1 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
4 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
5 (TARGET_PIPELINE_N8): Likewise.
6 (TARGET_PIPELINE_N10): Likewise.
7 (TARGET_PIPELINE_N13): Likewise.
8 (TARGET_PIPELINE_GRAYWOLF): Likewise.
10 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
12 * config/nds32/nds32-fpu.md: Update copyright year.
14 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
16 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
18 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
20 * config/nds32/nds32.c
21 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
22 * config/nds32/nds32.opt (minline-asm-r15): New option.
24 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
26 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
28 * config/nds32/nds32.md (abssi2): New pattern.
30 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
32 * config/i386/i386.md (rex64namesuffix): New mode attribute.
33 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
34 Merge insn pattern from sse_cvtsi2ss<round_name> and
35 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
36 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
37 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
38 using SWI48 mode iterator.
39 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
40 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
41 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
42 pattern from sse_cvttss2si<round_saeonly_name>
43 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
44 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
45 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
46 using SWI48 mode iterator.
47 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
48 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
49 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
50 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
51 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
52 using SWI48 mode iterator.
53 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
54 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
55 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
56 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
57 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
59 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
60 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
61 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
62 pattern from sse_cvttsd2si<round_saeonly_name>
63 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
65 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
67 * config/nds32/nds32-md-auxiliary.c
68 (nds32_valid_smw_lwm_base_p): Refine.
69 (nds32_output_smw_single_word): Refine.
70 (nds32_output_smw_double_word): New.
71 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
73 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
75 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
76 (nds32_output_stack_pop): Refine.
77 (nds32_expand_unaligned_load): Refine.
78 (nds32_expand_unaligned_store): Refine.
80 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
81 Chung-Ju Wu <jasonwucj@gmail.com>
83 * config/nds32/constants.md: Add TP_REGNUM constant.
84 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
85 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
87 * config/nds32/nds32-doubleword.md: Consider flag_pic.
88 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
89 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
90 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
91 and PIC code generation.
92 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
94 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
96 * config/nds32/nds32.md: Support TLS and PIC.
97 * config/nds32/nds32.c: Support TLS and PIC.
98 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
99 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
102 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
104 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
107 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
108 Chung-Ju Wu <jasonwucj@gmail.com>
110 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
111 * config/nds32/nds32-md-auxiliary.c
112 (symbolic_reference_mentioned_p): New.
113 (nds32_legitimize_ict_address): New.
114 (nds32_expand_ict_move): New.
115 (nds32_indirect_call_referenced_p): New.
116 (nds32_symbol_binds_local_p): Delete.
117 (nds32_long_call_p): Modify.
118 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
119 * config/nds32/nds32-protos.h
120 (symbolic_reference_mentioned_p): Declare.
121 (nds32_legitimize_ict_address): Declare.
122 (nds32_expand_ict_move): Declare.
123 (nds32_indirect_call_referenced_p): Declare.
124 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
125 (nds32_relax_group): Use nds32_ict_const_p as condition.
126 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
127 (nds32_asm_file_start): Output ict_model directive in asm code.
128 (nds32_legitimate_address_p): Consider indirect call.
129 (nds32_print_operand): Consider indirect call.
130 (nds32_print_operand_address): Consider indirect call.
131 (nds32_insert_attributes): Handle "indirect_call" attribute.
132 (TARGET_LEGITIMATE_ADDRESS_P): Define.
133 (TARGET_LEGITIMATE_CONSTANT_P): Define.
134 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
135 (TARGET_DELEGITIMIZE_ADDRESS): Define.
136 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
137 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
138 (TARGET_ICT_MODEL_SMALL): Define.
139 (TARGET_ICT_MODEL_LARGE): Define.
140 * config/nds32/nds32.md (movsi): Consider ict model.
141 (call, call_value): Consider ict model.
142 (sibcall, sibcall_value): Consider ict model.
143 * config/nds32/nds32.opt (mict-model): New option.
144 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
147 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
148 Monk Chiang <sh.chiang04@gmail.com>
149 Jim Wilson <jimw@sifive.com>
151 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
152 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
153 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
154 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
155 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
156 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
157 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
158 compute save_libcall_adjustment properly.
159 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
160 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
161 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
162 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
163 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
164 (ABI_SPEC): Handle mabi=ilp32e.
165 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
167 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
168 <-march>: Add rv32e as an example.
170 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
173 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
174 (intra_create_variable_infos): Handle C++ constructors.
176 2018-05-18 Martin Liska <mliska@suse.cz>
178 * passes.def: Remove a redundant pass.
180 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
183 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
185 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
187 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
189 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
191 (armv2, armv2a, armv3, armv3m): Delete architectures.
192 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
193 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
194 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
196 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
197 (*mulsidi3adddi): Likewise.
198 (mulsidi3): Likewise.
199 (*mulsidi3_nov6): Likewise.
200 (umulsidi3): Likewise.
201 (umulsidi3_nov6): Likewise.
202 (umaddsidi4): Likewise.
203 (*umulsidi3adddi): Likewise.
204 (smulsi3_highpart): Likewise.
205 (*smulsi3_highpart_nov6): Likewise.
206 (umulsi3_highpart): Likewise.
207 (*umulsi3_highpart_nov6): Likewise.
208 * config/arm/arm.h (arm_arch3m): Delete.
209 * config/arm/arm.c (arm_arch3m): Delete.
210 (arm_option_override_internal): Update armv3-related comment.
211 (arm_configure_build_target): Delete use of isa_bit_mode32.
212 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
213 (arm_rtx_costs_internal): Delete check of arm_arch3m.
214 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
217 * config/arm/arm-protos.h (arm_arch3m): Delete.
218 * config/arm/arm-tables.opt: Regenerate.
219 * config/arm/arm-tune.md: Likewise.
220 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
221 deleted architectures.
223 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
225 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
226 (armv5t, armv5te): New features.
227 (ARMv5, ARMv5e): Delete fgroups.
228 (ARMv5t, ARMv5te): Adjust for above changes.
230 (armv5, armv5e): Delete arches.
231 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
233 (*call_reg_arm): Likewise.
234 (*call_value_reg_armv5): Likewise.
235 (*call_value_reg_arm): Likewise.
236 (*call_symbol): Likewise.
237 (*call_value_symbol): Likewise.
238 (*sibcall_insn): Likewise.
239 (*sibcall_value_insn): Likewise.
241 (prefetch): Likewise.
242 (define_split and define_peephole2 dependent on arm_arch5):
244 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
246 (TARGET_ARM_QBIT): Likewise.
247 (TARGET_DSP_MULTIPLY): Likewise.
248 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
249 (arm_arch5, arm_arch5e): Delete.
250 (arm_arch5t, arm_arch5te): Declare.
251 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
252 (arm_arch5t): Declare.
253 (arm_option_reconfigure_globals): Update for the above.
254 (arm_options_perform_arch_sanity_checks): Update comment, replace
255 use of arm_arch5 with arm_arch5t.
256 (use_return_insn): Likewise.
257 (arm_emit_call_insn): Likewise.
258 (output_return_instruction): Likewise.
259 (arm_final_prescan_insn): Likewise.
260 (arm_coproc_builtin_available): Likewise.
261 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
262 arm_arch5e with arm_arch5t and arm_arch5te.
263 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
264 (arm_arch5t, arm_arch5te): Declare.
265 * config/arm/arm-tables.opt: Regenerate.
266 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
267 * config/arm/t-multilib: Likewise.
268 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
269 instead of arm_arch5.
270 (*call_reg_thumb1): Likewise.
271 (*call_value_reg_thumb1_v5): Likewise.
272 (*call_value_reg_thumb1): Likewise.
273 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
275 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
277 2018-05-18 Martin Liska <mliska@suse.cz>
279 PR gcov-profile/84846
280 * doc/gcov.texi: Document -t option of gcov tool.
282 2018-05-18 Martin Liska <mliska@suse.cz>
284 PR gcov-profile/84846
285 * gcov.c (print_usage): Add new -t option.
286 (process_args): Handle the option.
287 (generate_results): Use stdout as output when requested by
290 2018-05-18 Martin Liska <mliska@suse.cz>
292 PR gcov-profile/84846
293 * coverage.c (coverage_init): Write PWD to .gcno file.
294 * doc/gcov.texi: Document how working directory is printed.
295 * gcov-dump.c (dump_gcov_file): Print PWD.
296 * gcov.c (output_intermediate_file): Likewise.
297 (read_graph_file): Read PWD string.
298 (output_lines): Print PWD.
300 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
303 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
304 for retval and return false if all args to phi are zero.
306 2018-05-18 Richard Biener <rguenther@suse.de>
308 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
310 (evrp_dom_walker::before_dom_children): Call it.
312 2018-05-18 Richard Biener <rguenther@suse.de>
314 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
315 results when processing array refs with variable index.
317 2018-05-18 Toon Moene <toon@moene.org>
319 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
320 directly after that of -floop-interchange. Indicate that both
321 options are enabled by default when specifying -O3.
323 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
325 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
326 iterator. Delete separate integer-mode vec_set<mode> expander.
327 (aarch64_simd_vec_setv2di): Delete.
328 (vec_setv2di): Delete.
329 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
330 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
331 the "w, r" alternative.
333 2018-05-18 Martin Liska <mliska@suse.cz>
335 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
336 * tree-pass.h (make_pass_lower_switch_O0): New function.
337 * tree-switch-conversion.c (node_has_low_bound): Remove.
338 (node_has_high_bound): Likewise.
339 (node_is_bounded): Likewise.
340 (class pass_lower_switch): Make it a template type and create
342 (pass_lower_switch::execute): Add template argument.
343 (make_pass_lower_switch): New function.
344 (make_pass_lower_switch_O0): New function.
345 (do_jump_if_equal): Remove.
346 (emit_case_nodes): Simplify to just handle all 3 cases and leave
347 all the hard work to tree optimization passes.
349 2018-05-18 Martin Liska <mliska@suse.cz>
351 * dbgcnt.c (limit_low): Renamed from limit.
352 (limit_high): New variable.
353 (dbg_cnt_is_enabled): Check for upper limit.
354 (dbg_cnt): Adjust dumping.
355 (dbg_cnt_set_limit_by_index): Add new argument for high
357 (dbg_cnt_set_limit_by_name): Likewise.
358 (dbg_cnt_process_single_pair): Parse new format.
359 (dbg_cnt_process_opt): Use strtok.
360 (dbg_cnt_list_all_counters): Remove 'value' and add
362 * doc/invoke.texi: Document changes.
364 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
366 * doc/sourcebuild.texi (scalar_all_fma): Document.
367 * tree.def (FMA_EXPR): Delete.
368 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
369 * internal-fn.c (ternary_direct): New macro.
370 (expand_ternary_optab_fn): Likewise.
371 (direct_ternary_optab_supported_p): Likewise.
372 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
373 * builtins.c (fold_builtin_fma): Delete.
374 (fold_builtin_3): Don't call it.
375 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
376 * expr.c (expand_expr_real_2): Likewise.
377 * fold-const.c (operand_equal_p): Likewise.
378 (fold_ternary_loc): Likewise.
379 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
380 * gimple.c (DEFTREECODE): Likewise.
381 * gimplify.c (gimplify_expr): Likewise.
382 * optabs-tree.c (optab_for_tree_code): Likewise.
383 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
384 * tree-eh.c (operation_could_trap_p): Likewise.
385 (stmt_could_throw_1_p): Likewise.
386 * tree-inline.c (estimate_operator_cost): Likewise.
387 * tree-pretty-print.c (dump_generic_node): Likewise.
388 (op_code_prio): Likewise.
389 * tree-ssa-loop-im.c (stmt_cost): Likewise.
390 * tree-ssa-operands.c (get_expr_operands): Likewise.
391 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
392 * fold-const-call.h (fold_fma): Delete.
393 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
394 CFN_FNMA and CFN_FNMS.
396 * genmatch.c (combined_fn): New enum.
397 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
398 (commutative_op): New function.
399 (commutate): Use it. Handle more than 2 operands.
400 (dt_operand::gen_gimple_expr): Use commutative_op.
401 (parser::parse_expr): Allow :c to be used with non-binary
402 operators if the commutative operand is known.
403 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
404 CFN_FMS, CFN_FNMA and CFN_FNMS.
405 (backprop::process_assign_use): Remove FMA_EXPR handling.
406 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
407 (gen_hsa_fma): New function.
408 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
409 IFN_FNMA and IFN_FNMS.
410 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
411 * gimple-fold.h (follow_all_ssa_edges): Declare.
412 * gimple-fold.c (follow_all_ssa_edges): New function.
413 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
414 gimple_build interface and use follow_all_ssa_edges to fold the result.
415 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
416 instead of checking for optabs directly.
417 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
418 rather than FMA_EXPRs.
419 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
420 call to IFN_FMA instead of an FMA_EXPR.
422 2018-05-17 Jim Wilson <jimw@sifive.com>
424 * expr.c (do_tablejump): When converting index to Pmode, if we have a
425 sign extended promoted subreg, and the range does not have the sign bit
426 set, then do a sign extend.
428 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
429 test, check for sign extended subreg and/or constant operands, and
430 do a sign extend in that case.
432 2018-05-17 Steve Ellcey <sellcey@cavium.com>
434 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
435 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
437 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
438 Change logics_shift_reg to logics_shift_imm.
439 (thunderx2t99_fp_loadpair_basic): Delete.
440 (thunderx2t99_fp_storepair_basic): Delete.
441 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
442 (thunderx2t99_asimd_polynomial): Delete.
443 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
444 and neon_fp_mul_d_scalar_q.
445 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
446 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
447 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
448 (thunderx2t99_asimd_lut): Add missing tbl types.
449 (thunderx2t99_asimd_ext): Delete.
450 (thunderx2t99_asimd_load1_1_mult): Delete.
451 (thunderx2t99_asimd_load1_2_mult): Delete.
452 (thunderx2t99_asimd_load1_ldp): New.
453 (thunderx2t99_asimd_load1): New.
454 (thunderx2t99_asimd_load2): Add missing *load2* types.
455 (thunderx2t99_asimd_load3): New.
456 (thunderx2t99_asimd_load4): New.
457 (thunderx2t99_asimd_store1_1_mult): Delete.
458 (thunderx2t99_asimd_store1_2_mult): Delete.
459 (thunderx2t99_asimd_store2_mult): Delete.
460 (thunderx2t99_asimd_store2_onelane): Delete.
461 (thunderx2t99_asimd_store_stp): New.
462 (thunderx2t99_asimd_store1): New.
463 (thunderx2t99_asimd_store2): New.
464 (thunderx2t99_asimd_store3): New.
465 (thunderx2t99_asimd_store4): New.
467 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
469 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
470 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
472 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
473 Segher Boessenkool <segher@kernel.crashing.org>
476 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
479 2018-05-17 Richard Biener <rguenther@suse.de>
481 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
482 for pruning loop and prune defs feeding only already visited PHIs.
484 2018-05-17 Richard Biener <rguenther@suse.de>
486 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
488 2018-05-17 Bin Cheng <bin.cheng@arm.com>
489 Richard Biener <rguenther@suse.de>
491 PR tree-optimization/85793
492 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
493 for VMAT_ELEMENTWISE.
495 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
497 * internal-fn.h (lookup_internal_fn): Declare
498 * internal-fn.c (lookup_internal_fn): New function.
499 * gimple.c (gimple_build_call_from_tree): Handle calls to
501 * gimple-pretty-print.c (dump_gimple_call): Print "." before
502 internal function names.
503 * tree-pretty-print.c (dump_generic_node): Likewise.
504 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
506 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
508 * gimple-fold.h (gimple_build): Make the function forms take
509 combined_fn rather than built_in_function.
510 (gimple_simplify): Likewise.
511 * gimple-match-head.c (gimple_simplify): Likewise.
512 * gimple-fold.c (gimple_build): Likewise.
513 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
514 rather than gimple_build_call_internal.
515 (get_initial_defs_for_reduction): Likewise.
516 (vect_create_epilog_for_reduction): Likewise.
517 (vectorizable_live_operation): Likewise.
519 2018-05-17 Martin Liska <mliska@suse.cz>
521 * gimple-ssa-sprintf.c (format_directive): Do not use
522 space in between 'G_' and '('.
524 2018-05-17 Jakub Jelinek <jakub@redhat.com>
527 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
528 even if the mask is not all ones.
531 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
533 (ix86_gimple_fold_builtin): Likewise.
536 * config/i386/i386.c: Include tree-vector-builder.h.
537 (ix86_vector_shift_count): New function.
538 (ix86_fold_builtin): Fold shift builtins by scalar count.
539 (ix86_gimple_fold_builtin): Likewise.
541 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
542 _mm512_setzero): New intrinsics.
544 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
545 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
547 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
548 code generation for cases where splatting a value is not useful.
549 * simplify-rtx.c (simplify_ternary_operation): Simplify
550 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
551 mode to a vec_concat.
553 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
555 * config.gcc: Support "goldmont-plus".
556 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
558 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
559 PROCESSOR_GOLDMONT_PLUS.
560 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
561 (processor_target_table): Add "goldmont-plus".
562 (PTA_GOLDMONT_PLUS): Define.
563 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
564 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
565 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
566 (fold_builtin_cpu): Add "goldmont-plus".
567 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
568 (ix86_option_override_internal): Add "goldmont-plus".
569 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
570 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
571 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
572 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
574 2018-05-17 Richard Biener <rguenther@suse.de>
576 PR tree-optimization/85757
577 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
578 remove defs that only feed that PHI from further processing.
580 2018-05-16 Jim Wilson <jimw@sifive.com>
582 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
584 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
586 2018-05-16 Mark Wielaard <mark@klomp.org>
588 * dwarf2out.c (count_index_strings): New function.
589 (output_indirect_strings): Call count_index_strings and generate
590 header for dwarf_version >= 5.
592 2018-05-16 Mark Wielaard <mark@klomp.org>
594 * dwarf2out.c (dwarf_FORM): New function.
595 (set_indirect_string): Use dwarf_FORM.
596 (reset_indirect_string): Likewise.
597 (size_of_die): Likewise.
598 (value_format): Likewise.
599 (output_die): Likewise.
600 (add_skeleton_AT_string): Likewise.
601 (output_macinfo_op): Likewise.
602 (index_string): Likewise.
603 (output_index_string_offset): Likewise.
604 (output_index_string): Likewise.
605 (count_index_strings): Likewise.
607 2018-05-16 Carl Love <cel@us.ibm.com>
609 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
610 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
612 2018-05-16 Martin Jambor <mjambor@suse.cz>
614 * ipa-prop.c (ipa_free_all_edge_args): Remove.
615 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
617 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
619 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
620 (fnma<mode>4): Likewise.
621 (fms<mode>4): Likewise.
622 (fnms<mode>4): Likewise.
623 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
624 (aarch64_fnma<mode>4): Likewise.
625 (aarch64_fms<mode>4): Likewise.
626 (aarch64_fnms<mode>4): Likewise.
627 (aarch64_fnmadd<mode>4): Likewise.
629 2018-05-16 Jason Merrill <jason@redhat.com>
631 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
633 2018-05-16 Richard Biener <rguenther@suse.de>
635 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
636 (dump_stmt_cost): Declare.
637 (add_stmt_cost): Dump cost we add.
638 (add_stmt_costs): New function.
639 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
641 (vect_analyze_stmt): Adjust prototype.
642 (vectorizable_condition): Likewise.
643 (vectorizable_live_operation): Likewise.
644 (vectorizable_reduction): Likewise.
645 (vectorizable_induction): Likewise.
646 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
647 cost vector to pass to vectorizable_ and record afterwards.
648 (vect_model_reduction_cost): Take cost vector argument and adjust.
649 (vect_model_induction_cost): Likewise.
650 (vectorizable_reduction): Likewise.
651 (vectorizable_induction): Likewise.
652 (vectorizable_live_operation): Likewise.
653 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
654 SLP_TREE_NUMBER_OF_VEC_STMTS.
655 (vect_analyze_slp_cost_1): Remove.
