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10796 vfmsl_low, vfmsl_high): New set of builtins.
10797 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10798 (vfml_op): New code attribute.
10799 (VFMLHALVES): New int iterator.
10800 (VFML, VFMLSEL): New mode attributes.
10801 (V_reg): Define mapping for V2HF.
10802 (V_hi, V_lo): New mode attributes.
10803 (VF_constraint): Likewise.
10804 (vfml_half, vfml_half_selector): New int attributes.
10805 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10807 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10808 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10810 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10811 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10812 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10813 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10815 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10816 Document new effective target and option set.
10818 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10820 * config/arm/arm-cpus.in (armv8_4): New feature.
10821 (ARMv8_4a): New fgroup.
10822 (armv8.4-a): New arch.
10823 * config/arm/arm-tables.opt: Regenerate.
10824 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10825 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10826 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10827 Add matching rules for -march=armv8.4-a and extensions.
10828 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10830 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10833 * config/rx/rx.md (BW): New mode attribute.
10834 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10836 2018-01-11 Richard Biener <rguenther@suse.de>
10838 PR tree-optimization/83435
10839 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10840 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10841 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10843 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10844 Alan Hayward <alan.hayward@arm.com>
10845 David Sherwood <david.sherwood@arm.com>
10847 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10849 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10850 (aarch64_print_address_internal): Use it to check for a zero offset.
10852 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10853 Alan Hayward <alan.hayward@arm.com>
10854 David Sherwood <david.sherwood@arm.com>
10856 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10857 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10858 Return a poly_int64 rather than a HOST_WIDE_INT.
10859 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10860 rather than a HOST_WIDE_INT.
10861 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10862 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10863 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10864 final_offset from HOST_WIDE_INT to poly_int64.
10865 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10866 to_constant when getting the number of units in an Advanced SIMD
10868 (aarch64_builtin_vectorized_function): Check for a constant number
10870 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10872 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10873 attribute instead of GET_MODE_NUNITS.
10874 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10875 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10876 GET_MODE_SIZE for fixed-size registers.
10877 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10878 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10879 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10880 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10881 (aarch64_print_operand, aarch64_print_address_internal)
10882 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10883 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10884 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10885 Handle polynomial GET_MODE_SIZE.
10886 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10887 wider than SImode without modification.
10888 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10889 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10890 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10891 passing and returning SVE modes.
10892 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10893 rather than GEN_INT.
10894 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10895 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10896 (aarch64_allocate_and_probe_stack_space): Likewise.
10897 (aarch64_layout_frame): Cope with polynomial offsets.
10898 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10899 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10900 polynomial offsets.
10901 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10902 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10903 poly_int64 rather than a HOST_WIDE_INT.
10904 (aarch64_get_separate_components, aarch64_process_components)
10905 (aarch64_expand_prologue, aarch64_expand_epilogue)
10906 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10907 (aarch64_anchor_offset): New function, split out from...
10908 (aarch64_legitimize_address): ...here.
10909 (aarch64_builtin_vectorization_cost): Handle polynomial
10910 TYPE_VECTOR_SUBPARTS.
10911 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10913 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10914 number of elements from the PARALLEL rather than the mode.
10915 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10916 rather than GET_MODE_BITSIZE.
10917 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10918 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10919 (aarch64_expand_vec_perm_const_1): Handle polynomial
10920 d->perm.length () and d->perm elements.
10921 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10922 Apply to_constant to d->perm elements.
10923 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10924 polynomial CONST_VECTOR_NUNITS.
10925 (aarch64_move_pointer): Take amount as a poly_int64 rather
10927 (aarch64_progress_pointer): Avoid temporary variable.
10928 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10929 the mode attribute instead of GET_MODE.
10931 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10932 Alan Hayward <alan.hayward@arm.com>
10933 David Sherwood <david.sherwood@arm.com>
10935 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10936 x exists before using it.
10937 (aarch64_add_constant_internal): Rename to...
10938 (aarch64_add_offset_1): ...this. Replace regnum with separate
10939 src and dest rtxes. Handle the case in which they're different,
10940 including when the offset is zero. Replace scratchreg with an rtx.
10941 Use 2 additions if there is no spare register into which we can
10942 move a 16-bit constant.
10943 (aarch64_add_constant): Delete.
10944 (aarch64_add_offset): Replace reg with separate src and dest
10945 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10946 Use aarch64_add_offset_1.
10947 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10948 an rtx rather than an int. Take the delta as a poly_int64
10949 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10950 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10951 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10952 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10953 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10954 and aarch64_add_sp.
10955 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10956 aarch64_add_constant.
10958 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10960 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10961 Use scalar_float_mode.
10963 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10965 * config/aarch64/aarch64-simd.md
10966 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10967 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10968 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10969 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10970 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10971 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10972 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10973 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10974 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10975 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10977 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10980 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10981 targ_options->x_arm_arch_string is non NULL.
10983 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10985 * config/aarch64/aarch64.h
10986 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10988 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10991 * expmed.c (emit_store_flag_force): Swap if const op0
10992 and change VOIDmode to mode of op0.
10994 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10996 PR rtl-optimization/83761
10997 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10998 than bytes to mode_for_size.
11000 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11002 PR middle-end/83189
11003 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11004 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11007 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11009 PR middle-end/83575
11010 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11011 when in layout mode.
11012 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11013 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11016 2018-01-10 Michael Collison <michael.collison@arm.com>
11018 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11019 * config/aarch64/aarch64-option-extension.def: Add
11020 AARCH64_OPT_EXTENSION of 'fp16fml'.
11021 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11022 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11023 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11024 * config/aarch64/constraints.md (Ui7): New constraint.
11025 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11026 (VFMLA_SEL_W): Ditto.
11029 (VFMLA16_LOW): New int iterator.
11030 (VFMLA16_HIGH): Ditto.
11031 (UNSPEC_FMLAL): New unspec.
11032 (UNSPEC_FMLSL): Ditto.
11033 (UNSPEC_FMLAL2): Ditto.
11034 (UNSPEC_FMLSL2): Ditto.
11035 (f16mac): New code attribute.
11036 * config/aarch64/aarch64-simd-builtins.def
11037 (aarch64_fmlal_lowv2sf): Ditto.
11038 (aarch64_fmlsl_lowv2sf): Ditto.
11039 (aarch64_fmlalq_lowv4sf): Ditto.
11040 (aarch64_fmlslq_lowv4sf): Ditto.
11041 (aarch64_fmlal_highv2sf): Ditto.
11042 (aarch64_fmlsl_highv2sf): Ditto.
11043 (aarch64_fmlalq_highv4sf): Ditto.
11044 (aarch64_fmlslq_highv4sf): Ditto.
11045 (aarch64_fmlal_lane_lowv2sf): Ditto.
11046 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11047 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11048 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11049 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11050 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11051 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11052 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11053 (aarch64_fmlal_lane_highv2sf): Ditto.
11054 (aarch64_fmlsl_lane_highv2sf): Ditto.
11055 (aarch64_fmlal_laneq_highv2sf): Ditto.
11056 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11057 (aarch64_fmlalq_lane_highv4sf): Ditto.
11058 (aarch64_fmlsl_lane_highv4sf): Ditto.
11059 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11060 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11061 * config/aarch64/aarch64-simd.md:
11062 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11063 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11064 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11065 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11066 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11067 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11068 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11069 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11070 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11071 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11072 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11073 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11074 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11075 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11076 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11077 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11078 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11079 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11080 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11081 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11082 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11083 (vfmlsl_low_u32): Ditto.
11084 (vfmlalq_low_u32): Ditto.
11085 (vfmlslq_low_u32): Ditto.
11086 (vfmlal_high_u32): Ditto.
11087 (vfmlsl_high_u32): Ditto.
11088 (vfmlalq_high_u32): Ditto.
11089 (vfmlslq_high_u32): Ditto.
11090 (vfmlal_lane_low_u32): Ditto.
11091 (vfmlsl_lane_low_u32): Ditto.
11092 (vfmlal_laneq_low_u32): Ditto.
11093 (vfmlsl_laneq_low_u32): Ditto.
11094 (vfmlalq_lane_low_u32): Ditto.
11095 (vfmlslq_lane_low_u32): Ditto.
11096 (vfmlalq_laneq_low_u32): Ditto.
11097 (vfmlslq_laneq_low_u32): Ditto.
11098 (vfmlal_lane_high_u32): Ditto.
11099 (vfmlsl_lane_high_u32): Ditto.
11100 (vfmlal_laneq_high_u32): Ditto.
11101 (vfmlsl_laneq_high_u32): Ditto.
11102 (vfmlalq_lane_high_u32): Ditto.
11103 (vfmlslq_lane_high_u32): Ditto.
11104 (vfmlalq_laneq_high_u32): Ditto.
11105 (vfmlslq_laneq_high_u32): Ditto.
11106 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11107 (AARCH64_FL_FOR_ARCH8_4): New.
11108 (AARCH64_ISA_F16FML): New ISA flag.
11109 (TARGET_F16FML): New feature flag for fp16fml.
11110 (doc/invoke.texi): Document new fp16fml option.
11112 2018-01-10 Michael Collison <michael.collison@arm.com>
11114 * config/aarch64/aarch64-builtins.c:
11115 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11116 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11117 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11118 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11119 (AARCH64_ISA_SHA3): New ISA flag.
11120 (TARGET_SHA3): New feature flag for sha3.
11121 * config/aarch64/iterators.md (sha512_op): New int attribute.
11122 (CRYPTO_SHA512): New int iterator.
11123 (UNSPEC_SHA512H): New unspec.
11124 (UNSPEC_SHA512H2): Ditto.
11125 (UNSPEC_SHA512SU0): Ditto.
11126 (UNSPEC_SHA512SU1): Ditto.
11127 * config/aarch64/aarch64-simd-builtins.def
11128 (aarch64_crypto_sha512hqv2di): New builtin.
11129 (aarch64_crypto_sha512h2qv2di): Ditto.
11130 (aarch64_crypto_sha512su0qv2di): Ditto.
11131 (aarch64_crypto_sha512su1qv2di): Ditto.
11132 (aarch64_eor3qv8hi): Ditto.
11133 (aarch64_rax1qv2di): Ditto.