656 (vect_analyze_slp_cost): Likewise.
657 (vect_slp_analyze_node_operations): Take visited args and
658 a target cost vector. Avoid processing already visited stmt sets.
659 (vect_slp_analyze_operations): Use a local cost vector to gather
660 costs and register those of non-discarded instances.
661 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
662 (vect_schedule_slp_instance): Remove copying of
663 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
665 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
666 adding cost. Record cost entry location.
667 (vect_prologue_cost_for_slp_op): Function to compute cost of
668 a constant or invariant generated for SLP vect in the prologue,
669 split out from vect_analyze_slp_cost_1.
670 (vect_model_simple_cost): Make static. Adjust for SLP costing.
671 (vect_model_promotion_demotion_cost): Likewise.
672 (vect_model_store_cost): Likewise, make static.
673 (vect_model_load_cost): Likewise.
674 (vectorizable_bswap): Add cost vector arg and adjust.
675 (vectorizable_call): Likewise.
676 (vectorizable_simd_clone_call): Likewise.
677 (vectorizable_conversion): Likewise.
678 (vectorizable_assignment): Likewise.
679 (vectorizable_shift): Likewise.
680 (vectorizable_operation): Likewise.
681 (vectorizable_store): Likewise.
682 (vectorizable_load): Likewise.
683 (vectorizable_condition): Likewise.
684 (vectorizable_comparison): Likewise.
685 (can_vectorize_live_stmts): Likewise.
686 (vect_analyze_stmt): Likewise.
687 (vect_transform_stmt): Adjust calls to vectorizable_*.
688 * tree-vectorizer.c: Include gimple-pretty-print.h.
689 (dump_stmt_cost): New function.
691 2018-05-16 Richard Biener <rguenther@suse.de>
693 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
694 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
695 * tree-ssa-dse.c: Include tree-ssa-loop.h.
696 (check_name): New callback.
697 (dse_classify_store): Track cycles via a visited bitmap of PHI
698 defs and simplify handling of in-loop and across loop dead stores
699 and properly fail for loop-variant refs. Handle byte-tracking with
700 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
703 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
705 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
706 (vect_get_mask_type_for_stmt): Likewise.
707 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
709 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
710 to determine the statement's vector type and the vector type that
711 should be used for calculating nunits. Deal with cases in which
712 the type has to be deferred.
713 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
714 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
715 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
716 (vect_determine_vf_for_stmt): New functions, split out from...
717 (vect_determine_vectorization_factor): ...here.
718 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
719 (vect_get_mask_type_for_stmt): New functions, split out from
720 vect_determine_vectorization_factor.
722 2018-05-16 Richard Biener <rguenther@suse.de>
724 * tree-cfg.c (verify_gimple_assign_ternary): Properly
725 verify the [VEC_]COND_EXPR embedded comparison.
727 2018-05-15 Martin Sebor <msebor@redhat.com>
729 PR tree-optimization/85753
730 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
731 RECORD_TYPE in addition to ARRAY_TYPE.
733 2018-05-15 Martin Sebor <msebor@redhat.com>
736 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
738 2018-05-15 Richard Biener <rguenther@suse.de>
740 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
741 add by_clobber_p one. Change algorithm to collect all defs
742 representing uses we need to walk and try reducing them to
743 a single one before failing.
744 (dse_dom_walker::dse_optimize_stmt): Adjust.
746 2018-05-13 Mark Wielaard <mark@klomp.org>
748 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
749 (size_of_loc_descr): Likewise.
750 (output_loc_operands): Likewise.
751 (output_loc_operands_raw): Likewise.
752 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
753 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
754 (hash_loc_operands): Likewise.
755 (compare_loc_operands): Likewise.
757 2018-05-14 Mark Wielaard <mark@klomp.org>
759 * dwarf2out.c (count_index_addrs): New function.
760 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
762 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
764 PR tree-optimization/83648
765 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
766 return value as malloc candidate.
768 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
771 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
772 as true in call to suggest_attribute.
774 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
776 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
779 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
781 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
783 (aarch64_crypto_aesd_fused): Likewise.
785 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
787 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
788 (movsi_aarch64): Likewise.
789 (load_pairsi): Likewise.
790 (load_pairdi): Likewise.
791 (store_pairsi): Likewise.
792 (store_pairdi): Likewise.
793 (load_pairsf): Likewise.
794 (load_pairdf): Likewise.
795 (store_pairsf): Likewise.
796 (store_pairdf): Likewise.
797 (zero_extend): Likewise.
798 (trunc): Swap alternatives.
799 (fcvt_target): Add '?' to prefer w over r.
801 2018-05-14 Jakub Jelinek <jakub@redhat.com>
804 * config/i386/i386.md: Disallow non-commutative arithmetics in
805 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
806 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
807 in the peephole2 before it.
809 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
811 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
812 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
813 (ix86_handle_option): Handle -mcldemote.
814 * config.gcc: New header.
815 * config/i386/cldemoteintrin.h: New file.
816 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
817 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
819 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
820 OPTION_MASK_ISA_CLDEMOTE.
821 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
822 (ix86_valid_target_attribute_inner_p): Ditto.
823 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
824 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
825 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
826 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
827 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
829 * config/i386/i386.opt: Add -mcldemote.
830 * config/i386/x86intrin.h: New header.
831 * doc/invoke.texi: Add -mcldemote.
833 2018-05-14 Richard Biener <rguenther@suse.de>
835 * doc/match-and-simplify.texi: Adjust :s documentation.
837 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
839 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
840 intended memcpy size.
841 (REORDER_45): Likewise.
843 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
846 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
847 * vec.c (qsort_chk): Use gcc_qsort.
848 * Makefile.in (OBJS-libcommon): Add sort.o.
849 (build/sort.o): New target. Use it...
850 (BUILD_RTL): ... here, and...
851 (build/gencfn-macros): ... here, and...
852 (build/genmatch): ... here.
854 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
855 Chung-Ju Wu <jasonwucj@gmail.com>
857 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
858 * config/nds32/nds32-graywolf.md: New file.
859 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
860 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
862 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
863 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
864 * config/nds32/nds32.md (pipeline_model): Add graywolf.
865 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
866 * config/nds32/pipelines.md: Include n15 settings.
868 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
869 Chung-Ju Wu <jasonwucj@gmail.com>
871 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
872 * config/nds32/nds32-n13.md: New file.
873 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
874 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
876 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
877 * config/nds32/nds32.md (pipeline_model): Add n13.
878 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
879 * config/nds32/pipelines.md: Include n13 settings.
881 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
882 Chung-Ju Wu <jasonwucj@gmail.com>
884 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
885 * config/nds32/nds32-n10.md: New file.
886 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
887 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
889 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
890 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
891 * config/nds32/nds32.md (pipeline_model): Add n10.
892 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
893 * config/nds32/pipelines.md: Include n10 settings.
895 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
896 Kito Cheng <kito.cheng@gmail.com>
897 Chung-Ju Wu <jasonwucj@gmail.com>
899 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
900 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
901 Add enum values for DSP extension instructions.
902 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
904 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
905 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
907 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
908 * config/nds32/nds32-dspext.md: New file for DSP implementation.
909 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
910 * config/nds32/nds32-intrinsic.md: Likewise.
911 * config/nds32/nds32_intrinsic.h: Likewise.
912 * config/nds32/nds32-md-auxiliary.c: Likewise.
913 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
914 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
915 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
916 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
917 * config/nds32/nds32-protos.h: New declarations for DSP extension.
918 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
919 TYPE_DMAC in switch statement.
920 * config/nds32/nds32.c: New checking and implementation for DSP
921 extension instructions.
922 * config/nds32/nds32.h: Likewise.
923 * config/nds32/nds32.md: Likewise.
924 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
925 * config/nds32/predicates.md: Implement new predicates for DSP
928 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
930 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
931 Reformat alternatives and attributes so it is easier to identify
932 which constraints/attributes go with which instruction.
933 (mov<mode>_hardfloat32, FMOVE64): Likewise.
934 (mov<mode>_softfloat32, FMOVE64): Likewise.
935 (mov<mode>_hardfloat64, FMOVE64): Likewise.
936 (mov<mode>_softfloat64, FMOVE64): Likewise.
938 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
940 * doc/extend.texi (PowerPC Built-in Functions): Rename this
942 (Basic PowerPC Built-in Functions): The new name of the
943 subsection previously known as "PowerPC Built-in Functions".
944 (Basic PowerPC Built-in Functions Available on all Configurations):
946 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
947 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
948 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
949 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
951 2018-05-11 Martin Jambor <mjambor@suse.cz>
954 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
957 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
960 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
962 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
964 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
965 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
966 (ix86_handle_option): Handle -mwaitpkg.
967 * config.gcc: New header.
968 * config/i386/cpuid.h (bit_WAITPKG): New bit.
969 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
970 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
972 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
973 OPTION_MASK_ISA_WAITPKG.
974 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
975 (ix86_option_override_internal): Add PTA_WAITPKG.
976 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
977 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
979 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
980 __builtin_ia32_umwait and __builtin_ia32_tpause.
981 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
982 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
983 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
984 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
985 UNSPECV_TPAUSE): New.
986 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
987 * config/i386/i386.opt: Add -mwaitpkg.
988 * config/i386/waitpkgintrin.h: New file.
989 * config/i386/x86intrin.h: New header.
990 * doc/invoke.texi: Add -mwaitpkg.
992 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
995 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
997 (cortex-m0): Use armv6s-m isa.
998 (cortex-m0plus): Likewise.
999 (cortex-m1): Likewise.
1000 (cortex-m0.small-multiply): Likewise.
1001 (cortex-m0plus.small-multiply): Likewise.
1002 (cortex-m1.small-multiply): Likewise.
1004 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1005 Jakub Jelinek <jakub@redhat.com>
1007 PR tree-optimization/85692
1008 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1009 source permute as well.
1011 2018-05-11 Martin Liska <mliska@suse.cz>
1014 * doc/extend.texi: Document LLVM style format for no_sanitize
1017 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1019 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1020 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1021 (mode_supports_vsx_dform_quad): Likewise.
1022 (mode_supports_vmx_dform): Move these functions to be next to the
1023 other mode_supports functions.
1024 (mode_supports_dq_form): Likewise.
1025 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1026 mode_supports_dq_form.
1027 (reg_offset_addressing_ok_p): Likewise.
1028 (offsettable_ok_by_alignment): Likewise.
1029 (rs6000_legitimate_offset_address_p): Likewise.
1030 (legitimate_lo_sum_address_p): Likewise.
1031 (rs6000_legitimize_address): Likewise.
1032 (rs6000_legitimize_reload_address): Likewise.
1033 (rs6000_secondary_reload_inner): Likewise.
1034 (rs6000_preferred_reload_class): Likewise.
1035 (rs6000_output_move_128bit): Likewise.
1037 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1039 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1040 Generate SImode target register for null target.
1041 <case IX86_BUILTIN_XGETBV>: Ditto.
1042 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1043 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1045 2018-05-10 Carl Love <cel@us.ibm.com>
1047 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1048 dcbtt and dcbtstt if operands[2] is 0.
1050 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1053 * config/i386/sse.md (usadv64qi): New expander.
1055 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1057 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1058 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1059 -maltivec=be support.
1060 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1061 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1062 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1063 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1064 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1065 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1066 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1067 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1068 altivec_vsumsws): Adjust.
1069 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1070 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1071 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1073 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1074 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1075 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1076 (altivec_lve<VI_char>x): Delete expand.
1077 (*altivec_lve<VI_char>x_internal): Rename to...
1078 (altivec_lve<VI_char>x): ... this.
1079 (altivec_lvxl_<mode>): Delete expand.
1080 (*altivec_lvxl_<mode>_internal): Rename to ...
1081 (altivec_lvxl_<mode>): ... this.
1082 (altivec_stvxl_<mode>): Delete expand.
1083 (*altivec_stvxl_<mode>_internal): Rename to ...
1084 (altivec_stvxl_<mode>): ... this.
1085 (altivec_stve<VI_char>x): Delete expand.
1086 (*altivec_stve<VI_char>x_internal): Rename to ...
1087 (altivec_stve<VI_char>x): ... this.
1088 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1089 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1090 reduc_plus_scal_<mode>): Adjust.
1091 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1093 (rs6000_cpu_cpp_builtins): Adjust.
1094 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1095 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1096 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1097 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1098 -maltivec=be support.
1099 (rs6000_split_vec_extract_var): Adjust.
1100 (rs6000_split_v4si_init): Adjust.
1101 (swap_selector_for_mode): Delete.
1102 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1103 altivec_expand_stvex_be): Delete.
1104 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1105 -maltivec=be support.
1106 (rs6000_gimple_fold_builtin): Ditto.
1107 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1109 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1110 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1111 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1112 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1113 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1114 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1115 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1116 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1117 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1118 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1119 anonymous split): Adjust.
1120 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1121 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1123 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1125 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1126 when --with-gxx-include-dir is also specified.
1127 * configure: Regenerate.
1129 2018-05-09 Jim Wilson <jimw@sifive.com>
1132 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1133 * config/riscv/t-withmultilib: New.
1134 * config/riscv/withmultilib.h: New.
1135 * doc/install.texi: Document RISC-V --with-multilib-list support.
1137 2018-05-09 Richard Biener <rguenther@suse.de>
1139 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1141 (vect_bb_vectorization_profitable_p): Adjust. Compute
1142 actual scalar cost using the cost vector and the add_stmt_cost
1145 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1147 PR rtl-optimization/85645
1148 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1149 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1151 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1153 PR rtl-optimization/85645
1154 * shrink-wrap.c (spread_components): Return a boolean saying if
1155 anything was changed.
1156 (try_shrink_wrapping_separate): Iterate spread_components until
1157 nothing changes anymore.
1159 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1161 PR rtl-optimization/85645
1162 * regrename.c (build_def_use): Also kill the chains that include the
1163 destination of a REG_CFA_REGISTER note.
1165 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1167 PR rtl-optimization/85645
1168 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1169 insn that has a REG_CFA_REGISTER note.
1171 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1173 * cfgexpand.c (expand_clobber): New function.
1174 (expand_gimple_stmt_1): Use it.
1175 * tree-vect-stmts.c (vect_clobber_variable): New function,
1177 (vectorizable_simd_clone_call): ...here.
1178 (vectorizable_store): Emit a clobber either side of an
1179 IFN_STORE_LANES sequence.
1180 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1182 2018-05-09 Tom de Vries <tom@codesourcery.com>
1185 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1186 (define_insn "trap_if_false"): Add exit after trap.
1188 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1190 PR rtl-optimization/85638
1191 * bb-reorder.c: Include common/common-target.h.
1192 (create_forwarder_block): New function extracted from...
1193 (fix_up_crossing_landing_pad): ...here. Rename into...
1194 (dw2_fix_up_crossing_landing_pad): ...this.
1195 (sjlj_fix_up_crossing_landing_pad): New function.
1196 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1197 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1198 from both partitions and exit the loop after one iteration.
1200 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1203 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1205 (Basic PowerPC Built-in Functions): The new name of the
1206 subsection previously known as "PowerPC Built-in Functions".
1207 (Basic PowerPC Built-in Functions Available on all Configurations):
1209 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1211 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1213 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1215 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1218 2018-05-08 Jim Wilson <jimw@sifive.com>
1220 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1221 (LD_EMUL_SUFFIX): New.
1222 (LINK_SPEC): Use it.
1224 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1226 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1228 (Basic PowerPC Built-in Functions): The new name of the
1229 subsection previously known as "PowerPC Built-in Functions".
1230 (Basic PowerPC Built-in Functions Available on all Configurations):
1232 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1234 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1236 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1238 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1241 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1244 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1245 after cmpelim optimization.
1247 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1249 * config.gcc: Support "goldmont".
1250 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1251 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1253 * config/i386/i386.c (m_GOLDMONT): Define.
1254 (processor_target_table): Add "goldmont".
1255 (PTA_GOLDMONT): Define.
1256 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1257 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1258 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1259 (fold_builtin_cpu): Add "goldmont".
1260 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1261 (ix86_option_override_internal): Add "goldmont".
1262 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1263 (processor_type): Add PROCESSOR_GOLDMONT.
1264 * config/i386/i386.md: Add CPU "glm".
1265 * config/i386/glm.md: New file.
1266 * config/i386/x86-tune.def: Add m_GOLDMONT.
1267 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1269 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1272 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1274 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1275 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1276 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1279 * config/i386/i386.c (ix86_fold_builtin): Handle
1280 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1283 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1284 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1286 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1289 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1290 (check_arch): Likewise.
1291 (check_fpu): Return the result rather than printing it.
1292 (end arch): Fix operator precedence.
1293 (end cpu): Likewise.
1294 (END): Print the result from check_fpu.
1296 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1297 Alan Hayward <alan.hayward@arm.com>
1298 David Sherwood <david.sherwood@arm.com>
1300 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1301 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1302 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1303 (*fcmuo<mode>_and): New patterns.
1305 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1307 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1308 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1309 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1310 (cmp_op, sve_imm_con): New code attributes.
1311 (SVE_COND_INT_CMP, imm_con): Delete.
1312 (cmp_op): Remove above unspecs from int attribute.
1313 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1315 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1316 comparison-specific unspecs.
1317 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1318 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1319 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1320 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1321 (*vec_fcm<cmp_op><mode>): Rename to...
1322 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1323 (*vec_fcmuo<mode>): Rename to...
1324 (*fcmuo<mode>): ...this and adjust likewise.
1325 (*pred_fcm<cmp_op><mode>): New pattern.
1326 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1327 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1329 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1331 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1332 (aarch64_emit_sve_predicated_cond): New function.
1333 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1334 (aarch64_emit_unspec_cond_or): Replace with...
1335 (aarch64_emit_sve_or_conds): ...this new function. Use
1336 aarch64_emit_sve_ptrue_op for the individual comparisons and
1337 aarch64_emit_binop to OR them together.
1338 (aarch64_emit_inverted_unspec_cond): Replace with...
1339 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1340 aarch64_emit_sve_ptrue_op for the comparison and
1341 aarch64_emit_unop to invert the result.
1342 (aarch64_expand_sve_vec_cmp_float): Update after the above
1343 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1345 2018-05-07 Nathan Sidwell <nathan@acm.org>
1347 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1348 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1349 (Backwards Compatibility): Likewise.
1351 2018-05-07 Luis Machado <luis.machado@linaro.org>
1355 2018-05-07 Luis Machado <luis.machado@linaro.org>
1357 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1358 <prefetch_dynamic_strides>: New const bool field.
1359 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1360 prefetch_dynamic_strides.
1361 (exynosm1_prefetch_tune): Likewise.
1362 (thunderxt88_prefetch_tune): Likewise.
1363 (thunderx_prefetch_tune): Likewise.
1364 (thunderx2t99_prefetch_tune): Likewise.
1365 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1366 (aarch64_override_options_internal): Update to set
1367 PARAM_PREFETCH_DYNAMIC_STRIDES.
1368 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1369 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1370 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1371 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1372 prefetch-dynamic-strides setting.
1374 2018-05-07 Luis Machado <luis.machado@linaro.org>
1376 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1377 <minimum_stride>: New const int field.
1378 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1379 minimum_stride field.
1380 (exynosm1_prefetch_tune): Likewise.
1381 (thunderxt88_prefetch_tune): Likewise.
1382 (thunderx_prefetch_tune): Likewise.
1383 (thunderx2t99_prefetch_tune): Likewise.
1384 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1385 (aarch64_override_options_internal): Update to set
1386 PARAM_PREFETCH_MINIMUM_STRIDE.
1387 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1388 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1389 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1390 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1391 stride is constant and is below the minimum stride threshold.
1393 2018-05-07 Luis Machado <luis.machado@linaro.org>
1395 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1398 2018-05-07 Luis Machado <luis.machado@linaro.org>
1400 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1401 <prefetch_dynamic_strides>: New const bool field.
1402 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1403 prefetch_dynamic_strides.
1404 (exynosm1_prefetch_tune): Likewise.
1405 (thunderxt88_prefetch_tune): Likewise.
1406 (thunderx_prefetch_tune): Likewise.
1407 (thunderx2t99_prefetch_tune): Likewise.
1408 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1409 (aarch64_override_options_internal): Update to set
1410 PARAM_PREFETCH_DYNAMIC_STRIDES.
1411 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1412 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1413 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1414 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1415 prefetch-dynamic-strides setting.
1417 2018-05-07 Luis Machado <luis.machado@linaro.org>
1419 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1420 <minimum_stride>: New const int field.
1421 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1422 minimum_stride field.
1423 (exynosm1_prefetch_tune): Likewise.
1424 (thunderxt88_prefetch_tune): Likewise.
1425 (thunderx_prefetch_tune): Likewise.
1426 (thunderx2t99_prefetch_tune): Likewise.
1427 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1428 (aarch64_override_options_internal): Update to set
1429 PARAM_PREFETCH_MINIMUM_STRIDE.
1430 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1431 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1432 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1433 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1434 stride is constant and is below the minimum stride threshold.
1436 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1439 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1440 the type is addressable. Don't force op into register if it has
1443 2018-05-05 Roland McGrath <mcgrathr@google.com>
1446 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1447 any section for which we don't know a specific type it should have,
1448 regardless of name. Previously this was done only for the exact
1449 names ".init_array", ".fini_array", and ".preinit_array".
1450 (default_elf_asm_named_section): Add comment about
1451 relationship with default_section_type_flags and SECTION_NOTYPE.
1452 (get_section): Don't consider it a type conflict if one side has
1453 SECTION_NOTYPE and the other doesn't, as long as neither has the
1454 SECTION_BSS et al used in the default_section_type_flags logic.
1456 2018-05-05 Tom de Vries <tom@codesourcery.com>
1459 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1460 (workaround_barsyncs): New function.
1461 (nvptx_reorg): Use workaround_barsyncs.
1462 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1463 (define_expand "nvptx_membar_cta"): New define_expand.
1464 (define_insn "*nvptx_membar_cta"): New insn.
1466 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1468 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1469 To improve optimization opportunities.
1470 * builtin-types.def: The new needed builtin types for the above.
1472 2018-05-04 Richard Biener <rguenther@suse.de>
1474 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1475 * gimple-ssa-store-merging.c
1476 (imm_store_chain_info::output_merged_store): Remove redundant create,
1477 release split_store vector contents on failure.
1478 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1479 scalar stmt vector on cache hit.
1481 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1483 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1485 * config.gcc (powerpc-xilinx-eabi*): Remove.
1486 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1488 (fusion_addis_mem_combo_load): Ditto.
1489 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1491 (rs6000_cpu_cpp_builtins): Ditto.
1492 * config/rs6000/rs6000-linux.c
1493 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1494 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1495 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1497 (rs6000_setup_reg_addr_masks): Ditto.
1498 (rs6000_init_hard_regno_mode_ok): Ditto.
1499 (rs6000_option_override_internal): Ditto.
1500 (legitimate_lo_sum_address_p): Ditto.
1501 (rs6000_legitimize_address): Ditto.
1502 (rs6000_legitimize_reload_address): Ditto.
1503 (rs6000_legitimate_address_p): Ditto.
1504 (abi_v4_pass_in_fpr): Ditto.
1505 (setup_incoming_varargs): Ditto.
1506 (rs6000_gimplify_va_arg): Ditto.
1507 (rs6000_split_multireg_move): Ditto.
1508 (rs6000_savres_strategy): Ditto.
1509 (rs6000_emit_prologue_components): Ditto.
1510 (rs6000_emit_epilogue_components): Ditto.
1511 (rs6000_emit_prologue): Ditto.
1512 (rs6000_emit_epilogue): Ditto.
1513 (rs6000_elf_file_end): Ditto.
1514 (rs6000_function_value): Ditto.
1515 (rs6000_libcall_value): Ditto.
1516 * config/rs6000/rs6000.h: Ditto.
1517 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1518 (TARGET_MINMAX): ... this. New.
1519 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1520 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1521 (*movsi_internal1_single): Delete.
1522 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1523 mfpu=, mxilinx-fpu): Delete.
1524 * config/rs6000/singlefp.h: Delete.
1525 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1526 * config/rs6000/t-rs6000: Ditto.
1527 * config/rs6000/t-xilinx: Delete.
1528 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1529 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1530 (VStype_simple): Delete.
1531 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1532 * config/rs6000/xfpu.h: Delete.
1533 * config/rs6000/xfpu.md: Delete.
1534 * config/rs6000/xilinx.h: Delete.
1535 * config/rs6000/xilinx.opt: Delete.
1536 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1537 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1539 2018-05-04 Tom de Vries <tom@codesourcery.com>
1542 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1545 2018-05-04 Richard Biener <rguenther@suse.de>
1548 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1549 (expand_complex_div_wide): Likewise.
1550 (expand_complex_operations_1): Likewise.
1551 (expand_complex_libcall): Preserve EH info of the original stmt.
1552 (tree_lower_complex): Handle removed blocks.
1553 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1554 on complex multiplication and division libcall builtins.
1556 2018-05-04 Richard Biener <rguenther@suse.de>
1559 * fold-const.c (negate_expr_p): Restrict negation of operand
1560 zero of a division to when we know that can happen without
1562 (fold_negate_expr_1): Likewise.
1564 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1567 * real.h (real_nextafter): Declare.
1568 * real.c (real_nextafter): New function.
1569 * fold-const-call.c (fold_const_nextafter): New function.
1570 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1571 CASE_CFN_NEXTTOWARD.
1572 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1573 even when arg1_mode is different from arg0_mode.
1575 2018-05-03 Nathan Sidwell <nathan@acm.org>
1577 * doc/extend.texi (Deprecated Features): Remove
1579 (Backwards Compatibility): Likewise.
1580 * doc/invoke.texi (C++ Language Options): Likewise.
1581 (C++ Dialect Options): Likewise.
1583 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1586 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1587 _mm512_mask_mullox_epi64): New intrinsics.
1589 2018-05-03 Tom de Vries <tom@codesourcery.com>
1592 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1593 dump files): Add offload-tree.
1595 2018-05-03 Richard Biener <rguenther@suse.de>
1597 PR tree-optimization/85615
1598 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1599 to loops not nested in BBs loop father to avoid creating multi-entry
1602 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1604 PR tree-optimization/70291
1605 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1606 arguments. Change return type to tree. Emit libcall as a new
1607 statement rather than replacing existing one when inplace_p is true.
1608 (expand_complex_multiplication_components): New function.
1609 (expand_complex_multiplication): Expand floating-point complex
1610 multiplication using the above.
1611 (expand_complex_division): Rename inner_type parameter to type.
1612 Update expand_complex_libcall call-site.
1613 (expand_complex_operations_1): Update expand_complex_multiplication
1614 and expand_complex_division call-sites.
1616 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1619 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1620 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1621 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1622 the highest significant bit of the shift count mask is clear. In
1623 check whether and[sq]i3 is needed verify that all significant bits
1624 of the shift count other than the highest are set.
1626 2018-05-02 Tom de Vries <tom@codesourcery.com>
1629 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1630 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1631 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1632 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1633 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1634 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1635 __builtin_goacc_parlevel_size.
1637 2018-05-02 Richard Biener <rguenther@suse.de>
1639 PR tree-optimization/85597
1640 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1641 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1644 2018-05-02 Tom de Vries <tom@codesourcery.com>
1647 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1648 dump files): Add ltrans-tree.
1650 2018-05-02 Tom de Vries <tom@codesourcery.com>
1653 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1654 dump files): Add wpa-ipa.
1656 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1658 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1659 powerpc*-*-linux*paired* target.
1660 * config/rs6000/750cl.h: Delete.
1661 * config/rs6000/paired.h: Delete.
1662 * config/rs6000/paired.md: Delete.
1663 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1665 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1666 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1667 comment. Remove paired float support.
1668 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1669 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1671 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1672 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1674 * config/rs6000/rs6000.c: Remove paired float support.
1675 (paired_expand_vector_init, paired_expand_vector_move,
1676 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1677 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1678 paired_expand_builtin, paired_expand_predicate_builtin,
1679 paired_init_builtins): Delete.
1680 * config/rs6000/rs6000.h: Remove paired float support.
1681 * config/rs6000/rs6000.md: Remove paired float support.
1682 (move_from_CR_ov_bit): Delete.
1683 * config/rs6000/rs6000.opt (mpaired): Delete.
1684 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1685 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1687 2018-05-02 Richard Biener <rguenther@suse.de>
1690 * gimplify.c (gimplify_save_expr): When in SSA form allow
1691 SAVE_EXPRs to compute to SSA vars.
1693 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1696 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1697 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1698 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1699 clobber operands[2], instead use a new pseudo. Formatting fixes.
1701 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1703 PR tree-optimization/85586
1704 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1705 exit early for statements in the same group if the accesses are
1708 2018-05-02 Tom de Vries <tom@codesourcery.com>
1711 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1714 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1716 PR tree-optimization/85143
1717 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1719 2018-05-01 Tom de Vries <tom@codesourcery.com>
1722 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1723 not found" error message.
1725 2018-05-01 Tom de Vries <tom@codesourcery.com>
1728 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1729 * vec.c (test_ordered_remove_if): New function.
1730 (vec_c_tests): Call test_ordered_remove_if.
1731 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1732 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1733 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1734 VEC_ORDERED_REMOVE_IF.
1736 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1738 PR tree-optimization/82665
1739 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1740 pointer subtraction where arguments come from a memchr call.
1742 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1744 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1745 --push-state --as-needed and --pop-state instead of --as-needed and
1746 --no-as-needed if ld supports it.
1747 * configure: Regenerated.
1750 * doc/install.texi2html: Replace _002d with - and _002a with * in
1751 generated html files using sed.
1753 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1756 * gcc-rich-location.c (blank_line_before_p): New function.
1757 (use_new_line): New function.
1758 (gcc_rich_location::add_fixit_insert_formatted): New function.
1759 * gcc-rich-location.h
1760 (gcc_rich_location::add_fixit_insert_formatted): New function.
1762 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1764 * selftest.c (assert_streq): Rename "expected" and "actual" to
1765 "val1" and "val2". Extend NULL-handling to cover both inputs
1766 symmetrically, while still requiring both to be non-NULL for a pass.
1767 * selftest.h (assert_streq): Rename "expected" and "actual" to
1769 (ASSERT_EQ): Likewise.
1770 (ASSERT_EQ_AT): Likewise.
1771 (ASSERT_KNOWN_EQ): Likewise.
1772 (ASSERT_KNOWN_EQ_AT): Likewise.
1773 (ASSERT_NE): Likewise.
1774 (ASSERT_MAYBE_NE): Likewise.
1775 (ASSERT_MAYBE_NE_AT): Likewise.
1776 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1777 the assertion to pass.
1778 (ASSERT_STREQ_AT): Likewise.
1780 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1782 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1783 interaction with -pie.
1785 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1787 * selftest.h: Fix alphabetization of per-source-file selftest
1790 2018-04-30 Jason Merrill <jason@redhat.com>
1792 PR c++/61982 - dead stores to destroyed objects.
1793 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1796 2018-04-30 Jason Merrill <jason@redhat.com>
1798 * tree.c (build_clobber): New.
1799 * tree.h: Declare it.
1800 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1802 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1804 * diagnostic-show-locus.c (layout::layout): Update for
1805 location_get_source_line returning a char_span.
1806 (struct char_span): Move to input.h.
1807 (struct correction): Update for fields in char_span becoming
1809 (struct source_line): Update for location_get_source_line
1810 returning a char_span.
1811 (layout::print_line): Likewise.
1812 * edit-context.c (edited_file::print_content): Likewise.
1813 (edited_file::print_diff_hunk): Likewise.
1814 (edited_file::print_run_of_changed_lines): Likewise.
1815 (edited_file::get_num_lines): Likewise.
1816 (edited_line::edited_line): Likewise.
1817 * final.c (asm_show_source): Likewise.
1818 * input.c (location_get_source_line): Convert return type
1819 from const char * to char_span, losing the final "line_len"
1821 (dump_location_info): Update for the above.
1822 (get_substring_ranges_for_loc): Likewise. Use a char_span
1823 when handling the literal within the line.
1824 (test_reading_source_line): Update for location_get_source_line
1825 returning a char_span.
1826 * input.h (class char_span): Move here from
1827 diagnostic-show-locus.c, converting from a struct to a class.
1828 Make data members private.
1829 (char_span::operator bool): New.
1830 (char_span::length): New.
1831 (char_span::get_buffer): New.
1832 (char_span::operator[]): New.
1833 (char_span::subspan): Make const.
1834 (char_span::xstrdup): New.
1835 (location_get_source_line): Convert return type from const char *
1836 to char_span, losing the final "line_size" param.
1838 2018-04-30 Jan Hubicka <jh@suse.cz>
1840 * lto-wrapper.c (ltrans_priorities): New static var.
1841 (cmp_priority): New.
1842 (run_gcc): Read priorities and if doing parallel build order
1843 the Makefile by them.
1845 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1847 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1849 2018-04-30 Richard Biener <rguenther@suse.de>
1851 * tree-cfg.c (verify_address): Remove base argument, add
1852 flag whether to check TREE_ADDRESSABLE and do that.
1853 (verify_expr): Remove.
1854 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1855 (verify_gimple_assign_single): Likewise.
1856 (verify_gimple_switch): Likewise.
1857 (verify_expr_location_1): Dereference tp once. Add (disabled)
1858 piece from verify_expr.
1859 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1861 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1863 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1865 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1867 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1868 (small_data_pattern): Likewise.
1869 (arc_rewrite_small_data): Likewise.
1870 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1871 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1872 (get_symbol_alignment): New function.
1873 (legitimate_small_data_address_p): Likewise.
1874 (legitimate_scaled_address): Update, call
1875 legitimate_small_data_address_p.
1876 (output_sdata): New static variable.
1877 (arc_print_operand): Update how we handle small data operands.
1878 (arc_print_operand_address): Likewise.
1879 (arc_legitimate_address_p): Update, use
1880 legitimate_small_data_address_p.
1881 (arc_rewrite_small_data_p): Remove.
1882 (arc_rewrite_small_data_1): Likewise.
1883 (arc_rewrite_small_data): Likewise.
1884 (small_data_pattern): Likewise.
1885 (compact_sda_memory_operand): Update to use
1886 legitimate_small_data_address_p and get_symbol_alignment.
1887 (prepare_move_operands): Don't rewite sdata pattern.
1888 (prepare_extend_operands): Remove.
1889 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1891 (zero_extendqisi2): Likewise.
1892 (zero_extendhisi2): Likewise.
1893 (extendqihi2): Likewise.
1894 (extendqisi2): Likewise.
1895 (extendhisi2): Likewise.
1899 * config/arc/constraints.md (Usd): Change it to memory constraint.
1901 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1903 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1904 as source of std instructions.
1905 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1906 allow 6-bit constants as source for store instructions.
1907 (movdi_insn): Update instruction pattern to allow 6-bit constants
1908 as source for store instructions.
1910 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1912 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1914 2018-04-30 Nathan Sidwell <nathan@acm.org>
1915 Sandra Loosemore <sandra@codesourcery.com>
1917 * dumpfile.c (dump_open): Allow '-' for stdout.
1918 * doc/invoke.texi (Developer Options): Document dump filename
1919 determination early. Document stdin/stdout selection.
1921 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1923 Microblaze Target: PIC data text relative
1925 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1926 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1928 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1929 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1930 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1931 New addressing mode for data-text relative position indepenedent code.
1932 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1933 'ADDRESS_SYMBOLIC_TXT_REL'.
1934 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1935 (microblaze_legitimate_pic_operand): Exclude function calls from
1936 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1937 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1939 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1940 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1941 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1942 for 'address + offset'.
1943 (microblaze_expand_prologue): Add new function prologue call for
1945 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1946 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1947 table in case of TARGET_PIC_DATA_TEXT_REL.
1948 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1949 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1950 Add new macros 'UNSPEC_TEXT',
1951 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1952 + exclude function calls from 'UNSPEC_PLT' in case of data text
1954 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1955 new target hook for generating address diff vector tables in case of
1957 * doc/tm.texi : Regenerate.
1958 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1959 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1960 of addr diff vector generation.
1961 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1962 target hook definition.
1963 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1964 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1965 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1966 option for data text relative.
1968 2018-04-30 Richard Biener <rguenther@suse.de>
1970 * tree-chrec.h (evolution_function_is_constant_p): Remove
1972 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1974 2018-04-30 Richard Biener <rguenther@suse.de>
1977 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1979 2018-04-30 Richard Biener <rguenther@suse.de>
1981 PR tree-optimization/28364
1982 PR tree-optimization/85275
1983 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1984 copying first exit test.
1986 2018-04-28 Mark Wielaard <mark@klomp.org>
1988 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1990 (dwarf_AT): Handle DW_AT_addr_base.
1991 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1993 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1996 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1997 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1998 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1999 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2001 2018-04-28 Richard Biener <rguenther@suse.de>
2003 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2004 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2005 to reflect use. Only add interesting stmts.
2007 2018-04-27 Martin Jambor <mjambor@suse.cz>
2010 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2011 the jump function allows for passing through aggregate values.
2013 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2015 * input.h (in_system_header_at): Convert from macro to inline
2017 (from_macro_expansion_at): Likewise.
2018 (from_macro_definition_at): Likewise.
2020 2018-04-27 Jeff Law <law@redhat.com>
2022 * config.gcc: Mark tile* targets as deprecated/obsolete.
2024 2018-04-27 Richard Biener <rguenther@suse.de>
2026 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2029 2018-04-27 Richard Biener <rguenther@suse.de>
2031 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2033 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2035 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2036 with Yd constraint. Set "preferred_for_speed" attribute from
2037 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2039 (*movdi_internal): Ditto.
2040 (movti_interunit splitters): Remove
2041 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2042 (movdi_interunit splitters): Ditto.
2043 * config/i386/constraints.md (Ye): Remove.
2044 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2046 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2049 * config/aarch64/constraints.md (Usg): Limit to 31.
2052 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2054 PR tree-optimization/85529
2055 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2056 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2057 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2058 zero extension or masking of the MSB bit.
2059 (optimize_range_tests): Add FIRST_BB argument, pass it through
2060 to optimize_range_tests_var_bound.
2061 (maybe_optimize_range_tests, reassociate_bb): Adjust
2062 optimize_range_tests callers.
2064 2018-04-26 Richard Biener <rguenther@suse.de>
2065 Jakub Jelinek <jakub@redhat.com>
2067 * cgraph.h (symbol_table): Just declare debug method here.
2068 * symtab.c (symbol_table::debug): Define.
2070 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2072 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2074 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2076 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2077 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2078 (*movdi_internal): Substitute Yi and Yj constraint with x
2079 and Ym and Yn constraint with y constraint. Update "isa"
2080 attribute and set "preferred_for_speed" attribute from
2081 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2082 (*movsi_internal): Ditto.
2083 (*movdf_internal): Ditto.
2084 (*movsf_internal): Ditto.
2085 (*zero_extendsidi2): Ditto.
2086 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2087 (sse2_loadld): Ditto.
2088 (*vec_extract<ssevecmodelower>_0): Ditto.
2089 (*vec_extractv4si_0_zext_sse4): Ditto.
2090 (vec_concatv2di): Ditto.
2091 (*vec_dup<mode>): Ditto.
2092 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2093 * config/i386/constraints.md (Yi): Remove.
2098 2018-04-26 Nathan Sidwell <nathan@acm.org>
2100 * dumpfile.c (dump_open): New.
2101 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2102 (dump_finish): Detect stdio/stderr by value not name.