11134 (aarch64_xarqv2di): Ditto.
11135 (aarch64_bcaxqv8hi): Ditto.
11136 * config/aarch64/aarch64-simd.md:
11137 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11138 (aarch64_crypto_sha512su0qv2di): Ditto.
11139 (aarch64_crypto_sha512su1qv2di): Ditto.
11140 (aarch64_eor3qv8hi): Ditto.
11141 (aarch64_rax1qv2di): Ditto.
11142 (aarch64_xarqv2di): Ditto.
11143 (aarch64_bcaxqv8hi): Ditto.
11144 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11145 (vsha512h2q_u64): Ditto.
11146 (vsha512su0q_u64): Ditto.
11147 (vsha512su1q_u64): Ditto.
11148 (veor3q_u16): Ditto.
11149 (vrax1q_u64): Ditto.
11150 (vxarq_u64): Ditto.
11151 (vbcaxq_u16): Ditto.
11152 * config/arm/types.md (crypto_sha512): New type attribute.
11153 (crypto_sha3): Ditto.
11154 (doc/invoke.texi): Document new sha3 option.
11156 2018-01-10 Michael Collison <michael.collison@arm.com>
11158 * config/aarch64/aarch64-builtins.c:
11159 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11160 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11161 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11162 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11163 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11164 (AARCH64_ISA_SM4): New ISA flag.
11165 (TARGET_SM4): New feature flag for sm4.
11166 * config/aarch64/aarch64-simd-builtins.def
11167 (aarch64_sm3ss1qv4si): Ditto.
11168 (aarch64_sm3tt1aq4si): Ditto.
11169 (aarch64_sm3tt1bq4si): Ditto.
11170 (aarch64_sm3tt2aq4si): Ditto.
11171 (aarch64_sm3tt2bq4si): Ditto.
11172 (aarch64_sm3partw1qv4si): Ditto.
11173 (aarch64_sm3partw2qv4si): Ditto.
11174 (aarch64_sm4eqv4si): Ditto.
11175 (aarch64_sm4ekeyqv4si): Ditto.
11176 * config/aarch64/aarch64-simd.md:
11177 (aarch64_sm3ss1qv4si): Ditto.
11178 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11179 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11180 (aarch64_sm4eqv4si): Ditto.
11181 (aarch64_sm4ekeyqv4si): Ditto.
11182 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11183 (sm3part_op): Ditto.
11184 (CRYPTO_SM3TT): Ditto.
11185 (CRYPTO_SM3PART): Ditto.
11186 (UNSPEC_SM3SS1): New unspec.
11187 (UNSPEC_SM3TT1A): Ditto.
11188 (UNSPEC_SM3TT1B): Ditto.
11189 (UNSPEC_SM3TT2A): Ditto.
11190 (UNSPEC_SM3TT2B): Ditto.
11191 (UNSPEC_SM3PARTW1): Ditto.
11192 (UNSPEC_SM3PARTW2): Ditto.
11193 (UNSPEC_SM4E): Ditto.
11194 (UNSPEC_SM4EKEY): Ditto.
11195 * config/aarch64/constraints.md (Ui2): New constraint.
11196 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11197 * config/arm/types.md (crypto_sm3): New type attribute.
11198 (crypto_sm4): Ditto.
11199 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11200 (vsm3tt1aq_u32): Ditto.
11201 (vsm3tt1bq_u32): Ditto.
11202 (vsm3tt2aq_u32): Ditto.
11203 (vsm3tt2bq_u32): Ditto.
11204 (vsm3partw1q_u32): Ditto.
11205 (vsm3partw2q_u32): Ditto.
11206 (vsm4eq_u32): Ditto.
11207 (vsm4ekeyq_u32): Ditto.
11208 (doc/invoke.texi): Document new sm4 option.
11210 2018-01-10 Michael Collison <michael.collison@arm.com>
11212 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11213 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11214 (AARCH64_FL_FOR_ARCH8_4): New.
11215 (AARCH64_FL_V8_4): New flag.
11216 (doc/invoke.texi): Document new armv8.4-a option.
11218 2018-01-10 Michael Collison <michael.collison@arm.com>
11220 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11221 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11222 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11223 * config/aarch64/aarch64-option-extension.def: Add
11224 AARCH64_OPT_EXTENSION of 'sha2'.
11225 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11226 (crypto): Disable sha2 and aes if crypto disabled.
11227 (crypto): Enable aes and sha2 if enabled.
11228 (simd): Disable sha2 and aes if simd disabled.
11229 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11231 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11232 (TARGET_SHA2): New feature flag for sha2.
11233 (TARGET_AES): New feature flag for aes.
11234 * config/aarch64/aarch64-simd.md:
11235 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11236 conditional on TARGET_AES.
11237 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11238 (aarch64_crypto_sha1hsi): Make pattern conditional
11240 (aarch64_crypto_sha1hv4si): Ditto.
11241 (aarch64_be_crypto_sha1hv4si): Ditto.
11242 (aarch64_crypto_sha1su1v4si): Ditto.
11243 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11244 (aarch64_crypto_sha1su0v4si): Ditto.
11245 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11246 (aarch64_crypto_sha256su0v4si): Ditto.
11247 (aarch64_crypto_sha256su1v4si): Ditto.
11248 (doc/invoke.texi): Document new aes and sha2 options.
11250 2018-01-10 Martin Sebor <msebor@redhat.com>
11252 PR tree-optimization/83781
11253 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11256 2018-01-11 Martin Sebor <msebor@gmail.com>
11257 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11259 PR tree-optimization/83501
11260 PR tree-optimization/81703
11262 * tree-ssa-strlen.c (get_string_cst): Rename...
11263 (get_string_len): ...to this. Handle global constants.
11264 (handle_char_store): Adjust.
11266 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11267 Jim Wilson <jimw@sifive.com>
11269 * config/riscv/riscv-protos.h (riscv_output_return): New.
11270 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11271 (riscv_attribute_table, riscv_output_return),
11272 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11273 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11274 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11276 (riscv_expand_prologue): Add early return for naked function.
11277 (riscv_expand_epilogue): Likewise.
11278 (riscv_function_ok_for_sibcall): Return false for naked function.
11279 (riscv_set_current_function): New.
11280 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11281 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11282 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11283 * doc/extend.texi (RISC-V Function Attributes): New.
11285 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11287 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11288 check for 128-bit long double before checking TCmode.
11289 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11290 128-bit long doubles before checking TFmode or TCmode.
11291 (FLOAT128_IBM_P): Likewise.
11293 2018-01-10 Martin Sebor <msebor@redhat.com>
11295 PR tree-optimization/83671
11296 * builtins.c (c_strlen): Unconditionally return zero for the empty
11298 Use -Warray-bounds for warnings.
11299 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11300 for non-constant array indices with COMPONENT_REF, arrays of
11301 arrays, and pointers to arrays.
11302 (gimple_fold_builtin_strlen): Determine and set length range for
11303 non-constant character arrays.
11305 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11307 PR middle-end/81897
11308 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11311 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11313 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11315 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11318 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11319 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11320 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11321 indexed_or_indirect_operand predicate.
11322 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11323 (*vsx_le_perm_load_v8hi): Likewise.
11324 (*vsx_le_perm_load_v16qi): Likewise.
11325 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11326 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11327 (*vsx_le_perm_store_v8hi): Likewise.
11328 (*vsx_le_perm_store_v16qi): Likewise.
11329 (eight unnamed splitters): Likewise.
11331 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11333 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11334 * config/rs6000/emmintrin.h: Likewise.
11335 * config/rs6000/mmintrin.h: Likewise.
11336 * config/rs6000/xmmintrin.h: Likewise.
11338 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11341 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11343 * tree.c (tree_nop_conversion): Return true for location wrapper
11345 (maybe_wrap_with_location): New function.
11346 (selftest::check_strip_nops): New function.
11347 (selftest::test_location_wrappers): New function.
11348 (selftest::tree_c_tests): Call it.
11349 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11350 (maybe_wrap_with_location): New decl.
11351 (EXPR_LOCATION_WRAPPER_P): New macro.
11352 (location_wrapper_p): New inline function.
11353 (tree_strip_any_location_wrapper): New inline function.
11355 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11358 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11359 stack_realign_offset for the largest alignment of stack slot
11361 (ix86_find_max_used_stack_alignment): New function.
11362 (ix86_finalize_stack_frame_flags): Use it. Set
11363 max_used_stack_alignment if we don't realign stack.
11364 * config/i386/i386.h (machine_function): Add
11365 max_used_stack_alignment.
11367 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11369 * config/arm/arm.opt (-mbranch-cost): New option.
11370 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11373 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11376 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11377 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11379 2018-01-10 Richard Biener <rguenther@suse.de>
11382 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11383 early out so it also covers the case where we have a non-NULL
11386 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11388 PR tree-optimization/83753
11389 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11390 for non-strided grouped accesses if the number of elements is 1.
11392 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11395 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11396 * i386.h (TARGET_USE_GATHER): Define.
11397 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11399 2018-01-10 Martin Liska <mliska@suse.cz>
11402 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11403 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11405 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11406 CLEANUP_NO_PARTITIONING is not set.
11408 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11410 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11411 for vectors, as a partial revert of r254296.
11412 * rtl.h (const_vec_p): Delete.
11413 (const_vec_duplicate_p): Don't test for vector CONSTs.
11414 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11415 * expmed.c (make_tree): Likewise.
11418 * common.md (E, F): Use CONSTANT_P instead of checking for
11420 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11421 checking for CONST_VECTOR.
11423 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11425 PR middle-end/83575
11426 * predict.c (force_edge_cold): Handle in more sane way edges
11427 with no prediction.
11429 2018-01-09 Carl Love <cel@us.ibm.com>
11431 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11433 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11434 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11435 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11436 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11437 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11438 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11439 * config/rs6000/rs6000-protos.h: Add extern defition for
11440 rs6000_generate_float2_double_code.
11441 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11443 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11444 (float2_v2df): Add define_expand.
11446 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11449 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11450 op_mode in the force_to_mode call.
11452 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11454 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11455 instead of checking each element individually.
11456 (aarch64_evpc_uzp): Likewise.
11457 (aarch64_evpc_zip): Likewise.
11458 (aarch64_evpc_ext): Likewise.
11459 (aarch64_evpc_rev): Likewise.
11460 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11461 instead of checking each element individually. Return true without
11463 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11464 whether all selected elements come from the same input, instead of
11465 checking each element individually. Remove calls to gen_rtx_REG,
11466 start_sequence and end_sequence and instead assert that no rtl is
11469 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11471 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11472 order of HIGH and CONST checks.
11474 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11476 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11477 if the destination isn't an SSA_NAME.
11479 2018-01-09 Richard Biener <rguenther@suse.de>
11481 PR tree-optimization/83668
11482 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11484 (canonicalize_loop_form): ... here, renamed from ...
11485 (canonicalize_loop_closed_ssa_form): ... this and amended to
11486 swap successor edges for loop exit blocks to make us use
11487 the RPO order we need for initial schedule generation.
11489 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11491 PR tree-optimization/64811
11492 * match.pd: When optimizing comparisons with Inf, avoid
11493 introducing or losing exceptions from comparisons with NaN.
11495 2018-01-09 Martin Liska <mliska@suse.cz>
11498 * asan.c (shadow_mem_size): Add gcc_assert.
11500 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11502 Don't save registers in main().
11505 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11506 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11507 * config/avr/avr.c (avr_set_current_function): Don't error if
11508 naked, OS_task or OS_main are specified at the same time.
11509 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11511 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11513 * common/config/avr/avr-common.c (avr_option_optimization_table):
11514 Switch on -mmain-is-OS_task for optimizing compilations.
11516 2018-01-09 Richard Biener <rguenther@suse.de>
11518 PR tree-optimization/83572
11519 * graphite.c: Include cfganal.h.
11520 (graphite_transform_loops): Connect infinite loops to exit
11521 and remove fake edges at the end.
11523 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11525 * ipa-inline.c (edge_badness): Revert accidental checkin.
11527 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11530 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11531 symbols; not inline clones.
11533 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11536 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11537 hard registers. Formatting fixes.
11539 PR preprocessor/83722
11540 * gcc.c (try_generate_repro): Pass
11541 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11542 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11545 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11546 Kito Cheng <kito.cheng@gmail.com>
11548 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11549 (riscv_leaf_function_p): Delete.
11550 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11552 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11554 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11556 (do_ifelse): New function.
11557 (do_isel): New function.
11558 (do_sub3): New function.
11559 (do_add3): New function.
11560 (do_load_mask_compare): New function.
11561 (do_overlap_load_compare): New function.
11562 (expand_compare_loop): New function.
11563 (expand_block_compare): Call expand_compare_loop() when appropriate.
11564 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11565 option description.
11566 (-mblock-compare-inline-loop-limit): New option.
11568 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11571 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11572 Reverse order of second and third operands in first alternative.
11573 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11574 of first and second elements in UNSPEC_VPERMR vector.
11575 (altivec_expand_vec_perm_le): Likewise.
11577 2018-01-08 Jeff Law <law@redhat.com>
11579 PR rtl-optimizatin/81308
11580 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11581 (process_switch): If group_case_labels makes a change, then set
11583 (pass_convert_switch::execute): If a switch is converted, then
11584 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11586 PR rtl-optimization/81308
11587 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11590 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11592 PR target/83663 - Revert r255946
11593 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11594 generation for cases where splatting a value is not useful.
11595 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11596 across a vec_duplicate and a paradoxical subreg forming a vector
11597 mode to a vec_concat.
11599 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11601 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11602 -march=armv8.3-a variants.
11603 * config/arm/t-multilib: Likewise.
11604 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11606 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11608 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11610 (cceq_ior_compare_complement): Give it a name so I can use it, and
11611 change boolean_or_operator predicate to boolean_operator so it can
11612 be used to generate a crand.
11613 (eqne): New code iterator.
11614 (bd/bd_neg): New code_attrs.
11615 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11616 a single define_insn.
11617 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11618 decrement (bdnzt/bdnzf/bdzt/bdzf).
11619 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11620 with the new names of the branch decrement patterns, and added the
11621 names of the branch decrement conditional patterns.
11623 2018-01-08 Richard Biener <rguenther@suse.de>
11625 PR tree-optimization/83563
11626 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11629 2018-01-08 Richard Biener <rguenther@suse.de>
11631 PR middle-end/83713
11632 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11634 2018-01-08 Richard Biener <rguenther@suse.de>
11636 PR tree-optimization/83685
11637 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11638 references to abnormals.
11640 2018-01-08 Richard Biener <rguenther@suse.de>
11643 * dwarf2out.c (output_indirect_strings): Handle empty
11644 skeleton_debug_str_hash.
11645 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11647 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11649 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11650 (emit_store_direct): Likewise.
11651 (arc_trampoline_adjust_address): Likewise.
11652 (arc_asm_trampoline_template): New function.
11653 (arc_initialize_trampoline): Use asm_trampoline_template.
11654 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11655 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11656 * config/arc/arc.md (flush_icache): Delete pattern.
11658 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11660 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11661 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11664 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11667 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11668 by not USED_FOR_TARGET.
11669 (make_pass_resolve_sw_modes): Likewise.
11671 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11673 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11676 2018-01-08 Richard Biener <rguenther@suse.de>
11678 PR middle-end/83580
11679 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11681 2018-01-08 Richard Biener <rguenther@suse.de>
11683 PR middle-end/83517
11684 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11686 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11688 PR middle-end/81897
11689 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11690 basic blocks with a small number of successors.
11691 (convert_control_dep_chain_into_preds): Improve handling of
11693 (dump_predicates): Split apart into...
11694 (dump_pred_chain): ...here...
11695 (dump_pred_info): ...and here.
11696 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11697 (can_chain_union_be_invalidated_p): Improve check for invalidation
11699 (uninit_uses_cannot_happen): Avoid unnecessary if
11700 convert_control_dep_chain_into_preds yielded nothing.
11702 2018-01-06 Martin Sebor <msebor@redhat.com>
11704 PR tree-optimization/83640
11705 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11706 subtracting negative offset from size.
11707 (builtin_access::overlap): Adjust offset bounds of the access to fall
11708 within the size of the object if possible.
11710 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11712 PR rtl-optimization/83699
11713 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11714 extract_bit_field_as_subreg to cases in which the extracted
11715 value is also a vector.
11717 * lra-constraints.c (process_alt_operands): Test for the equivalence
11718 substitutions when detecting a possible reload cycle.
11720 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11723 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11724 by default if flag_selective_schedling{,2}. Formatting fixes.
11726 PR rtl-optimization/83682
11727 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11728 if it has non-VECTOR_MODE element mode.
11729 (vec_duplicate_p): Likewise.
11731 PR middle-end/83694
11732 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11733 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11735 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11738 * config/i386/i386-builtin.def
11739 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11740 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11741 Require also OPTION_MASK_ISA_AVX512F in addition to
11742 OPTION_MASK_ISA_GFNI.
11743 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11744 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11745 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11746 to OPTION_MASK_ISA_GFNI.
11747 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11748 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11749 OPTION_MASK_ISA_AVX512BW.
11750 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11751 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11752 addition to OPTION_MASK_ISA_GFNI.
11753 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11754 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11755 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11756 to OPTION_MASK_ISA_GFNI.
11757 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11758 a requirement for all ISAs rather than any of them with a few
11760 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11762 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11763 bitmasks to be enabled with 3 exceptions, instead of requiring any
11764 enabled ISA with lots of exceptions.
11765 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11766 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11767 Change avx512bw in isa attribute to avx512f.
11768 * config/i386/sgxintrin.h: Add license boilerplate.
11769 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11770 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11771 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11772 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11774 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11775 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11776 temporarily sse2 rather than sse if not enabled already.
11779 * config/i386/sse.md (VI248_VLBW): Rename to ...
11780 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11781 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11782 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11783 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11784 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11785 mode iterator instead of VI248_VLBW.
11787 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11789 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11790 (record_modified): Skip clobbers; add debug output.
11791 (param_change_prob): Use sreal frequencies.
11793 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11795 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11796 punt for user-aligned variables.
11798 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11800 * tree-chrec.c (chrec_contains_symbols): Return true for
11803 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11806 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11807 of (x|y) == x for BICS pattern.
11809 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11811 PR tree-optimization/83605
11812 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11813 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11816 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11818 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11819 * config/epiphany/rtems.h: New file.
11821 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11822 Uros Bizjak <ubizjak@gmail.com>
11825 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11826 QIreg_operand instead of register_operand predicate.
11827 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11828 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11829 comments instead of -fmitigate[-_]rop.
11831 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11834 * cgraphunit.c (symbol_table::compile): Switch to text_section
11835 before calling assembly_start debug hook.
11836 * run-rtl-passes.c (run_rtl_passes): Likewise.
11839 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11841 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11842 range_int_cst_p rather than !symbolic_range_p before calling
11843 extract_range_from_multiplicative_op_1.
11845 2018-01-04 Jeff Law <law@redhat.com>
11847 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11848 redundant test in assertion.
11850 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11852 * doc/rtl.texi: Document machine_mode wrapper classes.
11854 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11856 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11857 using tree_to_uhwi.
11859 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11861 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11862 the VEC_PERM_EXPR fold to fail.
11864 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11867 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11868 to switched_sections.
11870 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11873 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11874 test for d.testing.
11876 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11879 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11880 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11882 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11885 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11886 is BLKmode and bitpos not zero or mode change is needed.
11888 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11891 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11894 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11897 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11898 instead of MULT rtx. Update all corresponding splitters.
11899 (*saddl_se): Ditto.
11900 (*ssub<modesuffix>): Ditto.
11901 (*ssubl_se): Ditto.
11902 (*cmp_sadd_di): Update split patterns.
11903 (*cmp_sadd_si): Ditto.
11904 (*cmp_sadd_sidi): Ditto.
11905 (*cmp_ssub_di): Ditto.
11906 (*cmp_ssub_si): Ditto.
11907 (*cmp_ssub_sidi): Ditto.
11908 * config/alpha/predicates.md (const23_operand): New predicate.
11909 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11910 Look for ASHIFT, not MULT inner operand.