2104 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2106 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2108 2018-04-26 Tom de Vries <tom@codesourcery.com>
2111 * config/nvptx/nvptx.c (verify_neutering_jumps)
2112 (verify_neutering_labels): New function
2113 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2115 2018-04-26 Tom de Vries <tom@codesourcery.com>
2118 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2119 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2122 2018-04-26 Richard Biener <rguenther@suse.de>
2123 Tom de Vries <tom@codesourcery.com>
2126 * lto-streamer-out.c (output_function): Fixup loops if required to match
2127 discovery done in the reader.
2129 2018-04-26 Richard Biener <rguenther@suse.de>
2131 PR tree-optimization/85116
2132 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2133 have a loop exit from the single latch predecessor. Remove
2134 case of header with just condition.
2135 (ch_base::copy_headers): Exclude infinite loops from any
2137 (pass_ch::execute): Record exits.
2139 2018-04-26 Richard Biener <rguenther@suse.de>
2141 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2142 prologue cost vector and pass it to vect_get_load_cost.
2143 (vect_get_peeling_costs_all_drs): Likewise.
2144 (vect_peeling_hash_get_lowest_cost): Likewise.
2145 (vect_enhance_data_refs_alignment): Likewise.
2147 2018-04-26 Richard Biener <rguenther@suse.de>
2150 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2151 checking of integer<->pointer conversions.
2152 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2153 sign-/zero-extending pointer types.
2154 (expand_omp_for_static_chunk): Likewise.
2156 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2157 Jean Lee <xiaoyur347@gmail.com>
2159 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2160 (TARGET_ASAN_SHADOW_OFFSET): Define.
2161 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2162 true for -fsanitize=address.
2164 2018-04-25 Mark Wielaard <mark@klomp.org>
2166 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2169 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2171 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2172 than "alu", remove explicit "memory" and "imm_disp" attributes.
2173 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2176 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2177 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2178 gen_lowpart_no_emit.
2180 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2183 * config/i386/i386.c (ix86_expand_builtin): Change memory
2184 operand to XI, extend p0 to Pmode.
2185 * config/i386/i386.md: Change unspec volatile and operand
2186 1 mode to XI, change operand 0 mode to P.
2188 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2190 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2191 GET_MODE_MASK before any checking.
2192 (nds32_can_use_bset_p): Likewise.
2193 (nds32_can_use_btgl_p): Likewise.
2195 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2197 * config/nds32/nds32-doubleword.md: New define_split pattern for
2198 illegal register number.
2200 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2202 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2204 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2206 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2208 2018-04-25 Richard Biener <rguenther@suse.de>
2210 * lto-streamer.h (LTO_major_version): Bump to 8.
2212 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2214 * BASE-VER: Set to 9.0.0.
2216 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2218 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2219 in __abskf2 and __powikf2.
2221 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2224 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2225 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2226 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2227 Use the above on operand 2. Reindent.
2228 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2230 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2233 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2235 * config/i386/i386.opt (mcet): Removed.
2236 * doc/install.texi: Remove -mcet documentation.
2237 * doc/invoke.texi: Likewise.
2239 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2242 * doc/install.texi: Remove -mcet from bootstrap-cet.
2244 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2247 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2248 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2252 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2253 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2254 containing a CONST_VECTOR.
2256 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2258 * doc/install.texi: Update newlib dependency for nvptx.
2260 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2263 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2264 instead of INTVAL when shifting x left.
2266 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2268 PR tree-optimization/85478
2269 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2270 vect_grouped_store_supported for single element vectors.
2272 2018-04-24 Richard Biener <rguenther@suse.de>
2275 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2276 load cost increase to the case of non-constant step.
2278 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2281 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2282 destination if any_malformed_asm.
2284 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2287 * expr.c (store_field): In the bitfield case, if the value comes from
2288 a function call and is returned in registers by means of a PARALLEL,
2289 do not change the mode of the temporary unless BLKmode and VOIDmode.
2291 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2293 PR rtl-optimization/85423
2294 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2295 dependencies to debug insns when the previous insn is non-debug.
2297 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2299 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2300 enums into a single definition.
2301 (fls): Fix predicates and printing.
2304 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2306 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2307 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2308 and short u6 immediate.
2309 (check_if_valid_sleep_operand): Remove.
2310 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2312 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2314 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2315 flag_always_save_lp condition.
2316 * config/nds32/nds32.opt (malways-save-lp): New option.
2318 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2320 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2321 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2322 * config/nds32/nds32.h
2323 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2324 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2326 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2328 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2329 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2331 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2332 Chung-Ju Wu <jasonwucj@gmail.com>
2334 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2336 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2337 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2339 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2341 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2343 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2345 * config/nds32/nds32-protos.h (nds32_data_alignment,
2346 nds32_local_alignment): Declare.
2347 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2348 nds32_local_alignment): New functions.
2349 (TARGET_CONSTANT_ALIGNMENT): Define.
2350 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2352 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2354 * config/nds32/nds32.c
2355 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2356 (TARGET_MODES_TIEABLE_P): Likewise.
2358 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2360 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2363 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2364 Chung-Ju Wu <jasonwucj@gmail.com>
2366 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2367 for unaligned access.
2368 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2370 * config/nds32/nds32-intrinsic.md: Likewise.
2371 * config/nds32/nds32_intrinsic.h: Likewise.
2372 * config/nds32/nds32.h (nds32_builtins): Likewise.
2373 * config/nds32/nds32.opt (munaligned-access): New option.
2374 * config/nds32/nds32.c (nds32_asm_file_start): Display
2375 flag_unaligned_access status.
2377 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2379 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2380 -mno-relax is present.
2381 * config/riscv/linux.h (LINK_SPEC): Ditto.
2383 2018-04-20 Martin Sebor <msebor@redhat.com>
2386 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2388 (gimple_fold_builtin_stxcpy_chk): Same.
2389 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2391 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2394 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2395 __powikf2 when long double is IEEE 128-bit.
2397 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2399 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2400 step to make sure stack always aligned.
2402 2018-04-20 Carl Love <cel@us.ibm.com>
2405 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2406 size check for arg0.
2408 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2409 Tom de Vries <tom@codesourcery.com>
2412 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2413 Emit insns for calls too.
2414 (nvptx_find_par): Always look for worker-level predecessor insn.
2415 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2417 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2418 (nvptx_process_pars): Propagate frames for calls.
2420 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2423 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2425 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2426 (ix86_handle_option): Don't handle OPT_mibt.
2427 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2429 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2431 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2432 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2433 (ix86_target_macros): Define __CET__ with flag_cf_protection
2434 for -fcf-protection.
2435 * config/i386/i386.c (isa2_opts): Remove -mibt.
2436 * config/i386/i386.h (TARGET_IBT): Removed.
2437 (TARGET_IBT_P): Likewise.
2438 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2439 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2440 * config/i386/i386.opt (mcet): Update help message.
2443 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2444 -mcet as an alias for -mshstk.
2446 2018-04-20 Richard Biener <rguenther@suse.de>
2449 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2450 complexity by forcing a single use of the multiply operand.
2452 2018-04-20 Martin Jambor <mjambor@suse.cz>
2455 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2456 recursion dependency to only apply to non-clones.
2458 2018-04-20 Martin Jambor <mjambor@suse.cz>
2461 * ipa-cp.c (create_specialized_node): Check that clones of
2462 self-recursive edges exist during IPA-CP.
2464 2018-04-19 Toon Moene <toon@moene.org>
2466 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2469 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2471 PR tree-optimization/85467
2472 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2473 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2474 VECTOR_CST element to type.
2476 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2479 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2480 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2481 (builtin_longjmp): Likewise.
2482 (save_stack_nonlocal): New pattern.
2483 (restore_stack_nonlocal): Likewise.
2485 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2488 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2489 Replace ASM_OUTPUT_LABEL with fprintf.
2491 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2494 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2495 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2496 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2497 define __IBT__ and __SHSTK__ for -fcf-protection.
2498 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2500 (ix86_trampoline_init): Likewise.
2501 (x86_output_mi_thunk): Likewise.
2502 (ix86_notrack_prefixed_insn_p): Likewise.
2503 (ix86_option_override_internal): Don't disallow -fcf-protection.
2504 * config/i386/i386.md (rdssp<mode>): Also enable for
2506 (incssp<mode>): Likewise.
2507 (nop_endbr): Likewise.
2508 * config/i386/i386.opt (mcet): Change help message to built-in
2512 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2513 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2514 enable CET built-in functions.
2516 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2518 * common/config/i386/i386-common.c
2519 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2520 OPTION_MASK_ISA_MOVDIRI_UNSET,
2521 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2522 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2523 * config.gcc (movdirintrin.h): New header.
2524 * config/i386/cpuid.h (bit_MOVDIRI,
2525 bit_MOVDIR64B): New bits.
2526 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2528 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2529 (VOID, PVOID, PCVOID)): New function types.
2530 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2531 __builtin_ia32_directstoreu_u64,
2532 __builtin_ia32_movdir64b): New builtins.
2533 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2534 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2536 (ix86_valid_target_attribute_inner_p): Ditto.
2537 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2538 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2539 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2540 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2541 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2542 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2543 (movdiri<mode>, movdir64b_<mode>): New.
2544 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2545 * config/i386/immintrin.h: Include movdirintrin.h.
2546 * config/i386/movdirintrin.h: New file.
2547 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2549 2018-04-19 Richard Biener <rguenther@suse.de>
2552 * cfg.c (clear_bb_flags): When loop state says we have
2553 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2555 2018-04-19 Richard Biener <rguenther@suse.de>
2557 PR tree-optimization/84737
2558 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2559 copying restrict info.
2560 (vect_setup_realignment): Use it.
2561 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2562 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2563 the first DR to all generated stores.
2564 (vectorizable_load): Likewise for loads.
2566 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2568 PR tree-optimization/85446
2569 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2570 the integral and pointer types to have the same precision.
2572 * doc/install.texi: Document --disable-cet being the default and
2575 2018-04-18 Martin Liska <mliska@suse.cz>
2577 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2580 2018-04-18 Martin Liska <mliska@suse.cz>
2583 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2586 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2587 arguments if they are comparable.
2589 2018-04-18 Martin Liska <mliska@suse.cz>
2592 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2595 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2598 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2601 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2602 ENDBR after calling __morestack.
2604 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2607 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2608 by using gcc_base_ver to generate a gcc_driver_version, and use
2609 it when generating GCC_DRIVER_NAME.
2610 * configure: Regenerate.
2612 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2615 * config.gcc: Obsolete powerpc*-*-*spe*.
2617 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2620 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2621 (stabstr_D): Change type of unum from unsigned int to
2622 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2625 2018-04-17 Jim Wilson <jimw@sifive.com>
2628 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2629 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2630 Set arg_pointer_offset after using pretend_args_size.
2632 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2634 PR rtl-optimization/85431
2635 * dse.c (record_store): Ignore zero width stores.
2638 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2639 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2640 __builtin_stack_restore rather than after it.
2641 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2642 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2643 argument instead of virtual_dynamic_stack_rtx.
2645 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2647 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2649 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2650 Add note to error message to explain internal mapping of overloaded
2651 built-in function name to non-overloaded built-in function name.
2652 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2655 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2658 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2659 where the inputs overlap with the output.
2661 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2663 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2664 (=v, v) alternative and explicit "memory" attribute.
2665 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2666 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2668 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2669 "sselog1" type instead of "sselog".
2670 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2671 "sselog". Remove explicit "memory" attribute.
2672 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2673 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2675 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2676 "sselog1" type instead of "sselog". Remove explicit "memory"
2678 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2679 use "sselog1" type instead of "sselog". Remove explicit "memory"
2681 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2682 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2684 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2685 "sselog1" type instead of "sselog". Remove explicit "memory"
2687 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2688 use "sselog1" type instead of "sselog". Remove explicit "memory"
2692 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2695 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2698 2018-04-17 Martin Jambor <mjambor@suse.cz>
2701 * ipa-cp.c (create_specialized_node): Call
2702 expand_all_artificial_thunks if necessary.
2704 2018-04-17 Martin Liska <mliska@suse.cz>
2707 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2708 in message, remote space in between '_G' and '('.
2710 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2713 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2714 avx512f_vmcmp<mode>3<round_saeonly_name>,
2715 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2716 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2717 avx512f_rndscale<mode><round_saeonly_name>,
2718 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2719 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2720 Use %<iptr>2 instead of %2 for -masm=intel.
2721 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2722 avx512f_vcvttss2usi<round_saeonly_name>,
2723 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2725 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2726 avx512f_vcvttsd2usi<round_saeonly_name>,
2727 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2728 Use %q1 instead of %1 for -masm=intel.
2729 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2730 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2731 of %3 for -masm=intel.
2732 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2734 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2736 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2738 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2739 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2741 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2743 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2744 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2745 %g1 and one with %0 and %1.
2746 (avx512er_vmrcp28<mode><round_saeonly_name>,
2747 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2749 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2750 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2751 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2752 of %0 and %{%4%} for -masm=intel.
2753 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2754 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2755 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2756 order of %0 and %{%5%}%{z%} for -masm=intel.
2758 2018-04-17 Jan Hubicka <jh@suse.cz>
2761 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2763 2018-04-17 Martin Liska <mliska@suse.cz>
2766 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2767 for target_clone error message. Make default implementation
2768 clone to be a local declaration.
2769 (separate_attrs): Add new argument and check for an empty
2771 (expand_target_clones): Handle it.
2772 (ipa_target_clone): Make redirection just for target_clones
2775 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2776 Tom de Vries <tom@codesourcery.com>
2779 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2780 tiled basic blocks without omp continue statements.
2782 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2785 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2786 vec_extract expression as having side effects to make sure it gets
2789 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2792 * config/i386/i386.c (get_builtin_code_for_version): Check
2795 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2798 * gcc/config.gcc: Support "skylake".
2799 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2801 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2802 (processor_target_table): Add "skylake".
2803 (ix86_option_override_internal): Add "skylake".
2804 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2805 PROCESSOR_CANNONLAKE.
2806 (get_builtin_code_for_version): Fix priority for
2807 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2808 PROCESSOR_SKYLAKE-AVX512.
2809 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2810 (processor_type): Add PROCESSOR_SKYLAKE.
2812 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2813 Jason Merrill <jason@redhat.com>
2816 * convert.c (convert_to_integer_1): Use direct recursion for
2817 enumeral types and types with a precision less than the number
2818 of bits in their mode.
2820 2018-04-16 Julia Koval <julia.koval@intel.com>
2823 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2824 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2826 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2829 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2830 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2831 and -mno-direct-move.
2833 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2836 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2837 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2838 Ensure negative shifts result in {0}.
2840 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2842 PR rtl-optimization/79916
2843 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2844 regs (if any) to define how to gnerate SD moves when LRA is in
2847 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2849 PR rtl-optimization/85393
2850 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2851 * except.c (expand_dw2_landing_pad_for_region): Make static.
2852 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2853 a label and unconditional jump to old_bb, rather than
2854 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2857 PR rtl-optimization/85376
2858 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2859 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2860 instead of a specific value.
2862 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2863 Bin Cheng <bin.cheng@arm.com>
2865 PR tree-optimization/82965
2866 PR tree-optimization/83991
2867 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2868 by_profile_only parameter.
2869 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2870 information if the loop was predicted to iterate too many times.
2871 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2873 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2876 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2879 2018-04-13 Martin Liska <mliska@suse.cz>
2880 Jakub Jelinek <jakub@redhat.com>
2883 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2884 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2885 * builtins.c (expand_builtin_memory_copy_args): Use
2886 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2887 handle dest_addr == pc_rtx.
2889 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2892 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2893 asked to not generate direct moves.
2894 (fix_trunc<mode>si2_stfiwx): Similar.
2895 (fix_trunc<mode>si2_internal): Similar.
2897 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2900 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2901 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2902 lookup if dest in some wider mode is known to be const0_rtx and
2903 if so, record permanent equivalence for it to be ZERO_EXTEND of
2904 the narrower mode destination.
2906 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2908 * lto-streamer-out.c (output_function): Revert 259346.
2909 * omp-expand.c (expand_oacc_for): Likewise.
2911 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2913 PR rtl-optimization/85354
2914 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2915 * sel-sched.c (sel_global_init): ... here.
2917 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2920 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2921 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2922 mode for PE-COFF targets.
2923 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2924 (i386_pe_asm_lto_end): Likewise.
2925 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2926 (TARGET_ASM_LTO_END): Likewise.
2927 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2928 (i386_pe_asm_lto_start): New function.
2929 (i386_pe_asm_lto_end): Likewise.
2931 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2932 Richard Biener <rguenther@suse.de>
2935 * lto-streamer-out.c (output_function): Fix CFG loop state before
2937 * omp-expand.c (expand_oacc_for): Handle calls to internal
2938 functions like regular functions.
2940 2018-04-12 Richard Biener <rguenther@suse.de>
2943 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2944 for the early LTO debug to properly generate references to it
2945 during DIE emission. Do not re-use that for the skeleton for
2947 (dwarf2out_early_finish): Likewise.
2949 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2952 * config/i386/sse.md
2953 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2954 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2955 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2956 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2957 and output is a reg, avoid creating invalid lowpart subreg, but
2958 instead split into a 512-bit move. Don't split if not AVX512VL,
2959 input is xmm16+ reg and output is a mem.
2960 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2961 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2962 xmm16+ reg and output is a mem.
2964 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2966 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2967 also for flag_dwarf2_cfi_asm.
2969 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2971 PR rtl-optimization/85342
2972 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2973 a bool scalar var inside of the loop instead. Don't try to update
2974 recog_data.operand after failed apply_change_group.
2976 2018-04-12 Tom de Vries <tom@codesourcery.com>
2979 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2980 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2981 array with flexible array member as array without given dimension.
2982 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2983 argument for undefined param to true.
2985 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2988 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2989 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2990 from PowerPC section.
2991 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2992 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2993 help text that is too long.
2994 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2995 help text that is too long.
2996 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2997 help text that is too long.
2999 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3001 * config/alpha/alpha.md (stack_probe_internal): Rename
3002 from "probe_stack". Update all callers.
3004 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3006 PR rtl-optimization/84566
3007 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3008 sched_macro_fuse_insns.
3010 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3013 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3014 (compute_block_dependences): ... from here.
3016 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3018 PR tree-optimization/85331
3019 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3020 from int to HOST_WIDE_INT.
3022 2018-04-11 Martin Jambor <mjambor@suse.cz>
3025 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3026 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3027 not the same as the source val.
3028 (cgraph_edge_brings_value_p): New parameter.
3029 (gather_edges_for_value): Pass destination value to
3030 cgraph_edge_brings_value_p.
3031 (perhaps_add_new_callers): Likewise.
3032 (get_info_about_necessary_edges): Likewise and exclude values brought
3033 only by self-recursive edges.
3034 (create_specialized_node): Redirect only clones of self-calling edges.
3035 (+self_recursive_pass_through_p): New function.
3036 (find_more_scalar_values_for_callers_subset): Use it.
3037 (find_aggregate_values_for_callers_subset): Likewise.
3038 (known_aggs_to_agg_replacement_list): Removed.
3039 (decide_whether_version_node): Re-calculate known constants for all
3040 remaining context clones.
3042 2018-04-11 Richard Biener <rguenther@suse.de>
3045 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3046 from early DWARF output.
3047 (dwarf2out_early_finish): Output line info unconditionally into
3048 early DWARF and add reference to it.
3050 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3053 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3054 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3055 other than V2DFmode using iptr mode attribute.
3056 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3058 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3060 PR rtl-optimization/84659
3061 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3063 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3066 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3068 (output_loc_list): Pass address of a dummy size variable even in the
3069 locview handling loop.
3070 (index_location_lists): Add comment on why skip_loc_list_entry can't
3073 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3076 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3079 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3082 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3083 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3084 and -mstring-compare-inline-limit.
3086 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3089 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3090 for stack clash protection in a register whenever we need it to be in
3093 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3095 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3096 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3098 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3101 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3103 (mlong-double-): Ditto.
3104 * config/rs6000/sysv4.opt (msdata=): Ditto.
3105 (mtls-size=): Ditto.