11911 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11913 2018-01-04 Martin Liska <mliska@suse.cz>
11915 PR gcov-profile/83669
11916 * gcov.c (output_intermediate_file): Add version to intermediate
11918 * doc/gcov.texi: Document new field 'version' in intermediate
11919 file format. Fix location of '-k' option of gcov command.
11921 2018-01-04 Martin Liska <mliska@suse.cz>
11924 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11926 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11928 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11930 2018-01-03 Martin Sebor <msebor@redhat.com>
11932 PR tree-optimization/83655
11933 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11934 checking calls with invalid arguments.
11936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11938 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11939 (vectorizable_mask_load_store): Delete.
11940 (vectorizable_call): Return false for masked loads and stores.
11941 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11942 instead of gimple_assign_rhs1.
11943 (vectorizable_load): Handle IFN_MASK_LOAD.
11944 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11946 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11948 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11950 (vectorizable_mask_load_store): ...here.
11951 (vectorizable_load): ...and here.
11953 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11955 * tree-vect-stmts.c (vect_build_all_ones_mask)
11956 (vect_build_zero_merge_argument): New functions, split out from...
11957 (vectorizable_load): ...here.
11959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11961 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11963 (vectorizable_mask_load_store): ...here.
11964 (vectorizable_store): ...and here.
11966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11968 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11970 (vectorizable_mask_load_store): ...here.
11972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11974 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11975 (vect_model_store_cost): Take a vec_load_store_type instead of a
11977 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11978 (vect_model_store_cost): Take a vec_load_store_type instead of a
11980 (vectorizable_mask_load_store): Update accordingly.
11981 (vectorizable_store): Likewise.
11982 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11986 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11987 IFN_MASK_LOAD calls here rather than...
11988 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11991 Alan Hayward <alan.hayward@arm.com>
11992 David Sherwood <david.sherwood@arm.com>
11994 * expmed.c (extract_bit_field_1): For vector extracts,
11995 fall back to extract_bit_field_as_subreg if vec_extract
11998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11999 Alan Hayward <alan.hayward@arm.com>
12000 David Sherwood <david.sherwood@arm.com>
12002 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12003 they are variable or constant sized.
12004 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12005 slots for constant-sized data.
12007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12008 Alan Hayward <alan.hayward@arm.com>
12009 David Sherwood <david.sherwood@arm.com>
12011 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12012 handling COND_EXPRs with boolean comparisons, try to find a better
12013 basis for the mask type than the boolean itself.
12015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12017 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12018 is calculated and how it can be overridden.
12019 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12020 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12022 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12026 Alan Hayward <alan.hayward@arm.com>
12027 David Sherwood <david.sherwood@arm.com>
12029 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12030 Remove the mode argument.
12031 (aarch64_simd_valid_immediate): Remove the mode and inverse
12033 * config/aarch64/iterators.md (bitsize): New iterator.
12034 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12035 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12036 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12037 aarch64_simd_valid_immediate.
12038 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12039 (aarch64_reg_or_bic_imm): Likewise.
12040 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12041 with an insn_type enum and msl with a modifier_type enum.
12042 Replace element_width with a scalar_mode. Change the shift
12043 to unsigned int. Add constructors for scalar_float_mode and
12044 scalar_int_mode elements.
12045 (aarch64_vect_float_const_representable_p): Delete.
12046 (aarch64_can_const_movi_rtx_p)
12047 (aarch64_simd_scalar_immediate_valid_for_move)
12048 (aarch64_simd_make_constant): Update call to
12049 aarch64_simd_valid_immediate.
12050 (aarch64_advsimd_valid_immediate_hs): New function.
12051 (aarch64_advsimd_valid_immediate): Likewise.
12052 (aarch64_simd_valid_immediate): Remove mode and inverse
12053 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12054 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12055 and aarch64_float_const_representable_p on the result.
12056 (aarch64_output_simd_mov_immediate): Remove mode argument.
12057 Update call to aarch64_simd_valid_immediate and use of
12058 simd_immediate_info.
12059 (aarch64_output_scalar_simd_mov_immediate): Update call
12062 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12063 Alan Hayward <alan.hayward@arm.com>
12064 David Sherwood <david.sherwood@arm.com>
12066 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12067 (mode_nunits): Likewise CONST_MODE_NUNITS.
12068 * machmode.def (ADJUST_NUNITS): Document.
12069 * genmodes.c (mode_data::need_nunits_adj): New field.
12070 (blank_mode): Update accordingly.
12071 (adj_nunits): New variable.
12072 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12074 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12075 listed in adj_nunits.
12076 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12077 listed in adj_nunits. Don't emit case statements for such modes.
12078 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12079 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12080 nothing if adj_nunits is nonnull.
12081 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12082 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12083 (emit_mode_fbit): Update use of print_maybe_const_decl.
12084 (emit_move_size): Likewise. Treat the array as non-const
12086 (emit_mode_adjustments): Handle adj_nunits.
12088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12090 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12091 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12092 (VECTOR_MODES): Use it.
12093 (make_vector_modes): Take the prefix as an argument.
12095 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12096 Alan Hayward <alan.hayward@arm.com>
12097 David Sherwood <david.sherwood@arm.com>
12099 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12100 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12101 for MODE_VECTOR_BOOL.
12102 * machmode.def (VECTOR_BOOL_MODE): Document.
12103 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12104 (make_vector_bool_mode): New function.
12105 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12107 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12108 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12110 * stor-layout.c (int_mode_for_mode): Likewise.
12111 * tree.c (build_vector_type_for_mode): Likewise.
12112 * varasm.c (output_constant_pool_2): Likewise.
12113 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12114 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12115 for MODE_VECTOR_BOOL.
12116 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12117 of mode class checks.
12118 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12119 instead of a list of mode class checks.
12120 (expand_vector_scalar_condition): Likewise.
12121 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12123 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12124 Alan Hayward <alan.hayward@arm.com>
12125 David Sherwood <david.sherwood@arm.com>
12127 * machmode.h (mode_size): Change from unsigned short to
12129 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12130 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12131 or if measurement_type is not polynomial.
12132 (fixed_size_mode::includes_p): Check for constant-sized modes.
12133 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12134 return a poly_uint16 rather than an unsigned short.
12135 (emit_mode_size): Change the type of mode_size from unsigned short
12136 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12137 (emit_mode_adjustments): Cope with polynomial vector sizes.
12138 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12140 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12142 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12143 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12144 * caller-save.c (setup_save_areas): Likewise.
12145 (replace_reg_with_saved_mem): Likewise.
12146 * calls.c (emit_library_call_value_1): Likewise.
12147 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12148 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12149 (gen_lowpart_for_combine): Likewise.
12150 * convert.c (convert_to_integer_1): Likewise.
12151 * cse.c (equiv_constant, cse_insn): Likewise.
12152 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12153 (cselib_subst_to_values): Likewise.
12154 * dce.c (word_dce_process_block): Likewise.
12155 * df-problems.c (df_word_lr_mark_ref): Likewise.
12156 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12157 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12158 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12159 (rtl_for_decl_location): Likewise.
12160 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12161 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12162 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12163 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12164 (expand_expr_real_1): Likewise.
12165 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12166 (pad_below): Likewise.
12167 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12168 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12169 * ira.c (get_subreg_tracking_sizes): Likewise.
12170 * ira-build.c (ira_create_allocno_objects): Likewise.
12171 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12172 (ira_sort_regnos_for_alter_reg): Likewise.
12173 * ira-costs.c (record_operand_costs): Likewise.
12174 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12175 (resolve_simple_move): Likewise.
12176 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12177 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12178 (lra_constraints): Likewise.
12179 (CONST_POOL_OK_P): Reject variable-sized modes.
12180 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12181 (add_pseudo_to_slot, lra_spill): Likewise.
12182 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12183 * optabs-query.c (get_best_extraction_insn): Likewise.
12184 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12185 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12186 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12187 * recog.c (offsettable_address_addr_space_p): Likewise.
12188 * regcprop.c (maybe_mode_change): Likewise.
12189 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12190 * regrename.c (build_def_use): Likewise.
12191 * regstat.c (dump_reg_info): Likewise.
12192 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12193 (find_reloads, find_reloads_subreg_address): Likewise.
12194 * reload1.c (eliminate_regs_1): Likewise.
12195 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12196 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12197 (simplify_binary_operation_1, simplify_subreg): Likewise.
12198 * targhooks.c (default_function_arg_padding): Likewise.
12199 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12200 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12201 (verify_gimple_assign_ternary): Likewise.
12202 * tree-inline.c (estimate_move_cost): Likewise.
12203 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12204 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12205 (get_address_cost_ainc): Likewise.
12206 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12207 (vect_supportable_dr_alignment): Likewise.
12208 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12209 (vectorizable_reduction): Likewise.
12210 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12211 (vectorizable_operation, vectorizable_load): Likewise.
12212 * tree.c (build_same_sized_truth_vector_type): Likewise.
12213 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12214 * var-tracking.c (emit_note_insn_var_location): Likewise.
12215 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12216 (ADDR_VEC_ALIGN): Likewise.
12218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12219 Alan Hayward <alan.hayward@arm.com>
12220 David Sherwood <david.sherwood@arm.com>
12222 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12224 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12225 or if measurement_type is polynomial.
12226 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12227 * combine.c (make_extraction): Likewise.
12228 * dse.c (find_shift_sequence): Likewise.
12229 * dwarf2out.c (mem_loc_descriptor): Likewise.
12230 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12231 (extract_bit_field, extract_low_bits): Likewise.
12232 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12233 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12234 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12235 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12236 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12237 * reload.c (find_reloads): Likewise.
12238 * reload1.c (alter_reg): Likewise.
12239 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12240 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12241 * tree-if-conv.c (predicate_mem_writes): Likewise.
12242 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12243 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12244 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12245 * valtrack.c (dead_debug_insert_temp): Likewise.
12246 * varasm.c (mergeable_constant_section): Likewise.
12247 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12250 Alan Hayward <alan.hayward@arm.com>
12251 David Sherwood <david.sherwood@arm.com>
12253 * expr.c (expand_assignment): Cope with polynomial mode sizes
12254 when assigning to a CONCAT.
12256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12257 Alan Hayward <alan.hayward@arm.com>
12258 David Sherwood <david.sherwood@arm.com>
12260 * machmode.h (mode_precision): Change from unsigned short to
12262 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12264 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12265 or if measurement_type is not polynomial.
12266 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12267 in which the mode is already known to be a scalar_int_mode.
12268 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12269 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12271 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12272 for GET_MODE_PRECISION.
12273 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12274 for GET_MODE_PRECISION.
12275 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12277 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12278 (expand_field_assignment, make_extraction): Likewise.
12279 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12280 (get_last_value): Likewise.
12281 * convert.c (convert_to_integer_1): Likewise.
12282 * cse.c (cse_insn): Likewise.
12283 * expr.c (expand_expr_real_1): Likewise.
12284 * lra-constraints.c (simplify_operand_subreg): Likewise.
12285 * optabs-query.c (can_atomic_load_p): Likewise.
12286 * optabs.c (expand_atomic_load): Likewise.
12287 (expand_atomic_store): Likewise.
12288 * ree.c (combine_reaching_defs): Likewise.
12289 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12290 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12291 * tree.h (type_has_mode_precision_p): Likewise.
12292 * ubsan.c (instrument_si_overflow): Likewise.
12294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12295 Alan Hayward <alan.hayward@arm.com>
12296 David Sherwood <david.sherwood@arm.com>
12298 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12299 polynomial numbers of units.
12300 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12301 (valid_vector_subparts_p): New function.
12302 (build_vector_type): Remove temporary shim and take the number
12303 of units as a poly_uint64 rather than an int.
12304 (build_opaque_vector_type): Take the number of units as a
12305 poly_uint64 rather than an int.
12306 * tree.c (build_vector_from_ctor): Handle polynomial
12307 TYPE_VECTOR_SUBPARTS.
12308 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12309 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12310 (build_vector_from_val): If the number of units is variable,
12311 use build_vec_duplicate_cst for constant operands and
12312 VEC_DUPLICATE_EXPR otherwise.
12313 (make_vector_type): Remove temporary is_constant ().
12314 (build_vector_type, build_opaque_vector_type): Take the number of
12315 units as a poly_uint64 rather than an int.
12316 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12318 * cfgexpand.c (expand_debug_expr): Likewise.
12319 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12320 (store_constructor, expand_expr_real_1): Likewise.
12321 (const_scalar_mask_from_tree): Likewise.
12322 * fold-const-call.c (fold_const_reduction): Likewise.
12323 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12324 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12325 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12326 (fold_relational_const): Likewise.
12327 (native_interpret_vector): Likewise. Change the size from an
12328 int to an unsigned int.
12329 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12330 TYPE_VECTOR_SUBPARTS.
12331 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12332 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12333 duplicating a non-constant operand into a variable-length vector.
12334 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12335 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12336 * ipa-icf.c (sem_variable::equals): Likewise.
12337 * match.pd: Likewise.
12338 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12339 * print-tree.c (print_node): Likewise.
12340 * stor-layout.c (layout_type): Likewise.
12341 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12342 * tree-cfg.c (verify_gimple_comparison): Likewise.
12343 (verify_gimple_assign_binary): Likewise.
12344 (verify_gimple_assign_ternary): Likewise.
12345 (verify_gimple_assign_single): Likewise.
12346 * tree-pretty-print.c (dump_generic_node): Likewise.
12347 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12348 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12349 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12350 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12351 (vect_shift_permute_load_chain): Likewise.
12352 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12353 (expand_vector_condition, optimize_vector_constructor): Likewise.
12354 (lower_vec_perm, get_compute_type): Likewise.
12355 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12356 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12357 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12358 (vect_recog_mask_conversion_pattern): Likewise.
12359 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12360 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12361 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12362 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12363 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12364 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12365 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12366 (supportable_widening_operation): Likewise.
12367 (supportable_narrowing_operation): Likewise.
12368 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12370 * varasm.c (output_constant): Likewise.
12372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12373 Alan Hayward <alan.hayward@arm.com>
12374 David Sherwood <david.sherwood@arm.com>
12376 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12377 so that both the length == 3 and length != 3 cases set up their
12378 own permute vectors. Add comments explaining why we know the
12379 number of elements is constant.
12380 (vect_permute_load_chain): Likewise.
12382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12383 Alan Hayward <alan.hayward@arm.com>
12384 David Sherwood <david.sherwood@arm.com>
12386 * machmode.h (mode_nunits): Change from unsigned char to
12388 (ONLY_FIXED_SIZE_MODES): New macro.
12389 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12390 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12391 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12393 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12394 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12395 or if measurement_type is not polynomial.
12396 * genmodes.c (ZERO_COEFFS): New macro.
12397 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12399 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12400 Use ZERO_COEFFS when emitting initializers.
12401 * data-streamer.h (bp_pack_poly_value): New function.
12402 (bp_unpack_poly_value): Likewise.
12403 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12404 for GET_MODE_NUNITS.
12405 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12406 for GET_MODE_NUNITS.
12407 * tree.c (make_vector_type): Remove temporary shim and make
12408 the real function take the number of units as a poly_uint64
12409 rather than an int.
12410 (build_vector_type_for_mode): Handle polynomial nunits.
12411 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12412 * emit-rtl.c (const_vec_series_p_1): Likewise.
12413 (gen_rtx_CONST_VECTOR): Likewise.
12414 * fold-const.c (test_vec_duplicate_folding): Likewise.
12415 * genrecog.c (validate_pattern): Likewise.
12416 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12417 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12418 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12419 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12420 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12421 * rtlanal.c (subreg_get_info): Likewise.
12422 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12423 (vect_grouped_load_supported): Likewise.
12424 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12425 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12426 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12427 (simplify_const_unary_operation, simplify_binary_operation_1)
12428 (simplify_const_binary_operation, simplify_ternary_operation)
12429 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12430 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12431 instead of CONST_VECTOR_NUNITS.
12432 * varasm.c (output_constant_pool_2): Likewise.
12433 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12434 explicit-encoded elements in the XVEC for variable-length vectors.
12436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12438 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12441 Alan Hayward <alan.hayward@arm.com>
12442 David Sherwood <david.sherwood@arm.com>
12444 * coretypes.h (fixed_size_mode): Declare.
12445 (fixed_size_mode_pod): New typedef.
12446 * builtins.h (target_builtins::x_apply_args_mode)
12447 (target_builtins::x_apply_result_mode): Change type to
12448 fixed_size_mode_pod.
12449 * builtins.c (apply_args_size, apply_result_size, result_vector)
12450 (expand_builtin_apply_args_1, expand_builtin_apply)
12451 (expand_builtin_return): Update accordingly.
12453 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12455 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12456 * cselib.c (cselib_hash_rtx): Likewise.
12457 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12458 CONST_VECTOR encoding.
12460 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12461 Jeff Law <law@redhat.com>
12464 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12465 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12466 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12467 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12470 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12471 explicitly probe *sp in a noreturn function if there were any callee
12472 register saves or frame pointer is needed.
12474 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12477 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12478 BLKmode for ternary, binary or unary expressions.
12481 * var-tracking.c (delete_vta_debug_insn): New inline function.
12482 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12483 insns from get_insns () to NULL instead of each bb separately.
12484 Use delete_vta_debug_insn. No longer static.
12485 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12486 delete_vta_debug_insns callers.
12487 * rtl.h (delete_vta_debug_insns): Declare.
12488 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12489 instead of variable_tracking_main.
12491 2018-01-03 Martin Sebor <msebor@redhat.com>
12493 PR tree-optimization/83603
12494 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12495 arguments past the endof the argument list in functions declared
12496 without a prototype.
12497 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12498 Avoid checking when arguments are null.
12500 2018-01-03 Martin Sebor <msebor@redhat.com>
12503 * doc/extend.texi (attribute const): Fix a typo.
12504 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12505 issuing -Wsuggest-attribute for void functions.
12507 2018-01-03 Martin Sebor <msebor@redhat.com>
12509 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12510 offset_int::from instead of wide_int::to_shwi.
12511 (maybe_diag_overlap): Remove assertion.
12512 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12513 * gimple-ssa-sprintf.c (format_directive): Same.
12514 (parse_directive): Same.
12515 (sprintf_dom_walker::compute_format_length): Same.
12516 (try_substitute_return_value): Same.
12518 2018-01-03 Jeff Law <law@redhat.com>
12520 PR middle-end/83654
12521 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12522 non-constant residual for zero at runtime and avoid probing in
12523 that case. Reorganize code for trailing problem to mirror handling
12526 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12528 PR tree-optimization/83501
12529 * tree-ssa-strlen.c (get_string_cst): New.
12530 (handle_char_store): Call get_string_cst.
12532 2018-01-03 Martin Liska <mliska@suse.cz>
12534 PR tree-optimization/83593
12535 * tree-ssa-strlen.c: Include tree-cfg.h.
12536 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12537 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12538 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12540 (strlen_dom_walker::before_dom_children): Call
12541 gimple_purge_dead_eh_edges. Dump tranformation with details
12543 (strlen_dom_walker::before_dom_children): Update call by adding
12544 new argument cleanup_eh.
12545 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12547 2018-01-03 Martin Liska <mliska@suse.cz>
12550 * cif-code.def (VARIADIC_THUNK): New enum value.
12551 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12554 2018-01-03 Jan Beulich <jbeulich@suse.com>
12556 * sse.md (mov<mode>_internal): Tighten condition for when to use
12557 vmovdqu<ssescalarsize> for TI and OI modes.
12559 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12561 Update copyright years.
12563 2018-01-03 Martin Liska <mliska@suse.cz>
12566 * ipa-visibility.c (function_and_variable_visibility): Skip
12567 functions with noipa attribure.
12569 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12571 * gcc.c (process_command): Update copyright notice dates.
12572 * gcov-dump.c (print_version): Ditto.
12573 * gcov.c (print_version): Ditto.
12574 * gcov-tool.c (print_version): Ditto.
12575 * gengtype.c (create_file): Ditto.
12576 * doc/cpp.texi: Bump @copying's copyright year.
12577 * doc/cppinternals.texi: Ditto.
12578 * doc/gcc.texi: Ditto.
12579 * doc/gccint.texi: Ditto.
12580 * doc/gcov.texi: Ditto.
12581 * doc/install.texi: Ditto.
12582 * doc/invoke.texi: Ditto.