3107 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3109 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3110 erroneous entries for
3111 "vector int vec_ldl (int, long int *)", and
3112 "vector unsigned int vec_ldl (int, unsigned long int *)".
3113 Add comments and entries for
3114 "vector bool char vec_ldl (int, bool char *)",
3115 "vector bool short vec_ldl (int, bool short *)",
3116 "vector bool int vec_ldl (int, bool int *)",
3117 "vector bool long long vec_ldl (int, bool long long *)",
3118 "vector pixel vec_ldl (int, pixel *)",
3119 "vector long long vec_ldl (int, long long *)",
3120 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3121 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3122 type tree bool_long_long_type_node and correct definition of
3123 bool_V2DI_type_node to make reference to this new type tree.
3124 (rs6000_mangle_type): Replace erroneous reference to
3125 bool_long_type_node with bool_long_long_type_node.
3126 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3127 comments to emphasize sign distinctions for char and int types and
3128 replace RS6000_BTI_bool_long constant with
3129 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3130 use of RS6000_BTI_pixel.
3131 (bool_long_type_node): Remove this macro definition.
3132 (bool_long_long_type_node): New macro definition
3134 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3136 PR rtl-optimization/85300
3137 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3138 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3139 simplify_unary_operation fails.
3141 2018-04-10 Martin Liska <mliska@suse.cz>
3143 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3144 cgraph_edge and ipa_ref.
3146 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3150 * config/i386/sse.md
3151 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3152 computation of the VEC_MERGE selector from mask.
3153 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3154 Fix decoding of the VEC_MERGE selector into mask.
3156 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3158 PR tree-optimization/85286
3159 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3161 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3163 * final.c (final_1): Set insn_last_address as well as
3164 insn_current_address.
3166 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3169 * explow.c (emit_stack_probe): Call validize_mem on memory location
3170 before passing it to gen_probe_stack. Create address operand and
3171 legitimize it for the probe_stack_address case.
3173 2018-04-09 Jan Hubicka <jh@suse.cz>
3176 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3177 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3178 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3180 (free_lang_data): Rebuild type inheritance graph.
3182 2018-04-09 Martin Sebor <msebor@redhat.com>
3184 * invoke.texi (-finline-small-functions): Mention other optimization
3186 (-findirect-inlining, -fpartial-inlining): Same.
3187 (-finline-functions-called-once): Same.
3188 (-freorder-blocks-and-partition): Same.
3190 2018-04-09 Jan Hubicka <jh@suse.cz>
3193 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3194 jumps; choose last target that matches the criteria (i.e.
3195 no partition changes for non-crossing jumps).
3196 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3197 support for redirecting crossing jumps to non-crossing.
3199 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3201 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3202 also for naked functions.
3204 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3206 * config/arc/arc.md (add_shift): New pattern.
3207 (add_shift2): Likewise.
3208 (sub_shift): Likewise.
3209 (sub_shift_cmp0_noout): Likewise.
3210 (compare_si_ashiftsi): Likewise.
3211 (xbfu_cmp0_noout): New combine pattern.
3212 (xbfu_cmp0"): Likewise.
3213 (movsi_set_cc_insn): Place the predicable variant first.
3214 (commutative_binary_cmp0_noout): Remove clobber.
3215 (commutative_binary_cmp0): New pattern.
3216 (noncommutative_binary_cmp0): Likewise.
3217 (noncommutative_binary_cmp0_noout): Likewise.
3218 (noncommutative_binary_comparison_result_used): Removed.
3219 (rsub_cmp0): New pattern.
3220 (rsub_cmp0_noout): Likewise.
3221 (extzvsi): Changed, keep only meaningful variants.
3222 (SQH, SEZ): New iterators.
3223 (SQH_postfix): New mode attribute.
3224 (SEZ_prefix): New code attribute.
3225 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3226 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3227 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3229 (noncommutative_operator): Check the availability of barrel
3232 2018-04-09 Richard Biener <rguenther@suse.de>
3234 PR tree-optimization/85284
3235 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3236 Only use the niter constraining form of simple_iv when the exit
3239 2018-04-09 Tom de Vries <tom@codesourcery.com>
3242 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3243 (define_expand "*memory_barrier"): New define_expand.
3244 (define_insn "memory_barrier"): New insn.
3246 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3248 PR rtl-optimization/80463
3249 PR rtl-optimization/83972
3250 PR rtl-optimization/83480
3252 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3253 correct producer for the insn.
3254 (tidy_control_flow): Fixup seqnos in case of debug insns.
3256 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3258 PR rtl-optimization/83913
3260 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3261 different sched-times when merging exprs.
3263 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3265 PR rtl-optimization/83962
3267 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3268 tidy_fallthru_edge and tidy_control_flow.
3270 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3272 PR rtl-optimization/83530
3274 * sel-sched.c (force_next_insn): New global variable.
3275 (remove_insn_for_debug): When force_next_insn is true, also leave only
3276 next insn in the ready list.
3277 (sel_sched_region): When the region wasn't scheduled, make another pass
3278 over it with force_next_insn set to 1.
3280 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3282 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3284 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3285 for interrupt control.
3286 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3287 functions for interrupt control.
3288 * config/nds32/nds32-intrinsic.md: Likewise.
3289 * config/nds32/nds32_intrinsic.h: Likewise.
3290 * config/nds32/nds32.h (nds32_builtins): Likewise.
3292 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3294 * config/nds32/nds32.c (nds32_init_machine_status,
3295 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3296 strict_aligned_p field.
3297 (nds32_expand_to_rtl_hook): New function.
3298 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3299 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3301 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3302 Chung-Ju Wu <jasonwucj@gmail.com>
3304 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3305 * config/nds32/nds32-n7.md: New file.
3306 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3307 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3309 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3310 * config/nds32/nds32.md (pipeline_model): Add n7.
3311 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3312 * config/nds32/pipelines.md: Include n7 settings.
3314 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3315 Chung-Ju Wu <jasonwucj@gmail.com>
3317 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3318 * config/nds32/nds32-e8.md: New file.
3319 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3320 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3322 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3323 * config/nds32/nds32.md (pipeline_model): Add e8.
3324 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3325 * config/nds32/pipelines.md: Include e8 settings.
3327 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3328 Chung-Ju Wu <jasonwucj@gmail.com>
3330 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3331 * config/nds32/nds32-n8.md: New file.
3332 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3333 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3335 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3336 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3337 * config/nds32/nds32.md (pipeline_model): Add n8.
3338 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3339 * config/nds32/pipelines.md: Include n8 settings.
3341 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3342 Chung-Ju Wu <jasonwucj@gmail.com>
3344 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3345 * config/nds32/nds32-n9-2r1w.md: New file.
3346 * config/nds32/nds32-n9-3r2w.md: New file.
3347 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3348 nds32_register_ports): New or modify for cpu n9.
3349 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3351 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3352 * config/nds32/nds32-utils.c: New file.
3353 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3354 TARGET_MUL_SLOW): Define.
3355 * config/nds32/nds32.md (pipeline_model): New attribute.
3356 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3357 New options that support cpu n9.
3358 * config/nds32/pipelines.md: Include n9 settings.
3359 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3361 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3363 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3364 information if necessary.
3365 (output_cond_branch_compare_zero): Likewise.
3366 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3367 (nds32_target_alignment): Refine for alignment.
3368 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3369 (FUNCTION_BOUNDARY): Modify.
3370 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3372 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3374 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3376 * config/nds32/constants.md (unspec_volatile_element): Add values for
3377 TLB operation and data prefetch.
3378 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3379 functions for TLB operation and data prefetch.
3380 * config/nds32/nds32-intrinsic.md: Likewise.
3381 * config/nds32/nds32_intrinsic.h: Likewise.
3382 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3383 (nds32_print_operand): Likewise.
3384 * config/nds32/nds32.h (nds32_builtins): Likewise.
3386 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3387 Andrew Pinski <pinsika@gcc.gnu.org>
3390 * match.pd: Use constant_boolean_node of correct type instead of
3391 boolean_true_node or boolean_false_node for simplifying
3392 pointer comparisons to zero.
3394 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3396 PR tree-optimization/80021
3397 * tree.c (verify_type_variant): Make error call in verify_variant_match
3398 translatable and remove final full stop.
3400 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3402 * config/nds32/constants.md (unspec_volatile_element): Add
3403 UNSPEC_VOLATILE_EH_RETURN.
3404 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3405 nds32_output_stack_pop): Support dwarf exception handling process.
3406 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3407 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3408 exception handling process.
3409 (nds32_compute_stack_frame): Likewise.
3410 (nds32_return_addr_rtx): Likewise.
3411 (nds32_initial_elimination_offset): Likewise.
3412 (nds32_expand_prologue): Likewise.
3413 (nds32_expand_epilogue): Likewise.
3414 (nds32_dynamic_chain_address): New function.
3415 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3417 (DYNAMIC_CHAIN_ADDRESS): Define.
3418 (EH_RETURN_DATA_REGNO): Define.
3419 (EH_RETURN_STACKADJ_RTX): Define.
3420 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3421 patterns for dwarf exception handling.
3423 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3425 * config/nds32/nds32.h: Clean up obsolete macros.
3427 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3429 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3430 Add enum values for particular instructions.
3431 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3432 particular intrinsic functions.
3433 * config/nds32/nds32-intrinsic.md: Likewise.
3434 * config/nds32/nds32_intrinsic.h: Likewise.
3435 * config/nds32/nds32.h (nds32_builtins): Likewise.
3436 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3437 (btst, ave): New patterns for particular instructions.
3439 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3441 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3442 Add enum values for atomic load/store and memory sync.
3443 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3445 * config/nds32/nds32-intrinsic.md: Likewise.
3446 * config/nds32/nds32_intrinsic.h: Likewise.
3447 * config/nds32/nds32.h (nds32_builtins): Likewise.
3449 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3451 PR tree-optimization/85257
3452 * fold-const.c (native_encode_vector): If not all elts could fit
3453 and off is -1, return 0 rather than offset.
3454 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3455 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3456 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3457 adjust buffer in native_interpret_expr call.
3459 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3461 * config/nds32/constants.md (unspec_volatile_element): Add cache
3462 control enum values.
3463 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3464 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3465 * config/nds32/nds32.c (nds32_cctl_names): New.
3466 (nds32_print_operand): Handle cache control register names.
3467 * config/nds32/nds32.h (nds32_builtins): New enum values.
3468 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3470 * config/nds32/nds32.md (type): Add mmu.
3471 * config/nds32/pipelines.md (simple_insn): Add mmu.
3473 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3475 * config/nds32/nds32.md (type): Remove call.
3476 * config/nds32/pipelines.md (simple_insn): Likewise.
3478 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3480 * config/nds32/constants.md (unspec_volatile_element): Add
3481 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3482 UNSPEC_VOLATILE_FMFCFG.
3483 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3484 description for fmfcfg and fmfcsr.
3485 (bdesc_1arg): Add fmtcsr.
3486 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3487 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3488 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3489 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3490 unspec_fmfcfg): New patterns.
3491 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3492 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3493 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3494 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3495 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3496 __nds32__fmfcfg): Define.
3498 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3500 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3501 intrinsic register names.
3502 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3503 intrinsic register enum values and macros.
3505 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3507 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3508 for load/store addressing form.
3509 (nds32_print_operand_address): Likewise.
3511 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3514 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3515 based on LABEL_REF. Remove useless assertion.
3516 (pic_address_needs_scratch): Fix formatting.
3517 (sparc_legitimize_pic_address): Minor tweaks.
3518 (sparc_delegitimize_address): Adjust assertion accordingly.
3519 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3520 into symbolic_operand.
3521 (movsi_high_pic_label_ref): Likewise.
3522 (movsi_lo_sum_pic_label_ref): Likewise.
3523 (movdi_pic_label_ref): Likewise.
3524 (movdi_high_pic_label_ref): Likewise.
3525 (movdi_lo_sum_pic_label_ref): Likewise.
3527 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3529 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3530 custom LIB_SPEC setup.
3532 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3533 Kito Cheng <kito.cheng@gmail.com>
3535 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3536 * config/riscv/freebsd.h: New.
3538 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3540 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3541 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3544 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3545 Kito Cheng <kito.cheng@gmail.com>
3547 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3548 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3549 * config/nds32/nds32-protos.h (nds32_output_call,
3550 nds32_output_return): Declare.
3551 * config/nds32/nds32.md: Refine all the call and return patterns.
3553 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3556 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3557 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3559 PR rtl-optimization/84872
3560 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3561 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3564 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3566 * expr.c (copy_blkmode_to_reg): Revert 254862.
3567 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3569 2018-04-06 Richard Biener <rguenther@suse.de>
3572 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3573 after seeing a component reference with an adjacent field. Treat
3574 refs to arrays at struct end of external decls similar to
3575 refs to unconstrained commons.
3577 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3580 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3581 look through SAVE_EXPRs with non-side-effects argument. Adjust
3583 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3586 2018-04-06 Richard Biener <rguenther@suse.de>
3589 * alias.c (find_base_term): New wrapper around find_base_term
3590 unwinding CSELIB_VAL_PTR changes.
3591 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3594 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3596 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3598 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3599 constant definitions.
3600 ("nop"): lr 0,0 -> nopr r0
3601 ("nop_lr0", "nop_lr1"): New insn definitions.
3603 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3605 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3606 NDS32_V3PUSH_AVAILABLE_P macro.
3608 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3609 Chung-Ju Wu <jasonwucj@gmail.com>
3611 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3612 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3613 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3614 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3615 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3616 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3617 * config/nds32/constraints.md: New constraints and checking for hard
3618 float configuration.
3619 * config/nds32/iterators.md: New mode iterator and attribute for hard
3620 float configuration.
3621 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3623 * config/nds32/nds32-fpu.md: New file.
3624 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3625 deal with hard float code generation.
3626 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3628 (abi_type, float_reg_number): New enum type.
3629 * config/nds32/nds32-predicates.c: New predicates for hard float.
3630 * config/nds32/nds32-protos.h: Declare functions for hard float.
3631 * config/nds32/nds32.c: Implementation for hard float configuration.
3632 * config/nds32/nds32.h: Definitions for hard float configuration.
3633 * config/nds32/nds32.md: Include hard float machine description and
3634 modify patterns for hard float configuration.
3635 * config/nds32/nds32.opt: New options for hard float configuration.
3636 * config/nds32/predicates.md: New predicates for hard float
3639 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3641 * common/config/nds32/nds32-common.c
3642 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3644 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3647 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3648 CONSTRUCTOR_ELT (ctor, ...)->value.
3650 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3653 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3655 2018-04-05 Tom de Vries <tom@codesourcery.com>
3658 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3661 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3662 Kito Cheng <kito.cheng@gmail.com>
3664 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3665 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3666 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3667 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3669 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3670 Kito Cheng <kito.cheng@gmail.com>
3672 * config/nds32/constraints.md (Ufe): New memory constraint.
3673 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3674 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3675 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3677 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3678 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3680 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3682 * config/nds32/nds32.md: Use optimize_size in the condition for
3683 alu-shift instructions.
3685 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3687 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3689 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3691 * config/nds32/nds32.md (negsi2): Refine pattern.
3693 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3694 Chung-Ju Wu <jasonwucj@gmail.com>
3696 * config/nds32/iterators.md (shift_rotate): New code iterator.
3697 (shift): New code attribute.
3698 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3699 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3700 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3701 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3702 bit-wise operations.
3703 (andsi3, *andsi3): Ditto.
3704 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3705 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3706 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3707 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3708 nds32_ior_operand, nds32_xor_operand): New predicates.
3710 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3712 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3713 (addsi3, subsi3): ... this.
3715 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3717 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3719 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3721 * config/nds32/nds32.md: Adjust indention.
3723 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3725 * config/nds32/nds32.md (feature): New attribute.
3727 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3729 * config/nds32/nds32.md (subtype): New attribute.
3731 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3734 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3735 expansion to perform a bitwise AND of the argument followed by a
3736 boolean negation of the result.
3738 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3740 PR rtl-optimization/84878
3741 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3742 the basic block. Assert the use reference is not artificial and that
3743 it has an associated insn.
3745 2018-04-04 Michael Matz <matz@suse.de>
3747 * builtins.c (compute_objsize): Pass correct operand
3748 to array_at_struct_end_p.
3750 2018-04-04 Richard Biener <rguenther@suse.de>
3753 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3754 from contexts for DINFO_LEVEL_TERSE and below.
3756 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3758 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3759 resiter_operand condition.
3760 * config/nds32/nds32.md (*move<mode>): Ditto.
3762 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3763 Monk Chiang <sh.chiang04@gmail.com>
3765 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3767 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3769 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3771 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3772 Kito Cheng <kito.cheng@gmail.com>
3774 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3775 nds32_cond_code_str, output_cond_branch,
3776 output_cond_branch_compare_zero, nds32_expand_cbranch,
3777 nds32_expand_cstore, nds32_expand_movcc,
3778 nds32_output_cbranchsi4_equality_zero,
3779 nds32_output_cbranchsi4_equality_reg,
3780 nds32_output_cbranchsi4_equality_reg_or_const_int,
3781 nds32_output_cbranchsi4_greater_less_zero: New functions.
3782 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3783 nds32_expand_cstore, nds32_expand_movcc,
3784 nds32_output_cbranchsi4_equality_zero,
3785 nds32_output_cbranchsi4_equality_reg,
3786 nds32_output_cbranchsi4_equality_reg_or_const_int,
3787 nds32_output_cbranchsi4_greater_less_zero): Declare.
3788 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3789 nds32_rimm11s_operand): New predicates.
3790 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3791 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3794 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3796 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3797 * config/nds32/nds32.md: Ditto.
3798 * config/nds32/pipelines.md: Ditto.
3800 2018-04-04 Richard Biener <rguenther@suse.de>
3802 PR tree-optimization/85168
3803 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3804 propagating abnormals.
3806 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3808 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3810 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3811 Kito Cheng <kito.cheng@gmail.com>
3813 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3814 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3815 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3816 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3817 * config/nds32/nds32.md (sibcall_internal): New.
3818 (sibcall_register): Remove.
3819 (sibcall_immediate): Remove.
3820 (sibcall_value_internal): New.
3821 (sibcall_value_register): Remove.
3822 (sibcall_value_immediate): Remove.
3823 * config/nds32/predicates.md (nds32_general_register_operand): New.
3824 (nds32_call_address_operand): New.
3826 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3828 PR rtl-optimization/85167
3829 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3830 bb_defs if *split_p, instead preinitialize it to NULL.
3832 PR tree-optimization/85156
3833 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3834 evaluating the argument multiple times.
3836 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3838 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3840 (_mm_cvtpd_ps): Likewise.
3841 (_mm_cvttpd_epi32): Likewise.
3842 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3843 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3844 vector, pixel, and bool following altivec.h include.
3846 2018-04-03 Martin Sebor <msebor@redhat.com>
3848 * doc/extend.texi (Common Function Attributes): Clarify.
3849 (const attribute): Likewise.
3850 (pure attribute): Likewise.
3852 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3855 * config/i386/i386.c (ix86_expand_vector_set): Use
3856 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3858 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3860 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3861 instructions when changing rounding bits to preserve precision bits
3862 in the x87 control word.
3864 2018-04-03 Martin Liska <mliska@suse.cz>
3866 PR tree-optimization/82491
3867 * rtl.h (strip_offset_and_add): Replace += suboffset with
3868 poly_uint64 () + suboffset.
3870 2018-03-29 Martin Liska <mliska@suse.cz>
3871 Martin Jambor <mjambor@suse.cz>
3874 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3875 param_type is not an integral or pointer type.
3877 2018-04-03 Richard Biener <rguenther@suse.de>
3879 * sese.h (recompute_all_dominators): Remove.
3881 2018-04-02 Martin Sebor <msebor@redhat.com>
3883 * doc/invoke.texi (-Wrestrict): Fix typos.
3885 2018-04-02 Jim Wilson <jimw@sifive.com>
3887 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3888 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3889 (<optab>di3, <optab>si3_extend): Likewise.
3890 (<optab>si3_mask, <optab>si3_mask_1): New.
3891 (<optab>di3_mask, <optab>di3_mask_1): New.
3892 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3893 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3894 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3896 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3898 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3901 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3903 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3904 (nds32_canonicalize_comparison): New function.
3906 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3907 Kito Cheng <kito.cheng@gmail.com>
3908 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3910 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3911 * config/nds32/constants.md (unspec_volatile_element): Add
3912 UNSPEC_VOLATILE_RELAX_GROUP.
3913 * config/nds32/nds32-relax-opt.c: New file.
3914 * config/nds32/nds32-predicates.c
3915 (nds32_symbol_load_store_p): New function.
3916 * config/nds32/nds32-protos.h
3917 (nds32_symbol_load_store_p): Declare function.
3918 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3919 * config/nds32/nds32.c
3920 (nds32_register_pass): New function to register pass.
3921 (nds32_register_passes): New function to register passes.
3922 * config/nds32/nds32.md (relax_group): New pattern.
3923 * config/nds32/nds32.opt (mrelax-hint): New option.
3924 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3926 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3928 * config/nds32/t-nds32: Modify files dependency.
3930 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3932 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3933 (PROFILE_HOOK): Define its implementation.
3935 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3937 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3938 type and 32-bit size.
3940 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3943 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3944 (V_128_256): New mode iterator.
3945 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3946 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3947 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3949 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3950 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3952 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3955 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3956 NaN inputs correctly.
3958 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3961 * config/rs6000/vsx.md (??r): New mode attribute.
3962 (*vsx_mov<mode>_64bit): Use it.
3963 (*vsx_mov<mode>_32bit): Likewise.
3965 2018-03-30 Martin Sebor <msebor@redhat.com>
3967 PR tree-optimization/84818
3968 * builtins.c (check_access): Use warning_n.
3970 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3973 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3975 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3978 2018-03-30 Julia Koval <julia.koval@intel.com>
3981 * x86-tune.def (movx, partial_reg_dependency): Enable for
3984 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3987 * lra-constraints.c (process_alt_operands): Move setting
3988 this_alternative_matches below.
3990 2018-03-29 Martin Liska <mliska@suse.cz>
3993 * doc/invoke.texi: Document how LTO works with debug info.
3994 Describe auto-load support of binutils. Mention 'x86-64'
3995 as valid option value of -march option.
3997 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3999 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4002 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4003 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4004 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4007 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4010 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4011 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4012 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4013 for RS6000_BTM_POWERPC64.
4014 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4015 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4016 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4021 2018-03-28 Carl Love <cel@us.ibm.com>
4024 2017-09-27 Carl Love <cel@us.ibm.com>
4026 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4027 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4028 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4031 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4033 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4034 instead of __vector bool.
4035 (_mm_max_pu8): Likewise.
4036 (_mm_min_pi16): Likewise.
4038 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4041 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4042 (DIVWEUO): Likewise.
4044 (DIVDEUO): Likewise.
4045 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4046 DIVWEUO and DIVDEUO.
4047 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4048 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4049 (div_extend): Likewise.
4050 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4052 (__builtin_divweuo): Likewise.
4053 (__builtin_divdeo): Likewise.
4054 (__builtin_divdeuo): Likewise.
4056 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4059 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4060 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4062 PR tree-optimization/82004
4063 * gimple-match-head.c (optimize_pow_to_exp): New function.
4064 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4065 Don't fold to exp if optimize_pow_to_exp is false.
4067 2018-03-28 Martin Liska <mliska@suse.cz>
4070 * calls.c (initialize_argument_information): Fix trailing space.
4071 * common.opt: Fix typo and provide better explanation for
4072 -fsanitize-coverage option.
4073 * config/i386/i386.opt: Fix typo.
4075 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4076 Martin Liska <mliska@suse.cz>
4079 * gimplify.c (asan_poison_variable): Don't do the check for
4080 gimplify_omp_ctxp here.
4081 (gimplify_decl_expr): Do it here.
4082 (gimplify_target_expr): Likewise.
4084 2018-03-28 Martin Liska <mliska@suse.cz>
4087 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4088 chkp_type_bounds_count if MPX is not enabled.
4090 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4092 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4094 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4097 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4098 function to create the function decl for complex long double
4099 multiply and divide for -mabi=ieeelongdouble.
4100 (init_float128_ieee): Call it.
4102 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4105 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4106 -fcf-protection=branch -mibt.
4107 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4109 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4112 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4114 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4117 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4118 extern array declarations.
4120 2018-03-27 Richard Biener <rguenther@suse.de>
4123 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4124 explicit single_use checks.
4126 2018-03-27 Richard Biener <rguenther@suse.de>
4128 PR tree-optimization/85082
4129 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4132 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4134 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4135 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4136 Turn on fasynchronous-unwind-tables and funwind-tables.
4138 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4141 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4142 (*bmi_blsr_<mode>_ccz): Ditto.
4144 2018-03-26 Tom de Vries <tom@codesourcery.com>
4146 PR tree-optimization/85063
4147 * omp-general.c (offloading_function_p): New function. Factor out
4149 * omp-offload.c (pass_omp_target_link::gate): ... here.
4150 * omp-general.h (offloading_function_p): Declare.
4151 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4152 with attribute omp declare target for offloading functions.
4154 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4156 PR tree-optimization/84005
4157 * tree-data-ref.h (get_base_for_alignment): Declare.
4158 * tree-data-ref.c (get_base_for_alignment_1): New function.
4159 (get_base_for_alignment): Likewise.
4160 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4161 get_base_for_alignment to find a suitable base object, instead
4162 of always using drb->base_address.
4164 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4167 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4168 known size by default.
4170 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4173 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4174 and non BLKmode operands.
4176 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4179 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4180 Clean up attributes.
4182 2018-03-23 Richard Biener <rguenther@suse.de>
4185 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4186 we are going to emit early debug for LTO.
4188 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4191 * function.c (match_asm_constraints_1): Don't optimize if input
4192 doesn't satisfy general_operand predicate for output's mode.
4195 * alias.c (write_dependence_p): Don't require for x_canonicalized
4196 non-VOIDmode if x has VOIDmode.
4199 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4200 just don't try to optimize it rather than assert it never happens.
4202 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4204 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4205 macro expansions for definition of ST_INTERNAL_<mode> and
4206 LD_INTERNAL_<mode> builtins.
4207 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4209 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4211 (altivec_expand_st_builtin): Likewise.
4212 (altivec_expand_builtin): Remove calls to deleted functions.
4213 (rs6000_address_for_altivec): Delete this function.
4214 * config/rs6000/vector.md: Remove expands for
4215 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4217 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4220 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4221 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4222 re-computing once computed.
4223 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4224 (arm_init_machine_status): Initialize
4225 machine->static_chain_stack_bytes.
4227 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4230 * doc/extend.texi: Add four new prototypes for vec_ld.
4231 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4232 definitions for more logical presentation.
4233 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4234 entries for V1TI variants of __builtin_altivec_ld builtin.
4235 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4236 handling of V1TI variant of LVX icode pattern.
4237 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4238 (rs6000_gimple_fold_builtin): Likewise.
4239 (altivec_init_builtins): Add code to define
4240 __builtin_altivec_lvx_v1ti function.
4242 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4245 * function.c (match_asm_constraints_1): Don't do the optimization
4246 if input isn't a REG, SUBREG, MEM or constant.
4248 2018-03-22 Tom de Vries <tom@codesourcery.com>
4250 PR tree-optimization/84956
4251 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4252 bb_has_abnormal_pred.
4254 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4257 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4258 DECL_INITIAL (decl) to decl at the end.
4259 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4262 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4264 * doc/extend.texi (__builtin_tgmath): Document when complex
4265 integer types are treated as _Complex _Float64.
4267 2018-03-21 Tom de Vries <tom@codesourcery.com>
4269 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4271 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4273 PR tree-optimization/84960
4274 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4275 if it is ENTRY block, move them into single succ of ENTRY in that case.
4277 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4279 PR tree-optimization/84811
4280 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4281 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4282 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4283 (known_subrange_p): Update accordingly. Cast each value involved
4284 in the size comparison, rather than casting the result of the
4287 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4289 PR tree-optimization/84982
4290 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4291 by flipping the least significant bit rather than all bits from
4292 bitpos to bitpos + bitsize - 1.
4294 2018-03-21 Nathan Sidwell <nathan@acm.org>
4296 * doc/extend.texi (Deprecated Features): Remove mention of
4297 long-deleted deprecations.
4299 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4302 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4303 * configure: Regenerate.
4305 2018-03-21 Tom de Vries <tom@codesourcery.com>
4307 PR tree-optimization/83126
4308 * tree-parloops.c (num_phis): New function.
4309 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4311 2018-03-21 Nathan Sidwell <nathan@acm.org>
4313 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4314 mention anon-struct/union members and trailing attributes.
4316 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4318 PR tree-optimization/84969
4319 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4320 builtin memset partitions if they set different rhs values.
4322 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4324 PR rtl-optimization/84989
4325 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4326 VEC_DUPLICATE with scalar result mode.
4328 2018-03-21 Martin Liska <mliska@suse.cz>
4331 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4332 not intended return statement.
4334 2018-03-21 Martin Liska <mliska@suse.cz>
4337 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4338 (chkp_find_bound_slots_1): Limit number of iterations.
4340 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4343 * Minor grammar fixes for x86 options.
4345 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4348 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4349 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4351 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4354 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4355 (altivec_lvx_<mode>_1op): Likewise.
4356 (altivec_stvx_<mode>_2op): Likewise.
4357 (altivec_stvx_<mode>_1op): Likewise.
4358 (altivec_lvx_<VM2:mode>): New define_expand.
4359 (altivec_stvx_<VM2:mode>): Likewise.
4360 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4361 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4362 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4363 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4364 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4365 (rs6000_gen_lvx): Likewise.
4366 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4367 (altivec_expand_stv_builtin): Likewise.
4368 (altivec_expand_builtin): Likewise.
4369 * config/rs6000/vector.md: Likewise.
4371 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4374 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4377 2018-03-20 Richard Biener <rguenther@suse.de>
4380 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4381 sign-conversions as zero, fall back to standard scalar_stmt
4384 2018-03-20 Martin Liska <mliska@suse.cz>
4387 * predict.c (rebuild_frequencies): Handle case when we have
4388 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4390 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4393 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4394 flag_section_anchors.
4395 * varasm.c (use_blocks_for_decl_p): Remove hack for
4396 dw2_force_const_mem.
4399 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4401 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4402 be created, use lowpart_subreg of operands[0] rather than operands[0]
4404 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4405 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4406 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4407 and n constraint instead of aarch64_shift_imm_di and Usd.
4408 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4409 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4411 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4414 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4415 to favor GPR over NEON registers.
4416 (<shift>di3_neon): Likewise.
4418 2018-03-20 Tom de Vries <tom@codesourcery.com>
4421 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4422 (nvptx_process_pars): Emit bar.sync asap and alap.
4424 2018-03-20 Tom de Vries <tom@codesourcery.com>
4427 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4428 seen_label if seen_label is already set.
4430 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4433 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4434 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4435 Use 1U instead of 1. Formatting fixes.
4438 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4439 instead of TREE_TYPE (s1) for the return value.
4441 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4443 PR tree-optimization/84946
4444 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4445 bitsize + bitsize in poly_uint64 rather than poly_int64.
4448 * dwarf2asm.c: Include fold-const.c.
4449 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4450 of decl rather than decl itself.
4452 PR rtl-optimization/84643
4453 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4455 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4458 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4459 calling assemble_variable.
4461 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4464 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4465 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4467 2018-03-19 Jim Wilson <jimw@sifive.com>
4470 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4471 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4472 (riscv_first_stack_step): Likewise.
4473 (riscv_option_override): Use STACK_BOUNDARY instead of
4475 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4477 (BIGGEST_ALIGNMENT): Set to 128.
4478 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4479 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4482 2018-03-19 Richard Biener <rguenther@suse.de>
4484 PR tree-optimization/84933
4485 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4486 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4488 2018-03-19 Richard Biener <rguenther@suse.de>
4490 PR tree-optimization/84859
4491 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4492 (cond_if_else_store_replacement): Perform sinking operation on
4493 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4494 Generalize what a BB with a single eligible store is.
4496 2018-03-19 Richard Biener <rguenther@suse.de>
4498 PR tree-optimization/84929
4499 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4500 chrec_is_positive against non-chrec arg.
4502 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4505 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4507 2018-03-18 Martin Liska <mliska@suse.cz>
4509 PR rtl-optimization/84635
4510 * regrename.c (build_def_use): Use matches_mode only when
4513 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4515 PR tree-optimization/84913
4516 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4517 vectorize chains of COND_EXPRs.
4519 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4521 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4523 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4525 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4527 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4529 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4531 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4532 Kito Cheng <kito.cheng@gmail.com>
4534 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4535 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4536 (nds32_adjust_reg_alloc_order): New function.
4537 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4539 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4541 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4542 nds32_print_operand, nds32_print_operand_address): Use
4543 HOST_WIDE_INT_PRINT_DEC instead.
4545 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4547 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4549 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4552 * config/i386/i386.c (initial_ix86_tune_features,
4553 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4555 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4556 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4557 rather than 1u << ix86_tune. Formatting fix.
4558 (ix86_option_override_internal): Change ix86_arch_mask from
4559 unsigned int to unsigned HOST_WIDE_INT, initialize to
4560 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4561 (ix86_function_specific_restore): Likewise.
4563 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4566 * postreload.c (reload_combine_recognize_pattern): Perform
4567 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4568 truncate_int_for_mode the result for the destination's mode.
4571 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4572 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4574 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4577 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4580 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4581 regno_allocno_class_array and sorted_pseudos.
4582 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4583 insns where regno is used.
4585 2018-03-16 Martin Liska <mliska@suse.cz>
4588 * multiple_target.c (create_dispatcher_calls): Redirect
4589 reference in the symbol table.
4591 2018-03-16 Martin Liska <mliska@suse.cz>
4594 * multiple_target.c (create_dispatcher_calls): Redirect also
4597 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4601 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4602 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4603 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4605 2018-03-16 Julia Koval <julia.koval@intel.com>
4607 * doc/invoke.texi (Skylake Server): Add CLWB.
4608 Cannonlake): Remove CLWB.
4610 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4612 PR tree-optimization/84841
4613 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4615 (FLOAT_ONE_CONST_TYPE): Define.
4616 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4617 (sort_by_operand_rank): Put entries with higher constant_type last
4618 rather than first to match comments.
4620 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4622 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4625 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4628 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4629 check and use address of target always.
4631 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4634 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4635 (indirect_thunk_bnd_needed): Likewise.
4636 (indirect_thunks_used): Likewise.
4637 (indirect_thunks_bnd_used): Likewise.
4638 (indirect_return_needed): New.
4639 (indirect_return_bnd_needed): Likewise.
4640 (output_indirect_thunk_function): Add a bool argument for
4642 (output_indirect_thunk_function): Don't generate alias for
4643 function return thunk.
4644 (ix86_code_end): Call output_indirect_thunk_function to generate
4645 function return thunks.
4646 (ix86_output_function_return): Set indirect_return_bnd_needed
4647 and indirect_return_needed instead of indirect_thunk_bnd_needed
4648 and indirect_thunk_needed.
4650 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4652 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4653 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4654 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4656 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4657 Paul Hua <paul.hua.gm@gmail.com>
4660 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4662 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4664 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4665 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4668 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4671 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4672 instead of GET_MODE_SIZE when comparing Units.
4674 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4677 * varasm.c (hash_section): Return an unchangeble hash value
4678 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4679 Return !aarch64_can_use_per_function_literal_pools_p ().
4681 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4684 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4685 rather than address of cmode as last argument to prepare_cmp_insn.
4687 2018-03-15 Julia Koval <julia.koval@intel.com>
4689 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4690 F_AVX512VNNI, F_AVX512BITALG): New.
4692 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4695 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4696 insn for floating-point loads and stores.
4698 2018-03-14 Carl Love <cel@us.ibm.com>
4700 * config/rs6000/rs6000-c.c: Add macro definitions for
4701 ALTIVEC_BUILTIN_VEC_PERMXOR.
4702 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4703 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4704 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4705 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4707 * config/doc/extend.texi: Add prototypes for vec_permxor.
4709 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4712 * diagnostic-show-locus.c (class layout_point): Convert m_line
4713 from int to linenum_type.
4714 (line_span::comparator): Use linenum "compare" function when
4715 comparing line numbers.
4716 (test_line_span): New function.
4717 (layout_range::contains_point): Convert param "row" from int to
4719 (layout_range::intersects_line_p): Likewise.
4720 (layout::will_show_line_p): Likewise.
4721 (layout::print_source_line): Likewise.
4722 (layout::should_print_annotation_line_p): Likewise.
4723 (layout::print_annotation_line): Likewise.
4724 (layout::print_leading_fixits): Likewise.
4725 (layout::annotation_line_showed_range_p): Likewise.
4726 (struct line_corrections): Likewise for field m_row.
4727 (line_corrections::line_corrections): Likewise for param "row".
4728 (layout::print_trailing_fixits): Likewise.
4729 (layout::get_state_at_point): Likewise.
4730 (layout::get_x_bound_for_row): Likewise.
4731 (layout::print_line): Likewise.
4732 (diagnostic_show_locus): Likewise for locals "last_line" and
4734 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4735 * input.c (selftest::test_linenum_comparisons): New function.
4736 (selftest::input_c_tests): Call it.
4737 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4738 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4739 * selftest.h (ASSERT_GT): New macro.
4740 (ASSERT_GT_AT): New macro.
4741 (ASSERT_LT): New macro.
4742 (ASSERT_LT_AT): New macro.
4744 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4746 PR rtl-optimization/84780
4747 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4749 2018-03-14 Martin Liska <mliska@suse.cz>
4751 * tree.c (record_node_allocation_statistics): Use
4752 get_stats_node_kind.
4753 (get_stats_node_kind): New function extracted from
4754 record_node_allocation_statistics.
4755 (free_node): Use get_stats_node_kind.
4757 2018-03-14 Richard Biener <rguenther@suse.de>
4759 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4760 that the value-set of ANTIC_IN doesn't grow.
4763 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4765 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4766 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4768 2018-03-14 Julia Koval <julia.koval@intel.com>
4770 * config.gcc (icelake-client, icelake-server): New.
4772 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4773 (initial_ix86_arch_features): Ditto.
4774 (PTA_SKYLAKE): Add SGX.
4775 (PTA_ICELAKE): Remove.
4776 (PTA_ICELAKE_CLIENT): New.
4777 (PTA_ICELAKE_SERVER): New.
4778 (ix86_option_override_internal): Split up icelake on icelake client and
4780 (get_builtin_code_for_version): Ditto.
4781 (fold_builtin_cpu): Ditto.
4782 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4783 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4784 * config/i386/i386.h (processor_type): Ditto.
4785 * doc/invoke.texi: Ditto.
4787 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4790 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4791 INTEGER_CST offset, add it together with bitpos / 8 and
4792 sign extend based on POINTER_SIZE.
4796 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4799 * config/i386/constraints.md (Yc): New register constraint.
4800 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4801 Use Yc constraint for alternative 2 of operand 0. Remove
4802 preferred_for_speed attribute.
4804 2018-03-14 Richard Biener <rguenther@suse.de>
4806 PR tree-optimization/84830
4807 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4808 with the old one to avoid oscillations.
4810 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4813 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4815 (assign_by_spills): Return a flag of reload assignment failure.
4816 Do not process the reload assignment failures. Do not spill other
4817 reload pseudos if they has the same reg class. Update n if
4819 (lra_assign): Add a return arg. Set up from the result of
4820 assign_by_spills call.
4821 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4822 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4823 usage_insns if it is not NULL.
4824 (spill_hard_reg_in_range): New function.
4825 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4826 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4827 function prototypes.
4828 (lra_assign): Change prototype.