12584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12586 * vector-builder.h (vector_builder::m_full_nelts): Change from
12587 unsigned int to poly_uint64.
12588 (vector_builder::full_nelts): Update prototype accordingly.
12589 (vector_builder::new_vector): Likewise.
12590 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12591 (vector_builder::operator ==): Likewise.
12592 (vector_builder::finalize): Likewise.
12593 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12594 Take the number of elements as a poly_uint64 rather than an
12596 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12597 from unsigned int to poly_uint64.
12598 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12599 (vec_perm_indices::new_vector): Likewise.
12600 (vec_perm_indices::length): Likewise.
12601 (vec_perm_indices::nelts_per_input): Likewise.
12602 (vec_perm_indices::input_nelts): Likewise.
12603 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12604 number of elements per input as a poly_uint64 rather than an
12605 unsigned int. Use the original encoding for variable-length
12606 vectors, rather than clamping each individual element.
12607 For the second and subsequent elements in each pattern,
12608 clamp the step and base before clamping their sum.
12609 (vec_perm_indices::series_p): Handle polynomial element counts.
12610 (vec_perm_indices::all_in_range_p): Likewise.
12611 (vec_perm_indices_to_tree): Likewise.
12612 (vec_perm_indices_to_rtx): Likewise.
12613 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12614 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12615 (tree_vector_builder::new_binary_operation): Handle polynomial
12616 element counts. Return false if we need to know the number
12617 of elements at compile time.
12618 * fold-const.c (fold_vec_perm): Punt if the number of elements
12619 isn't known at compile time.
12621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12623 * vec-perm-indices.h (vec_perm_builder): Change element type
12624 from HOST_WIDE_INT to poly_int64.
12625 (vec_perm_indices::element_type): Update accordingly.
12626 (vec_perm_indices::clamp): Handle polynomial element_types.
12627 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12628 (vec_perm_indices::all_in_range_p): Likewise.
12629 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12631 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12632 polynomial vec_perm_indices element types.
12633 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12634 * fold-const.c (fold_vec_perm): Likewise.
12635 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12636 * tree-vect-generic.c (lower_vec_perm): Likewise.
12637 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12638 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12639 element type to HOST_WIDE_INT.
12641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12642 Alan Hayward <alan.hayward@arm.com>
12643 David Sherwood <david.sherwood@arm.com>
12645 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12646 rather than an int. Use plus_constant.
12647 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12648 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12651 Alan Hayward <alan.hayward@arm.com>
12652 David Sherwood <david.sherwood@arm.com>
12654 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12655 a HOST_WIDE_INT to a poly_int64.
12657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12658 Alan Hayward <alan.hayward@arm.com>
12659 David Sherwood <david.sherwood@arm.com>
12661 * calls.c (load_register_parameters): Cope with polynomial
12662 mode sizes. Require a constant size for BLKmode parameters
12663 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12664 forces a parameter to be padded at the lsb end in order to
12665 fill a complete number of words, require the parameter size
12666 to be ordered wrt UNITS_PER_WORD.
12668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12669 Alan Hayward <alan.hayward@arm.com>
12670 David Sherwood <david.sherwood@arm.com>
12672 * reload1.c (spill_stack_slot_width): Change element type
12673 from unsigned int to poly_uint64_pod.
12674 (alter_reg): Treat mode sizes as polynomial.
12676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12677 Alan Hayward <alan.hayward@arm.com>
12678 David Sherwood <david.sherwood@arm.com>
12680 * reload.c (complex_word_subreg_p): New function.
12681 (reload_inner_reg_of_subreg, push_reload): Use it.
12683 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12684 Alan Hayward <alan.hayward@arm.com>
12685 David Sherwood <david.sherwood@arm.com>
12687 * lra-constraints.c (process_alt_operands): Reject matched
12688 operands whose sizes aren't ordered.
12689 (match_reload): Refer to this check here.
12691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12692 Alan Hayward <alan.hayward@arm.com>
12693 David Sherwood <david.sherwood@arm.com>
12695 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12696 that the mode size is in the set {1, 2, 4, 8, 16}.
12698 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12699 Alan Hayward <alan.hayward@arm.com>
12700 David Sherwood <david.sherwood@arm.com>
12702 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12703 Use plus_constant instead of gen_rtx_PLUS.
12705 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12706 Alan Hayward <alan.hayward@arm.com>
12707 David Sherwood <david.sherwood@arm.com>
12709 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12710 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12711 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12712 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12713 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12714 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12715 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12716 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12717 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12718 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12720 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12721 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12722 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12723 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12724 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12725 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12726 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12727 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12728 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12729 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12731 * expr.c (emit_move_resolve_push): Treat the input and result
12732 of PUSH_ROUNDING as a poly_int64.
12733 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12734 (emit_push_insn): Likewise.
12735 * lra-eliminations.c (mark_not_eliminable): Likewise.
12736 * recog.c (push_operand): Likewise.
12737 * reload1.c (elimination_effects): Likewise.
12738 * rtlanal.c (nonzero_bits1): Likewise.
12739 * calls.c (store_one_arg): Likewise. Require the padding to be
12740 known at compile time.
12742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12743 Alan Hayward <alan.hayward@arm.com>
12744 David Sherwood <david.sherwood@arm.com>
12746 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12747 Use plus_constant instead of gen_rtx_PLUS.
12749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12750 Alan Hayward <alan.hayward@arm.com>
12751 David Sherwood <david.sherwood@arm.com>
12753 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12754 rather than an int.
12756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12757 Alan Hayward <alan.hayward@arm.com>
12758 David Sherwood <david.sherwood@arm.com>
12760 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12761 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12762 via stack temporaries. Treat the mode size as polynomial too.
12764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12765 Alan Hayward <alan.hayward@arm.com>
12766 David Sherwood <david.sherwood@arm.com>
12768 * expr.c (expand_expr_real_2): When handling conversions involving
12769 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12770 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12771 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12780 Alan Hayward <alan.hayward@arm.com>
12781 David Sherwood <david.sherwood@arm.com>
12783 * combine.c (can_change_dest_mode): Handle polynomial
12784 REGMODE_NATURAL_SIZE.
12785 * expmed.c (store_bit_field_1): Likewise.
12786 * expr.c (store_constructor): Likewise.
12787 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12788 and polynomial REGMODE_NATURAL_SIZE.
12789 (gen_lowpart_common): Likewise.
12790 * reginfo.c (record_subregs_of_mode): Likewise.
12791 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12798 numbers of elements.
12800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12801 Alan Hayward <alan.hayward@arm.com>
12802 David Sherwood <david.sherwood@arm.com>
12804 * match.pd: Cope with polynomial numbers of vector elements.
12806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12807 Alan Hayward <alan.hayward@arm.com>
12808 David Sherwood <david.sherwood@arm.com>
12810 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12811 in a POINTER_PLUS_EXPR.
12813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12814 Alan Hayward <alan.hayward@arm.com>
12815 David Sherwood <david.sherwood@arm.com>
12817 * omp-simd-clone.c (simd_clone_subparts): New function.
12818 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12819 (ipa_simd_modify_function_body): Likewise.
12821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12822 Alan Hayward <alan.hayward@arm.com>
12823 David Sherwood <david.sherwood@arm.com>
12825 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12826 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12827 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12828 (expand_vector_condition, vector_element): Likewise.
12829 (subparts_gt): New function.
12830 (get_compute_type): Use subparts_gt.
12831 (count_type_subparts): Delete.
12832 (expand_vector_operations_1): Use subparts_gt instead of
12833 count_type_subparts.
12835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12836 Alan Hayward <alan.hayward@arm.com>
12837 David Sherwood <david.sherwood@arm.com>
12839 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12840 (vect_compile_time_alias): ...this new function. Do the calculation
12841 on poly_ints rather than trees.
12842 (vect_prune_runtime_alias_test_list): Update call accordingly.
12844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12845 Alan Hayward <alan.hayward@arm.com>
12846 David Sherwood <david.sherwood@arm.com>
12848 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12850 (vect_schedule_slp_instance): Likewise.
12852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12853 Alan Hayward <alan.hayward@arm.com>
12854 David Sherwood <david.sherwood@arm.com>
12856 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12857 constant and extern definitions for variable-length vectors.
12858 (vect_get_constant_vectors): Note that the number of units
12859 is known to be constant.
12861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12862 Alan Hayward <alan.hayward@arm.com>
12863 David Sherwood <david.sherwood@arm.com>
12865 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12866 of units as polynomial. Choose between WIDE and NARROW based
12869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12870 Alan Hayward <alan.hayward@arm.com>
12871 David Sherwood <david.sherwood@arm.com>
12873 * tree-vect-stmts.c (simd_clone_subparts): New function.
12874 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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-stmts.c (vectorizable_call): Treat the number of
12881 vectors as polynomial. Use build_index_vector for
12882 IFN_GOMP_SIMD_LANE.
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-stmts.c (get_load_store_type): Treat the number of
12889 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12890 for variable-length vectors.
12891 (vectorizable_mask_load_store): Treat the number of units as
12892 polynomial, asserting that it is constant if the condition has
12893 already been enforced.
12894 (vectorizable_store, vectorizable_load): Likewise.
12896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12897 Alan Hayward <alan.hayward@arm.com>
12898 David Sherwood <david.sherwood@arm.com>
12900 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12901 of units as polynomial. Punt if we can't tell at compile time
12902 which vector contains the final result.
12904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12905 Alan Hayward <alan.hayward@arm.com>
12906 David Sherwood <david.sherwood@arm.com>
12908 * tree-vect-loop.c (vectorizable_induction): Treat the number
12909 of units as polynomial. Punt on SLP inductions. Use an integer
12910 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12911 cast of such a series for variable-length floating-point
12914 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12915 Alan Hayward <alan.hayward@arm.com>
12916 David Sherwood <david.sherwood@arm.com>
12918 * tree.h (build_index_vector): Declare.
12919 * tree.c (build_index_vector): New function.
12920 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12921 of units as polynomial, forcibly converting it to a constant if
12922 vectorizable_reduction has already enforced the condition.
12923 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12924 to create a {1,2,3,...} vector.
12925 (vectorizable_reduction): Treat the number of units as polynomial.