4829 * lra.c (lra): Add code to deal with fails by splitting hard reg
4832 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4834 * config/riscv/riscv.opt (mrelax): New option.
4835 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4836 "norelax" when riscv_mrelax is disabled.
4837 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4839 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4842 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4843 reassociation for int modes.
4845 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4847 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4848 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4850 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4851 * config/aarch64/aarch64-sve.md
4852 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4853 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4854 (*extend<mode><Vwide>2): Rename to...
4855 (aarch64_sve_extend<mode><Vwide>2): ...this.
4856 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4857 renaming the old pattern to...
4858 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4860 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4861 define_expand, renaming the old pattern to...
4862 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4863 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4864 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4865 account when deciding which SVE instruction the optab should use.
4866 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4868 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4870 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4871 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4872 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4873 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4874 (tlsdesc_small_<mode>): Turn a define_expand and use
4875 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4876 (tlsdesc_small_advsimd_<mode>): ...this.
4877 (tlsdesc_small_sve_<mode>): New pattern.
4879 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4881 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4882 (UNSPEC_UMUL_HIGHPART): New constants.
4883 (MUL_HIGHPART): New int iteraor.
4884 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4885 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4887 (*<su>mul<mode>3_highpart): New define_insn.
4889 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4892 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4895 2018-03-13 Martin Liska <mliska@suse.cz>
4898 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4900 (sem_item_optimizer::~sem_item_optimizer): Release it.
4901 (sem_item_optimizer::merge_classes): Register variable aliases.
4902 (sem_item_optimizer::fixup_pt_set): New function.
4903 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4904 * ipa-icf.h: Declare new variables and functions.
4906 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4909 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4910 integer_pow2p@2 and test integer_pow2p in condition.
4911 (A < 0 ? C : 0): Similarly for @1.
4914 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4915 characters starting at p contain '\0' character, don't look beyond
4919 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4920 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4923 * reg-stack.c (change_stack): Change update_end var from int to
4924 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4925 also call set_block_for_insn on the newly added insns and rescan.
4928 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4929 on the last operand.
4932 * tree.c (stabilize_reference_1): Return save_expr (e) for
4933 STATEMENT_LIST even if it doesn't have side-effects.
4935 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4937 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4939 2018-03-12 Renlin Li <renlin.li@arm.com>
4941 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4942 aarch64_output_scalar_simd_mov_immediate.
4944 2018-03-12 Martin Sebor <msebor@redhat.com>
4946 PR tree-optimization/83456
4947 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4948 for perfectly overlapping calls to memcpy.
4949 (gimple_fold_builtin_memory_chk): Same.
4950 (gimple_fold_builtin_strcpy): Handle no-warning.
4951 (gimple_fold_builtin_stxcpy_chk): Same.
4952 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4954 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4956 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4957 parameter. Use it for SFmode.
4958 (rs6000_function_arg_advance_1): Adjust.
4959 (rs6000_function_arg): Adjust.
4960 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4962 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4964 PR rtl-optimization/84169
4965 PR rtl-optimization/84780
4966 * combine.c (can_combine_p): Check for a 2-insn combination whether
4967 the destination register is used between the two insns, too.
4969 2018-03-12 Richard Biener <rguenther@suse.de>
4971 PR tree-optimization/84803
4972 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4973 for refs DR analysis didn't process.
4975 2018-03-12 Richard Biener <rguenther@suse.de>
4977 PR tree-optimization/84777
4978 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4979 force-vectorize loops ignore whether we are optimizing for size.
4981 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4983 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4984 (TARGET_MD_ASM_ADJUST): Define.
4986 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4987 Kito Cheng <kito.cheng@gmail.com>
4988 Chung-Ju Wu <jasonwucj@gmail.com>
4990 * config/nds32/nds32.c (nds32_compute_stack_frame,
4991 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4992 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4993 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4994 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4995 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4996 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4997 * config/nds32/nds32.md (prologue, epilogue): Use macro
4998 NDS32_V3PUSH_AVAILABLE_P to do checking.
5000 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5003 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5004 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5005 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5006 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5007 addition of most attributes on !orig_type_die or the attribute not
5008 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5010 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5011 Chung-Ju Wu <jasonwucj@gmail.com>
5013 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5015 * config/nds32/nds32.opt (mvh): New option.
5017 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5018 Chung-Ju Wu <jasonwucj@gmail.com>
5020 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5022 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5023 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5026 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5027 Chung-Ju Wu <jasonwucj@gmail.com>
5029 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5031 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5032 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5034 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5035 Kito Cheng <kito.cheng@gmail.com>
5036 Chung-Ju Wu <jasonwucj@gmail.com>
5038 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5039 UNSPEC_FFMISM and UNSPEC_FLMISM.
5040 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5041 for ffb, ffmism and flmism.
5042 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5043 (unspec_ffmism): Ditto.
5044 (unspec_flmism): Ditto.
5045 (nds32_expand_builtin_impl): Check if string extension is available.
5046 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5047 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5049 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5052 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5055 * lra-assigns.c (assign_by_spills): Return a flag of reload
5056 assignment failure. Do not process the reload assignment
5057 failures. Do not spill other reload pseudos if they has the same
5059 (lra_assign): Add a return arg. Set up from the result of
5060 assign_by_spills call.
5061 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5062 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5063 usage_insns if it is not NULL.
5064 (spill_hard_reg_in_range): New function.
5065 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5066 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5067 function prototypes.
5068 (lra_assign): Change prototype.
5069 * lra.c (lra): Add code to deal with fails by splitting hard reg
5072 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5075 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5077 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5080 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5081 (dw_val_node): Add val_symbolic_view.
5082 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5083 (symview_upper_bound): New.
5084 (new_line_info_table): Initialize symviews_since_reset.
5085 (dwarf2out_source_line): Count symviews_since_reset and set
5086 symview_upper_bound.
5087 (dw_val_equal_p): Handle symview.
5088 (add_AT_symview): New.
5089 (print_dw_val): Handle symview.
5090 (attr_checksum, attr_checksum_ordered): Likewise.
5091 (same_dw_val_p, size_of_die): Likewise.
5092 (value_format, output_die): Likewise.
5093 (add_high_low_attributes): Use add_AT_symview for entry_view.
5094 (dwarf2out_finish): Reset symview_upper_bound, clear
5097 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5100 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5101 Add strict argument and use it.
5102 (rs6000_split_multireg_move): Update for new strict argument.
5103 (mem_operand_gpr): Disallow all non-offsettable addresses.
5104 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5106 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5109 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5110 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5111 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5114 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5115 decl, use remap_type if we want to use the type.
5117 2018-03-09 Martin Sebor <msebor@redhat.com>
5119 PR tree-optimization/84526
5120 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5122 (builtin_access::generic_overlap): Be prepared to handle non-array
5125 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5127 PR rtl-optimization/84682
5128 * lra-constraints.c (process_address_1): Check is_address flag
5129 for address constraints.
5130 (process_alt_operands): Likewise.
5131 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5132 preprocess_constraints.
5133 * recog.h (preprocess_constraints): Add oploc parameter.
5135 * recog.c (preprocess_constraints): Test address_operand for
5136 CT_ADDRESS constraints.
5138 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5141 * lra-assigns.c (assign_by_spills): Return a flag of reload
5142 assignment failure. Do not process the reload assignment
5143 failures. Do not spill other reload pseudos if they has the same
5145 (lra_assign): Add a return arg. Set up from the result of
5146 assign_by_spills call.
5147 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5148 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5149 usage_insns if it is not NULL.
5150 (spill_hard_reg_in_range): New function.
5151 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5152 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5153 function prototypes.
5154 (lra_assign): Change prototype.
5155 * lra.c (lra): Add code to deal with fails by splitting hard reg
5158 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5161 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5162 Accept complain bool parameter. Only emit errors if it is true.
5163 (arm_parse_cpu_option_name): Likewise.
5164 (arm_target_thumb_only): Adjust callers of the above.
5165 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5166 prototype to take a default true bool parameter.
5167 (arm_parse_arch_option_name): Likewise.
5169 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5170 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5174 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5175 * configure: Regenerate.
5176 * configure.ac ("linker --version-script option"): New.
5177 ("linker soname option"): New.
5179 2018-03-09 Richard Biener <rguenther@suse.de>
5181 PR tree-optimization/84775
5182 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5183 immediate uses of predicate stmts and mark them modified.
5186 PR tree-optimization/84178
5187 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5189 (version_loop_for_if_conversion): Delay update_ssa call.
5190 (tree_if_conversion): Delay update_ssa until after predicate
5193 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5196 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5197 when the function accesses prior frames.
5199 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5202 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5203 gen_llsym, otherwise call maybe_gen_llsym.
5206 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5207 has ',' character inside of it.
5209 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5212 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5213 as clobbering CC_REGNUM.
5215 2018-03-08 Richard Biener <rguenther@suse.de>
5218 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5219 (follow_copies_to_constant): Do not follow SSA names registered
5222 2018-03-08 Richard Biener <rguenther@suse.de>
5224 PR tree-optimization/84178
5225 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5227 (version_loop_for_if_conversion): Delay update_ssa call.
5228 (tree_if_conversion): Delay update_ssa until after predicate
5231 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5233 PR tree-optimization/84178
5234 * tree-if-conv.c (release_bb_predicate): Remove the
5235 the assertion that the stmts have NULL use_ops.
5236 Discard the statements, asserting that they haven't
5237 yet been added to a BB.
5239 2018-03-08 Richard Biener <rguenther@suse.de>
5241 PR tree-optimization/84746
5242 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5243 (phi_translate): Pass in destination ANTIC_OUT set.
5244 (phi_translate_1): Likewise. For a simplified result lookup
5245 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5246 (phi_translate_set): Adjust.
5247 (do_pre_regular_insertion): Likewise.
5248 (do_pre_partial_partial_insertion): Likewise.
5250 2018-03-08 Martin Liska <mliska@suse.cz>
5252 PR gcov-profile/84735
5253 * doc/gcov.texi: Document usage of profile files.
5254 * gcov-io.h: Document changes in the format.
5256 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5260 * dwarf2out.c (struct dw_line_info_table): Update comments for
5262 (FORCE_RESET_NEXT_VIEW): New.
5263 (FORCE_RESETTING_VIEW_P): New.
5264 (RESETTING_VIEW_P): Check for -1 too.
5265 (ZERO_VIEW_P): Likewise.
5266 (new_line_info_table): Force-reset next view.
5267 (dwarf2out_begin_function): Likewise.
5268 (dwarf2out_source_line): Simplify zero_view_p initialization.
5269 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5270 view directly. Omit view when omitting .loc at line 0.
5272 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5274 PR tree-optimization/84740
5275 * tree-switch-conversion.c (process_switch): Call build_constructors
5276 only if info.phi_count is non-zero.
5278 PR tree-optimization/84739
5279 * tree-tailcall.c (find_tail_calls): Check call arguments against
5280 DECL_ARGUMENTS (current_function_decl) rather than
5281 DECL_ARGUMENTS (func) when checking for tail recursion.
5283 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5285 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5286 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5287 Volker Reichelt's entry and add entries for people that perform
5288 GCC fuzzy testing and report numerous bugs.
5290 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5293 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5294 readonly data in sdata, if that is disabled.
5295 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5296 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5297 -mreadonly-in-sdata option.
5299 2018-03-07 Martin Sebor <msebor@redhat.com>
5301 PR tree-optimization/84468
5302 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5303 basic block when looking for nul assignment.
5305 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5308 * except.h (output_function_exception_table): Adjust prototype.
5309 * except.c (output_function_exception_table): Remove FNNAME parameter
5310 and add SECTION parameter. Ouput one part of the table at a time.
5311 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5312 the first part of the exception table and emit unwind directives.
5313 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5314 (i386_pe_seh_cold_init): Likewise.
5315 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5316 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5317 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5318 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5319 * config/i386/winnt.c: Include except.h.
5320 (struct seh_frame_state): Add reg_offset, after_prologue and
5321 in_cold_section fields.
5322 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5323 (i386_pe_seh_cold_init): New function.
5324 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5325 to seh->in_cold_section.
5326 (seh_emit_push): Record the offset of the push.
5327 (seh_emit_save): Record the offet of the save.
5328 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5329 Test seh->after_prologue to disregard the epilogue.
5330 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5331 (i386_pe_end_cold_function): New function.
5333 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5336 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5337 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5340 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5342 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5343 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5346 * multiple_target.c: Include tree-inline.h and intl.h.
5347 (expand_target_clones): Diagnose and fail if node->definition and
5348 !tree_versionable_function_p (node->decl).
5350 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5352 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5354 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5355 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5356 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5358 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5361 * combine.c (try_combine): Use reg_or_subregno instead of handling
5362 just paradoxical SUBREGs and REGs.
5364 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5366 * config/arc/arc.c (arc_finalize_pic): Remove function.
5367 (arc_must_save_register): We use single base PIC register, remove
5368 checks to save/restore the PIC register.
5369 (arc_expand_prologue): Likewise.
5370 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5372 (arc_verify_short): Likewise.
5373 (arc_attr_type): Likewise.
5374 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5375 (walk_stores): Likewise.
5376 (arc_address_cost): Make it static.
5377 (arc_verify_short): Likewise.
5378 (branch_dest): Likewise.
5379 (arc_attr_type): Likewise.
5380 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5381 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5382 (arc_final_prescan_insn): Remove inserting the nops due to
5383 hardware hazards. It is done in reorg step.
5384 (insn_length_variant_t): Remove.
5385 (insn_length_parameters_t): Likewise.
5386 (arc_insn_length_parameters): Likewise.
5387 (arc_get_insn_variants): Likewise.
5388 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5390 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5393 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5396 PR tree-optimization/84687
5397 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5399 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5401 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5403 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5404 Rename to ppc_speculation_barrier.
5405 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5406 __builtin_ppc_speculation_barrier.
5408 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5411 * combine.c (combine_simplify_rtx): Don't try to simplify if
5412 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5415 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5417 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5418 to 32 bytes when compiling for POWER9.
5420 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5423 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5424 regparm >= 3 with no arg reg available also for calls with
5425 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5428 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5430 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5432 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5435 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5437 2018-03-05 Richard Biener <rguenther@suse.de>
5439 PR tree-optimization/84486
5440 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5441 When inserting a __builtin_assume_aligned call set the LHS
5442 SSA name alignment info accordingly.
5444 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5446 PR tree-optimization/84114
5447 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5448 Avoid reassociation of FLOAT_MODE addition.
5450 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5452 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5453 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5454 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5455 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5456 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5457 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5458 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5460 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5461 __builtin_ia32_wbinvd): New builtins.
5462 (SPECIAL_ARGS2): New.
5463 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5464 (SPECIAL_ARGS2): New.
5465 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5466 (ix86_valid_target_attribute_inner_p): Ditto.
5467 (ix86_init_mmx_sse_builtins): Add special_args2.
5468 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5469 TARGET_WBNOINVD_P): New.
5470 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5471 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5472 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5473 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5474 * config/i386/pconfigintrin.h: New file.
5475 * config/i386/wbnoinvdintrin.h: Ditto.
5476 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5477 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5479 2018-03-05 Richard Biener <rguenther@suse.de>
5481 PR tree-optimization/84670
5482 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5484 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5485 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5486 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5487 doesn't grow if all successors (recursively) were visited at least
5490 2018-03-05 Richard Biener <rguenther@suse.de>
5492 PR tree-optimization/84650
5493 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5494 if executed in the loop pipeline.
5496 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5498 * doc/configfiles.texi (Configuration Files): Move info about
5499 conditionalizing $target-protos.h to...
5500 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5501 differs from $target-protos.h.
5503 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5504 Chung-Ju Wu <jasonwucj@gmail.com>
5506 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5507 * config/nds32/nds32-multiple.md (setmemsi): Define.
5508 * config/nds32/nds32-memory-manipulation.c
5509 (nds32_gen_dup_4_byte_to_word_value): New.
5510 (emit_setmem_word_loop): New.
5511 (emit_setmem_byte_loop): New.
5512 (nds32_expand_setmem_loop): New.
5513 (nds32_expand_setmem_loop_v3m): New.
5514 (nds32_expand_setmem_unroll): New.
5515 (nds32_expand_setmem): New.
5517 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5518 Chung-Ju Wu <jasonwucj@gmail.com>
5520 * config/nds32/nds32-memory-manipulation.c
5521 (nds32_emit_load_store): New.
5522 (nds32_emit_post_inc_load_store): New.
5523 (nds32_emit_mem_move): New.
5524 (nds32_emit_mem_move_block): New.
5525 (nds32_expand_movmemsi_loop_unknown_size): New.
5526 (nds32_expand_movmemsi_loop_known_size): New.
5527 (nds32_expand_movmemsi_loop): New.
5528 (nds32_expand_movmemsi_unroll): New.
5529 (nds32_expand_movmemqi): Rename ...
5530 (nds32_expand_movmemsi): ... to this.
5531 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5532 (movmemsi): ... to this.
5533 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5534 (nds32_expand_movmemsi): ... to this.
5536 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5537 Monk Chiang <sh.chiang04@gmail.com>
5538 Chung-Ju Wu <jasonwucj@gmail.com>
5540 * config/nds32/nds32-protos.h
5541 (nds32_expand_load_multiple): New arguments.
5542 (nds32_expand_store_multiple): Ditto.
5543 (nds32_valid_multiple_load_store): Rename ...
5544 (nds32_valid_multiple_load_store_p): ... to this.
5545 * config/nds32/nds32-memory-manipulation.c
5546 (nds32_expand_load_multiple): Refine implementation.
5547 (nds32_expand_store_multiple): Ditto.
5548 * config/nds32/nds32-multiple.md
5549 (load_multiple): Update nds32_expand_load_multiple interface.
5550 (store_multiple): Update nds32_expand_store_multiple interface.
5551 * config/nds32/nds32-predicates.c
5552 (nds32_valid_multiple_load_store): Rename ...
5553 (nds32_valid_multiple_load_store_p): ... to this and refine
5555 * config/nds32/predicates.md
5556 (nds32_load_multiple_and_update_address_operation): New predicate.
5557 (nds32_store_multiple_and_update_address_operation): New predicate.
5559 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5560 Chung-Ju Wu <jasonwucj@gmail.com>
5562 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5563 (combo): New attribute.
5564 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5566 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5568 * config/nds32/nds32.opt: Change -mcmodel= default value.
5570 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5571 Monk Chiang <sh.chiang04@gmail.com>
5572 Chung-Ju Wu <jasonwucj@gmail.com>
5574 * config/nds32/constants.md (unspec_element): New enum.
5575 * config/nds32/constraints.md (Umw): New constraint.
5576 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5577 * config/nds32/nds32-intrinsic.md: Likewise.
5578 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5579 (nds32_valid_smw_lwm_base_p): New.
5580 (nds32_output_smw_single_word): New.
5581 (nds32_output_lmw_single_word): New.
5582 (nds32_expand_unaligned_load): New.
5583 (nds32_expand_unaligned_store): New.
5584 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5585 (nds32_output_smw_single_word): Declare.
5586 (nds32_output_lmw_single_word): Declare.
5587 (nds32_expand_unaligned_load): Declare.
5588 (nds32_expand_unaligned_store): Declare.
5589 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5590 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5591 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5592 NDS32_BUILTIN_UASTORE_DW.
5593 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5596 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5597 Kito Cheng <kito.cheng@gmail.com>
5598 Chung-Ju Wu <jasonwucj@gmail.com>
5600 * config/nds32/nds32-intrinsic.c
5601 (nds32_expand_builtin_null_ftype_reg): Delete.
5602 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5603 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5604 (nds32_read_argument): New.
5605 (nds32_legitimize_target): Ditto.
5606 (nds32_legitimize_argument): Ditto.
5607 (nds32_check_constant_argument): Ditto.
5608 (nds32_expand_unop_builtin): Ditto.
5609 (nds32_expand_unopimm_builtin): Ditto.
5610 (nds32_expand_binop_builtin): Ditto.