12926 Choose vectype_in based on the largest scalar element size rather
12927 than the smallest number of units. Enforce the restrictions
12930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12931 Alan Hayward <alan.hayward@arm.com>
12932 David Sherwood <david.sherwood@arm.com>
12934 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12935 number of units as polynomial.
12937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12938 Alan Hayward <alan.hayward@arm.com>
12939 David Sherwood <david.sherwood@arm.com>
12941 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12942 * target.def (autovectorize_vector_sizes): Return the vector sizes
12943 by pointer, using vector_sizes rather than a bitmask.
12944 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12945 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12946 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12948 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12949 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12950 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12951 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12952 * omp-general.c (omp_max_vf): Likewise.
12953 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12954 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12955 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12956 * tree-vect-slp.c (vect_slp_bb): Likewise.
12957 * doc/tm.texi: Regenerate.
12958 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12960 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12961 the vector size as a poly_uint64 rather than an unsigned int.
12962 (current_vector_size): Change from an unsigned int to a poly_uint64.
12963 (get_vectype_for_scalar_type): Update accordingly.
12964 * tree.h (build_truth_vector_type): Take the size and number of
12965 units as a poly_uint64 rather than an unsigned int.
12966 (build_vector_type): Add a temporary overload that takes
12967 the number of units as a poly_uint64 rather than an unsigned int.
12968 * tree.c (make_vector_type): Likewise.
12969 (build_truth_vector_type): Take the number of units as a poly_uint64
12970 rather than an unsigned int.
12972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12973 Alan Hayward <alan.hayward@arm.com>
12974 David Sherwood <david.sherwood@arm.com>
12976 * target.def (get_mask_mode): Take the number of units and length
12977 as poly_uint64s rather than unsigned ints.
12978 * targhooks.h (default_get_mask_mode): Update accordingly.
12979 * targhooks.c (default_get_mask_mode): Likewise.
12980 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12981 * doc/tm.texi: Regenerate.
12983 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12984 Alan Hayward <alan.hayward@arm.com>
12985 David Sherwood <david.sherwood@arm.com>
12987 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12988 * omp-general.c (omp_max_vf): Likewise.
12989 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12990 (expand_omp_simd): Handle polynomial safelen.
12991 * omp-low.c (omplow_simd_context): Add a default constructor.
12992 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12993 (lower_rec_simd_input_clauses): Update accordingly.
12994 (lower_rec_input_clauses): Likewise.
12996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12997 Alan Hayward <alan.hayward@arm.com>
12998 David Sherwood <david.sherwood@arm.com>
13000 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13001 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13002 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13003 (vect_analyze_slp_cost): Likewise.
13004 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13005 (vect_model_load_cost): Likewise.
13007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13008 Alan Hayward <alan.hayward@arm.com>
13009 David Sherwood <david.sherwood@arm.com>
13011 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13012 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13013 from an unsigned int * to a poly_uint64_pod *.
13014 (calculate_unrolling_factor): New function.
13015 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13018 Alan Hayward <alan.hayward@arm.com>
13019 David Sherwood <david.sherwood@arm.com>
13021 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13022 from an unsigned int to a poly_uint64.
13023 (_loop_vec_info::slp_unrolling_factor): Likewise.
13024 (_loop_vec_info::vectorization_factor): Change from an int
13026 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13027 (vect_get_num_vectors): New function.
13028 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13029 (vect_get_num_copies): Use vect_get_num_vectors.
13030 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13031 to an unsigned int *.
13032 (vect_analyze_data_refs): Change min_vf from an int * to a
13034 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13035 than an unsigned HOST_WIDE_INT.
13036 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13037 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13038 to an unsigned int *.
13039 (vect_analyze_data_ref_dependences): Likewise.
13040 (vect_compute_data_ref_alignment): Handle polynomial vf.
13041 (vect_enhance_data_refs_alignment): Likewise.
13042 (vect_prune_runtime_alias_test_list): Likewise.
13043 (vect_shift_permute_load_chain): Likewise.
13044 (vect_supportable_dr_alignment): Likewise.
13045 (dependence_distance_ge_vf): Take the vectorization factor as a
13046 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13047 (vect_analyze_data_refs): Change min_vf from an int * to a
13049 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13050 vfm1 as a poly_uint64 rather than an int. Make the same change
13051 for the returned bound_scalar.
13052 (vect_gen_vector_loop_niters): Handle polynomial vf.
13053 (vect_do_peeling): Likewise. Update call to
13054 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13055 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13057 * tree-vect-loop.c (vect_determine_vectorization_factor)
13058 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13059 (vect_get_known_peeling_cost): Likewise.
13060 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13061 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13062 (vect_transform_loop): Likewise. Use the lowest possible VF when
13063 updating the upper bounds of the loop.
13064 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13065 rather than an int.
13066 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13067 polynomial unroll factors.
13068 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13069 (vect_make_slp_decision): Likewise.
13070 (vect_supported_load_permutation_p): Likewise, and polynomial
13072 (vect_analyze_slp_cost): Handle polynomial vf.
13073 (vect_slp_analyze_node_operations): Likewise.
13074 (vect_slp_analyze_bb_1): Likewise.
13075 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13076 than an unsigned HOST_WIDE_INT.
13077 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13078 (vectorizable_load): Handle polynomial vf.
13079 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13081 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13084 Alan Hayward <alan.hayward@arm.com>
13085 David Sherwood <david.sherwood@arm.com>
13087 * match.pd: Handle bit operations involving three constants
13088 and try to fold one pair.
13090 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13092 * tree-vect-loop-manip.c: Include gimple-fold.h.
13093 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13094 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13095 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13096 Add a path that uses a step of VF instead of 1, but disable it
13098 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13099 and niters_no_overflow parameters. Update calls to
13100 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13101 Create a new SSA name if the latter choses to use a ste other
13102 than zero, and return it via niters_vector_mult_vf_var.
13103 * tree-vect-loop.c (vect_transform_loop): Update calls to
13104 vect_do_peeling, vect_gen_vector_loop_niters and
13105 slpeel_make_loop_iterate_ntimes.
13106 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13107 (vect_gen_vector_loop_niters): Update declarations after above changes.
13109 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13111 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13112 128-bit round to integer instructions.
13113 (ceil<mode>2): Likewise.
13114 (btrunc<mode>2): Likewise.
13115 (round<mode>2): Likewise.
13117 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13119 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13120 unaligned VSX load/store on P8/P9.
13121 (expand_block_clear): Allow the use of unaligned VSX
13122 load/store on P8/P9.
13124 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13126 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13128 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13129 swap associated with both a load and a store.
13131 2018-01-02 Andrew Waterman <andrew@sifive.com>
13133 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13134 * config/riscv/riscv.md (clear_cache): Use it.
13136 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13138 * web.c: Remove out-of-date comment.
13140 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13142 * expr.c (fixup_args_size_notes): Check that any existing
13143 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13144 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13145 (emit_single_push_insn): ...here.
13147 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13149 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13150 (const_vector_encoded_nelts): New function.
13151 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13152 (const_vector_int_elt, const_vector_elt): Declare.
13153 * emit-rtl.c (const_vector_int_elt_1): New function.
13154 (const_vector_elt): Likewise.
13155 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13156 of CONST_VECTOR_ELT.
13158 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13160 * expr.c: Include rtx-vector-builder.h.
13161 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13162 directly on the tree encoding.
13163 (const_vector_from_tree): Likewise.
13164 * optabs.c: Include rtx-vector-builder.h.
13165 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13166 sequence of "u" values.
13167 * vec-perm-indices.c: Include rtx-vector-builder.h.
13168 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13169 directly on the vec_perm_indices encoding.
13171 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13173 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13174 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13175 * rtx-vector-builder.h: New file.
13176 * rtx-vector-builder.c: Likewise.
13177 * rtl.h (rtx_def::u2): Add a const_vector field.
13178 (CONST_VECTOR_NPATTERNS): New macro.
13179 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13180 (CONST_VECTOR_DUPLICATE_P): Likewise.
13181 (CONST_VECTOR_STEPPED_P): Likewise.
13182 (CONST_VECTOR_ENCODED_ELT): Likewise.
13183 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13184 (unwrap_const_vec_duplicate): Likewise.
13185 (const_vec_series_p): Check for a non-duplicated vector encoding.
13186 Say that the function only returns true for integer vectors.
13187 * emit-rtl.c: Include rtx-vector-builder.h.
13188 (gen_const_vec_duplicate_1): Delete.
13189 (gen_const_vector): Call gen_const_vec_duplicate instead of
13190 gen_const_vec_duplicate_1.
13191 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13192 (gen_const_vec_duplicate): Use rtx_vector_builder.
13193 (gen_const_vec_series): Likewise.
13194 (gen_rtx_CONST_VECTOR): Likewise.
13195 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13196 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13197 Build a new vector rather than modifying a CONST_VECTOR in-place.
13198 (handle_special_swappables): Update call accordingly.
13199 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13200 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13201 Build a new vector rather than modifying a CONST_VECTOR in-place.
13202 (handle_special_swappables): Update call accordingly.
13204 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13206 * simplify-rtx.c (simplify_const_binary_operation): Use
13207 CONST_VECTOR_ELT instead of XVECEXP.
13209 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13211 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13212 the selector elements to be different from the data elements
13213 if the selector is a VECTOR_CST.
13214 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13215 ssizetype for the selector.
13217 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13219 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13220 before testing each element individually.
13221 * tree-vect-generic.c (lower_vec_perm): Likewise.
13223 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13225 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13226 * selftest-run-tests.c (selftest::run_tests): Call it.
13227 * vector-builder.h (vector_builder::operator ==): New function.
13228 (vector_builder::operator !=): Likewise.
13229 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13230 (vec_perm_indices::all_from_input_p): New function.
13231 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13232 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13233 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13234 instead of reading the VECTOR_CST directly. Detect whether both
13235 vector inputs are the same before constructing the vec_perm_indices,
13236 and update the number of inputs argument accordingly. Use the
13237 utility functions added above. Only construct sel2 if we need to.
13239 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13241 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13242 the broadcast of the low byte.
13243 (expand_mult_highpart): Use an explicit encoding for the permutes.
13244 * optabs-query.c (can_mult_highpart_p): Likewise.
13245 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13246 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13247 (vectorizable_bswap): Likewise.
13248 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13249 explicit encoding for the power-of-2 permutes.
13250 (vect_permute_store_chain): Likewise.
13251 (vect_grouped_load_supported): Likewise.
13252 (vect_permute_load_chain): Likewise.
13254 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13256 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13257 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13258 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13259 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13260 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13261 (vect_gen_perm_mask_any): Likewise.
13263 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13265 * int-vector-builder.h: New file.
13266 * vec-perm-indices.h: Include int-vector-builder.h.
13267 (vec_perm_indices): Redefine as an int_vector_builder.
13268 (auto_vec_perm_indices): Delete.
13269 (vec_perm_builder): Redefine as a stand-alone class.
13270 (vec_perm_indices::vec_perm_indices): New function.
13271 (vec_perm_indices::clamp): Likewise.
13272 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13273 (vec_perm_indices::new_vector): New function.
13274 (vec_perm_indices::new_expanded_vector): Update for new
13275 vec_perm_indices class.
13276 (vec_perm_indices::rotate_inputs): New function.
13277 (vec_perm_indices::all_in_range_p): Operate directly on the
13278 encoded form, without computing elided elements.
13279 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13280 encoding. Update for new vec_perm_indices class.
13281 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13282 the given vec_perm_builder.
13283 (expand_vec_perm_var): Update vec_perm_builder constructor.
13284 (expand_mult_highpart): Use vec_perm_builder instead of
13285 auto_vec_perm_indices.
13286 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13287 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13288 or double series encoding as appropriate.
13289 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13290 vec_perm_indices instead of auto_vec_perm_indices.
13291 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13292 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13293 (vect_permute_store_chain): Likewise.
13294 (vect_grouped_load_supported): Likewise.
13295 (vect_permute_load_chain): Likewise.
13296 (vect_shift_permute_load_chain): Likewise.
13297 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13298 (vect_transform_slp_perm_load): Likewise.
13299 (vect_schedule_slp_instance): Likewise.
13300 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13301 (vectorizable_mask_load_store): Likewise.
13302 (vectorizable_bswap): Likewise.
13303 (vectorizable_store): Likewise.
13304 (vectorizable_load): Likewise.
13305 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13306 vec_perm_indices instead of auto_vec_perm_indices. Use
13307 tree_to_vec_perm_builder to read the vector from a tree.
13308 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13309 vec_perm_builder instead of a vec_perm_indices.
13310 (have_whole_vector_shift): Use vec_perm_builder and
13311 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13312 truncation to calc_vec_perm_mask_for_shift.
13313 (vect_create_epilog_for_reduction): Likewise.
13314 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13315 from auto_vec_perm_indices to vec_perm_indices.
13316 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13317 instead of changing individual elements.
13318 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13319 the vector in d.perm.
13320 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13321 from auto_vec_perm_indices to vec_perm_indices.
13322 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13323 instead of changing individual elements.
13324 (arm_vectorize_vec_perm_const): Use new_vector to install
13325 the vector in d.perm.
13326 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13327 Update vec_perm_builder constructor.
13328 (rs6000_expand_interleave): Likewise.
13329 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13330 (rs6000_expand_interleave): Likewise.
13332 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13334 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13335 to qimode could truncate the indices.
13336 * optabs.c (expand_vec_perm_var): Likewise.
13338 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13340 * Makefile.in (OBJS): Add vec-perm-indices.o.
13341 * vec-perm-indices.h: New file.
13342 * vec-perm-indices.c: Likewise.
13343 * target.h (vec_perm_indices): Replace with a forward class
13345 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13346 * optabs.h: Include vec-perm-indices.h.
13347 (expand_vec_perm): Delete.
13348 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13349 (expand_vec_perm_const): Declare.
13350 * target.def (vec_perm_const_ok): Replace with...
13351 (vec_perm_const): ...this new hook.
13352 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13353 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13354 * doc/tm.texi: Regenerate.
13355 * optabs.def (vec_perm_const): Delete.
13356 * doc/md.texi (vec_perm_const): Likewise.
13357 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13358 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13359 expand_vec_perm for constant permutation vectors. Assert that
13360 the mode of variable permutation vectors is the integer equivalent
13361 of the mode that is being permuted.
13362 * optabs-query.h (selector_fits_mode_p): Declare.
13363 * optabs-query.c: Include vec-perm-indices.h.
13364 (selector_fits_mode_p): New function.
13365 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13366 is defined, instead of checking whether the vec_perm_const_optab
13367 exists. Use targetm.vectorize.vec_perm_const instead of
13368 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13369 fit in the vector mode before using a variable permute.
13370 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13371 vec_perm_indices instead of an rtx.
13372 (expand_vec_perm): Replace with...
13373 (expand_vec_perm_const): ...this new function. Take the selector
13374 as a vec_perm_indices rather than an rtx. Also take the mode of
13375 the selector. Update call to shift_amt_for_vec_perm_mask.
13376 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13377 Use vec_perm_indices::new_expanded_vector to expand the original
13378 selector into bytes. Check whether the indices fit in the vector
13379 mode before using a variable permute.
13380 (expand_vec_perm_var): Make global.
13381 (expand_mult_highpart): Use expand_vec_perm_const.
13382 * fold-const.c: Includes vec-perm-indices.h.
13383 * tree-ssa-forwprop.c: Likewise.
13384 * tree-vect-data-refs.c: Likewise.
13385 * tree-vect-generic.c: Likewise.
13386 * tree-vect-loop.c: Likewise.
13387 * tree-vect-slp.c: Likewise.
13388 * tree-vect-stmts.c: Likewise.
13389 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13391 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13392 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13393 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13394 (aarch64_vectorize_vec_perm_const): ...this new function.
13395 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13396 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13397 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13398 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13399 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13400 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13401 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13403 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13404 check for NEON modes.
13405 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13406 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13407 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13408 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13410 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13411 the old VEC_PERM_CONST conditions.
13412 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13413 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13414 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13415 (ia64_vectorize_vec_perm_const_ok): Merge into...
13416 (ia64_vectorize_vec_perm_const): ...this new function.
13417 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13418 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13419 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13420 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13421 * config/mips/mips.c (mips_expand_vec_perm_const)
13422 (mips_vectorize_vec_perm_const_ok): Merge into...
13423 (mips_vectorize_vec_perm_const): ...this new function.
13424 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13425 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13426 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13427 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13428 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13429 (rs6000_expand_vec_perm_const): Delete.
13430 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13432 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13433 (altivec_expand_vec_perm_const_le): Take each operand individually.
13434 Operate on constant selectors rather than rtxes.
13435 (altivec_expand_vec_perm_const): Likewise. Update call to
13436 altivec_expand_vec_perm_const_le.
13437 (rs6000_expand_vec_perm_const): Delete.
13438 (rs6000_vectorize_vec_perm_const_ok): Delete.
13439 (rs6000_vectorize_vec_perm_const): New function.
13440 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13441 an element count and rtx array.
13442 (rs6000_expand_extract_even): Update call accordingly.
13443 (rs6000_expand_interleave): Likewise.
13444 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13445 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13446 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13447 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13448 (rs6000_expand_vec_perm_const): Delete.
13449 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13450 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13451 (altivec_expand_vec_perm_const_le): Take each operand individually.
13452 Operate on constant selectors rather than rtxes.
13453 (altivec_expand_vec_perm_const): Likewise. Update call to
13454 altivec_expand_vec_perm_const_le.
13455 (rs6000_expand_vec_perm_const): Delete.
13456 (rs6000_vectorize_vec_perm_const_ok): Delete.
13457 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13458 reference to the SPE evmerge intructions.
13459 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13460 an element count and rtx array.
13461 (rs6000_expand_extract_even): Update call accordingly.
13462 (rs6000_expand_interleave): Likewise.
13463 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13464 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13466 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13468 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13470 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13471 vector mode and that that mode matches the mode of the data
13473 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13474 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13475 directly using expand_vec_perm_1 when forcing selectors into
13477 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13479 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13481 * optabs-query.h (can_vec_perm_p): Delete.
13482 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13483 * optabs-query.c (can_vec_perm_p): Split into...
13484 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13485 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13486 particular selector is valid.
13487 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13488 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13489 (vect_grouped_load_supported): Likewise.
13490 (vect_shift_permute_load_chain): Likewise.
13491 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13492 (vect_transform_slp_perm_load): Likewise.
13493 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13494 (vectorizable_bswap): Likewise.
13495 (vect_gen_perm_mask_checked): Likewise.
13496 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13497 implementations of variable permutation vectors into account
13498 when deciding which selector to use.
13499 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13500 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13501 with a false third argument.
13502 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13503 to test whether the constant selector is valid and can_vec_perm_var_p
13504 to test whether a variable selector is valid.
13506 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13508 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13509 * optabs-query.c (can_vec_perm_p): Likewise.
13510 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13511 instead of vec_perm_indices.
13512 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13513 (vect_gen_perm_mask_checked): Likewise,
13514 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13515 (vect_gen_perm_mask_checked): Likewise,
13517 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13519 * optabs-query.h (qimode_for_vec_perm): Declare.
13520 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13521 (qimode_for_vec_perm): ...this new function.
13522 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13524 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13526 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13527 does not have a conditional at the top.
13529 2018-01-02 Richard Biener <rguenther@suse.de>
13531 * ipa-inline.c (big_speedup_p): Fix expression.
13533 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13536 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13539 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13543 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13544 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13545 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13546 cond_taken_branch_cost 3->4.
13548 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13550 PR tree-optimization/83581
13551 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13552 TODO_cleanup_cfg if any changes have been made.
13554 PR middle-end/83608
13555 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13556 convert_modes if target mode has the right side, but different mode
13559 PR middle-end/83609
13560 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13561 last argument when extracting from CONCAT. If either from_real or
13562 from_imag is NULL, use expansion through memory. If result is not
13563 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13564 the parts directly to inner mode, if even that fails, use expansion
13567 PR middle-end/83623
13568 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13569 check for bswap in mode rather than HImode and use that in expand_unop
13572 Copyright (C) 2018 Free Software Foundation, Inc.
13574 Copying and distribution of this file, with or without modification,
13575 are permitted in any medium without royalty provided the copyright
13576 notice and this notice are preserved.
projects / gcc.git / blob