5611 (nds32_builtin_decl_impl): Ditto.
5612 (builtin_description): Ditto.
5613 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5614 (nds32_init_builtins_impl): Ditto.
5615 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5616 (nds32_builtin_decl): New.
5617 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5618 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5620 2018-03-02 Jeff Law <law@redhat.com>
5622 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5623 (redundant_insn, fill_simple_delay_slots): Likewise.
5624 (fill_slots_from_thread): Likewise.
5625 * resource.c (mark_referenced_resources): Likewise.
5626 (mark_set_resources, find_dead_or_set_registers): Likewise.
5628 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5630 * substring-locations.h (format_warning_va): Formatting fix for
5632 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5634 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5635 * substring-locations.c: Include intl.h.
5636 (format_warning_va): Turned into small wrapper around
5637 format_warning_n_va, renamed to ...
5638 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5639 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5641 (format_warning_at_substring_n): New function.
5642 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5643 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5644 format_warning_at_substring with just a shorter name instead of
5645 const function pointer.
5646 (fmtwarn_n): New function.
5647 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5648 appropriate, get rid of all the fmtstr temporaries, move conditionals
5649 with G_() wrapped string literals directly into fmtwarn arguments,
5650 cast dir.len to (int), formatting fixes.
5652 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5654 * doc/invoke.texi: Remove "Cilk Plus" references.
5656 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5657 Richard Biener <rguenther@suse.de>
5660 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5661 for error or warning attributes if CALL_FROM_THUNK_P is set.
5664 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5667 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5668 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5671 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5672 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5674 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5675 instead of -1U in last predictors element's probability member.
5677 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5680 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5681 arguments if they are comparable.
5683 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5685 PR tree-optimization/84634
5686 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5687 masks and masked_loop_p with a single loop_masks, making sure it's
5688 null for bb vectorization.
5690 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5692 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5693 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5694 loop->force_vectorize to check whether there is no alias.
5696 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5699 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5701 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5703 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5704 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5705 instead of a loop around prev_real_insn.
5706 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5710 * config/i386/i386.c (ix86_print_operand): Use conditional
5711 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5714 2018-03-02 Richard Biener <rguenther@suse.de>
5716 PR tree-optimization/84427
5717 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5718 (bitmap_set_subtract_values): Rewrite to handle multiple
5721 (prune_clobbered_mems): Likewise.
5722 (phi_translate): Take edge instead of pred/phiblock.
5723 (phi_translate_1): Likewise.
5724 (phi_translate_set): Likewise. Insert all translated
5725 exprs for a value into the set, keeping possibly multiple
5726 expressions per value.
5727 (compute_antic_aux): Adjust for phi_translate changes.
5728 When intersecting union the expressions and prune those
5729 not in the final value set, keeping possibly multiple
5730 expressions per value. Do not use value-insertion
5731 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5732 all expressions. Add verification that the value-sets
5733 only shrink during iteration.
5734 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5735 (do_pre_regular_insertion): Likewise.
5736 (do_pre_partial_partial_insertion): Likewise.
5738 2018-03-02 Richard Biener <rguenther@suse.de>
5741 * config/darwin.c (saved_debug_info_level): New static global.
5742 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5743 (darwin_asm_lto_end): Restore debug info generation settings.
5745 2018-03-01 Martin Liska <mliska@suse.cz>
5748 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5751 2018-03-01 Richard Biener <rguenther@suse.de>
5754 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5755 type annotation with LTO when debug was disabled at compile-time.
5757 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5759 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5761 (mips_final_postscan_insn): Likewise.
5763 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5765 PR rtl-optimization/84528
5766 * alias.c (init_alias_target): Add commentary.
5767 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5768 a unique base value if the frame pointer is not eliminated
5769 to the stack pointer.
5771 2018-03-01 Tom de Vries <tom@codesourcery.com>
5773 PR rtl-optimization/83327
5774 * lra-int.h (hard_regs_spilled_into): Declare.
5775 * lra.c (hard_regs_spilled_into): Define.
5776 (init_reg_info): Init hard_regs_spilled_into.
5777 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5778 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5779 (process_bb_lives): Handle hard_regs_spilled_into.
5780 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5781 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5783 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5785 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5786 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5787 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5788 * config/rs6000/aix72.h: New file.
5790 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5792 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5793 instead of warning_at with conditional singular and plural messages
5797 * stor-layout.c (update_alignment_for_field): For
5798 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5799 && !DECL_PACKED (field), do the alignment update, just use
5800 only desired_align instead of MAX (type_align, desired_align)
5802 (place_field): Don't do known_align < desired_align handling
5803 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5804 is non-NULL, instead do it after rli->prev_field handling and
5805 only if not within a bitfield word. For DECL_PACKED (field)
5806 use type_align of BITS_PER_UNIT.
5808 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5810 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5811 superfluous parentheses and trailing spaces.
5813 2018-02-28 Richard Biener <rguenther@suse.de>
5815 PR tree-optimization/84584
5816 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5817 SCoPs with fake exit edge.
5819 2018-02-28 Martin Liska <mliska@suse.cz>
5822 * timevar.c (timer::print): Fix format to properly print 100%
5825 2018-02-28 Richard Biener <rguenther@suse.de>
5828 * genmatch.c (capture_info::walk_match): Do not mark
5829 captured expressions without operands as expr_p given
5830 they act more like predicates and should be subject to
5831 "lost tail" side-effect preserving.
5833 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5835 PR rtl-optimization/81611
5836 * auto-inc-dec.c (attempt_change): Move dead note from
5837 mem_insn if it's the next use of regno
5838 (find_address): Take address use of reg holding
5839 non-incremented value. Add parm to limit search to the named
5841 (merge_in_block): Attempt to use a mem insn that is the next
5842 use of the original regno.
5844 2018-02-27 Martin Sebor <msebor@redhat.com>
5847 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5848 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5850 2018-02-27 Martin Sebor <msebor@redhat.com>
5852 PR translation/84207
5853 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5854 n argument to unsigned HOST_WIDE_INT.
5855 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5856 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5857 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5858 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5860 2018-02-27 Richard Biener <rguenther@suse.de>
5862 PR tree-optimization/84512
5863 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5864 Do not use the estimate returned from record_stmt_cost for
5865 the scalar iteration cost but sum properly using add_stmt_cost.
5867 2018-02-27 Richard Biener <rguenther@suse.de>
5869 PR tree-optimization/84466
5870 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5871 Adjust last change to less strictly validate use operands.
5873 2018-02-27 Martin Liska <mliska@suse.cz>
5875 PR gcov-profile/84548
5876 * gcov.c (process_file): Allow partial overlap and consider it
5877 also as group functions.
5878 (output_lines): Properly calculate range of lines for a group.
5880 2018-02-27 Martin Liska <mliska@suse.cz>
5882 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5883 'ggc' suffixes. Change first column width.
5884 (timer::print): Fix formatting of the column.
5886 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5888 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5889 preserve inline entry blocks for the sake of debug inline
5890 entry point markers alone.
5891 (remove_unused_locals): Suggest in comments a better place to
5892 force the preservation of inline entry blocks that are
5893 otherwise unused, but do not preserve them.
5895 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5897 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5899 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5902 * config/i386/constraints.md (Bs): Replace
5903 ix86_indirect_branch_register with
5904 TARGET_INDIRECT_BRANCH_REGISTER.
5906 * config/i386/i386.md (indirect_jump): Likewise.
5907 (tablejump): Likewise.
5908 (*sibcall_memory): Likewise.
5909 (*sibcall_value_memory): Likewise.
5910 Peepholes of indirect call and jump via memory: Likewise.
5911 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5912 (*sibcall_value_GOT_32): Likewise.
5913 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5914 (GOT_memory_operand): Likewise.
5915 (call_insn_operand): Likewise.
5916 (sibcall_insn_operand): Likewise.
5917 (GOT32_symbol_operand): Likewise.
5918 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5920 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5922 PR rtl-optimization/83496
5923 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5924 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5925 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5926 redundant insn, if any.
5927 (relax_delay_slots): Likewise.
5928 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5930 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5932 PR tree-optimization/83965
5933 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5934 that grouped statements are part of a reduction chain. Return
5935 true if the statement is not marked as a reduction itself but
5937 (vect_recog_dot_prod_pattern): Don't check whether the statement
5938 is part of a group here.
5939 (vect_recog_sad_pattern): Likewise.
5940 (vect_recog_widen_sum_pattern): Likewise.
5942 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5945 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5947 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5950 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5952 (ix86_output_indirect_function_return): New prototype.
5953 (ix86_split_simple_return_pop_internal): Likewise.
5954 * config/i386/i386.c (indirect_return_via_cx): New.
5955 (indirect_return_via_cx_bnd): Likewise.
5956 (indirect_thunk_name): Handle return va CX_REG.
5957 (output_indirect_thunk_function): Create alias for
5958 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5959 (ix86_output_indirect_jmp): Remove the bool argument.
5960 (ix86_output_indirect_function_return): New function.
5961 (ix86_split_simple_return_pop_internal): Likewise.
5962 * config/i386/i386.md (*indirect_jump): Don't pass false
5963 to ix86_output_indirect_jmp.
5964 (*tablejump_1): Likewise.
5965 (simple_return_pop_internal): Change it to define_insn_and_split.
5966 Call ix86_split_simple_return_pop_internal to split it for
5968 (simple_return_indirect_internal): Call
5969 ix86_output_indirect_function_return instead of
5970 ix86_output_indirect_jmp.
5972 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5975 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5976 memset and value initialization afterwards.
5978 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5980 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5982 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5985 * common/config/aarch64/aarch64-common.c
5986 (aarch_option_optimization_table[]): Switch
5987 off fomit-frame-pointer
5989 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5990 Chung-Ju Wu <jasonwucj@gmail.com>
5992 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5994 (store_multiple): Ditto.
5996 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5998 * config.gcc: Add --with-cpu support for nds32 target.
5999 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6000 * config/nds32/nds32.opt: Add -mcpu= option.
6002 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6004 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6005 isel=yes): Warn for these deprecated options.
6007 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6009 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6010 ISA_2_5_MASKS_EMBEDDED.
6012 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6014 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6015 p->max as pointers rather than using iterative_hash_expr.
6017 2018-02-23 Carl Love <cel@us.ibm.com>
6019 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6020 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6022 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6023 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6024 P8V_BUILTIN_VEC_VUNSIGNED2.
6026 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6029 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6030 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6032 * lra-constraints.c (curr_insn_transform): Set up
6033 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6035 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6036 * lra-eliminations.c (spill_pseudos): Ditto.
6037 (process_insn_for_elimination): Ditto.
6038 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6039 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6040 LRA_NON_CLOBBERED_ALT.
6042 2018-02-22 Martin Sebor <msebor@redhat.com>
6044 PR tree-optimization/84480
6045 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6046 to maybe_diag_stxncpy_trunc. Call it.
6047 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6048 from gimple_fold_builtin_strcpy. Print inlining stack.
6049 (handle_builtin_stxncpy): Print inlining stack.
6050 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6052 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6055 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6056 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6057 and -fcheck-pointer-bounds are used together.
6058 (indirect_thunk_prefix): New enum.
6059 (indirect_thunk_need_prefix): New function.
6060 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6061 "_nt" instead of "_bnd" for NOTRACK prefix.
6062 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6063 (output_indirect_thunk_function): Likewise.
6065 (ix86_code_end): Update output_indirect_thunk_function calls.
6066 (ix86_output_indirect_branch_via_reg): Replace
6067 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6068 (ix86_output_indirect_branch_via_push): Likewise.
6069 (ix86_output_function_return): Likewise.
6070 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6071 incompatible with -fcf-protection=branch and
6072 -fcheck-pointer-bounds.
6074 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6077 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6078 Change gcc_assert call to output_operand_lossage.
6080 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6082 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6084 2018-02-22 DJ Delorie <dj@redhat.com>
6085 Sebastian Perta <sebastian.perta@renesas.com>
6086 Oleg Endo <olegendo@gcc.gnu.org>
6088 * config/rx/rx.c (rx_rtx_costs): New function.
6089 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6091 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6093 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6095 2018-02-22 Martin Liska <mliska@suse.cz>
6098 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6099 Add "native" as a possible value.
6101 2018-02-22 Martin Liska <mliska@suse.cz>
6104 * config/i386/i386.c (ix86_option_override_internal):
6105 Add "native" as a possible value for -march and -mtune.
6107 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6110 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6111 to all type variants.
6113 PR tree-optimization/84503
6114 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6115 width as info->bitpos + info->bitsize - start.
6116 (merged_store_group::merge_overlapping): Simplify width computation.
6117 (check_no_overlap): New function.
6118 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6119 start + width and last_order of the group, fail if check_no_overlap
6121 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6122 to group if check_no_overlap fails.
6124 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6126 * config/rs6000/altivec.md: Delete contraint arguments to
6127 define_expand, define_split, and define_peephole2, and in
6128 define_insn_and_split if always unused.
6129 * config/rs6000/darwin.md: Ditto.
6130 * config/rs6000/dfp.md: Ditto.
6131 * config/rs6000/rs6000.md: Ditto.
6132 * config/rs6000/sync.md: Ditto.
6133 * config/rs6000/vector.md: Ditto.
6134 * config/rs6000/vsx.md: Ditto.
6136 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6138 * config/rs6000/altivec.md: Write output control strings as braced
6139 blocks instead of double-quoted strings.
6140 * config/rs6000/darwin.md: Ditto.
6141 * config/rs6000/rs6000.md: Ditto.
6142 * config/rs6000/vector.md: Ditto.
6143 * config/rs6000/vsx.md: Ditto.
6145 2018-02-21 Jason Merrill <jason@redhat.com>
6147 PR c++/84314 - ICE with templates and fastcall attribute.
6148 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6150 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6152 * ipa-cp.c (determine_versionability): Fix comment typos.
6154 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6157 * ipa-cp.c (determine_versionability): Do not version functions caling
6160 2018-02-21 Martin Liska <mliska@suse.cz>
6163 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6164 Add "native" as a possible value.
6165 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6166 the macro when native cpu detection is available.
6168 2018-02-21 Martin Liska <mliska@suse.cz>
6171 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6172 Add "native" as a possible value.
6173 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6174 when native cpu detection is available.
6176 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6177 Martin Sebor <msebor@redhat.com>
6179 PR tree-optimization/84478
6180 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6182 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6183 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6184 support which is conservatively correct, for 2 only stay conservative
6185 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6186 argument to the 2 argument get_range_strlen, adjust 6 arg
6187 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6189 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6190 (gimple_fold_builtin_strlen): Pass true as last argument to
6193 2018-02-20 Martin Sebor <msebor@redhat.com>
6196 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6197 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6198 (builtin_memref::builtin_memref): Factor out parts into
6199 set_base_and_offset and call it.
6201 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6204 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6205 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6206 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6207 search at the associated MODE_INT.
6209 2018-02-20 Jeff Law <law@redhat.com>
6212 PR tree-optimization/81592
6214 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6215 for range data rather than using global data.
6216 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6217 range data rather than using global data.
6218 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6219 pass it to children as needed.
6220 (struct directive::fmtresult): Similarly.
6221 (struct directive::set_width): Similarly.
6222 (struct directive::set_precision): Similarly.
6223 (format_integer, format_directive, parse_directive): Similarly.
6224 (format_none): Accept unnamed vr_values parameter.
6225 (format_percent, format_floating, format_character): Similarly.
6226 (format_string, format_plain): Similarly.
6227 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6228 the EVRP range analyzer for range data rather than using global data.
6229 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6230 gimple-ssa-evrp-analyze.h
6231 (class sprintf_dom_walker): Add after_dom_children member function.
6232 Add evrp_range_analyzer member.
6233 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6234 range analyzer as needed.
6235 (sprintf_dom_walker::after_dom_children): New member function.
6236 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6238 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6239 (evrp_range_analyzer::pop_to_marker): Likewise.
6241 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6243 PR tree-optimization/84419
6244 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6245 with the required type if its current type is compatible but
6248 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6251 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6252 after vectorization.
6254 2018-02-20 Martin Liska <mliska@suse.cz>
6257 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6258 possible values if we don't have a hint.
6260 2018-02-20 Martin Liska <mliska@suse.cz>
6264 * final.c (shorten_branches): Build align_tab array with one
6266 * opts.c (finish_options): Add alignment option limit check.
6267 (MAX_CODE_ALIGN): Likewise.
6268 (MAX_CODE_ALIGN_VALUE): Likewise.
6269 * doc/invoke.texi: Document maximum allowed option value for
6270 all -falign-* options.
6272 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6275 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6276 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6277 * var-tracking.c (emit_note_insn_var_location): Remove all references
6278 to NOTE_INSN_CALL_ARG_LOCATION.
6279 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6280 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6281 Use copy_rtx_if_shared.
6282 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6283 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6284 (dwarf2out_var_location): Remove handling of
6285 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6287 * final.c (final_scan_insn): Remove all references to
6288 NOTE_INSN_CALL_ARG_LOCATION.
6289 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6290 before dumping final insns.
6291 * except.c (emit_note_eh_region_end): Remove all references to
6292 NOTE_INSN_CALL_ARG_LOCATION.
6293 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6294 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6295 * config/arc/arc.c (hwloop_optimize): Likewise.
6296 * config/arm/arm.c (create_fix_barrier): Likewise.
6297 * config/s390/s390.c (s390_chunkify_start): Likewise.
6298 * config/sh/sh.c (find_barrier): Likewise.
6299 * config/i386/i386.c (rest_of_insert_endbranch,
6300 ix86_seh_fixup_eh_fallthru): Likewise.
6301 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6302 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6303 * config/frv/frv.c (frv_function_prologue): Likewise.
6304 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6306 (note_outside_basic_block_p): Remove all references to
6307 NOTE_INSN_CALL_ARG_LOCATION.
6308 * gengtype.c (adjust_field_rtx_def): Likewise.
6309 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6311 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6312 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6315 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6318 PR tree-optimization/84452
6319 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6320 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6323 2018-02-19 Martin Liska <mliska@suse.cz>
6326 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6328 2018-02-19 Martin Liska <mliska@suse.cz>
6329 Richard Sandiford <richard.sandiford@linaro.org>
6331 PR tree-optimization/82491
6332 * gimple-fold.c (get_base_constructor): Make earlier bail out
6335 2018-02-19 Carl Love <cel@us.ibm.com>
6337 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6338 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6340 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6341 P8V_BUILTIN_VEC_NEG.
6343 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6345 * config/rl78/rl78.md (movdf): New define expand.
6347 2018-02-19 Martin Liska <mliska@suse.cz>
6350 * doc/invoke.texi: Fix typo.
6351 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6353 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6355 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6356 handle rs6000_single_float and rs6000_double_float specially for
6359 2018-02-16 Jeff Law <law@redhat.com>
6361 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6364 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6367 * ipa-inline.c (inline_small_functions): Fix a typo.
6369 2018-02-16 Nathan Sidwell <nathan@acm.org>
6371 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6373 2018-02-16 Carl Love <cel@us.ibm.com>
6375 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6376 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6377 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6378 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6379 expansion to P8V_BUILTIN_VEC_FLOAT2.
6381 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6383 PR rtl-optimization/70023
6384 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6385 src_regno into account.
6387 2018-02-16 Carl Love <cel@us.ibm.com>
6389 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6390 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6391 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6392 * config/rs6000/rs6000.c: Remove case statements for
6393 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6394 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6395 and P9V_BUILTIN_VEC_VINSERT4B.
6396 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6397 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6398 * config/rs6000/vsx.md:
6399 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6402 2018-02-16 Carl Love <cel@us.ibm.com>
6404 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6406 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6408 * config/rs6000/rs6000-c.c: Add the definitions for
6409 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6410 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6411 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6412 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6413 definition for insert4b and define insn *insert3b_internal.
6414 * doc/extend.texi: Add documentation for vec_extract4b.
6416 2018-02-16 Nathan Sidwell <nathan@acm.org>
6418 * doc/extend.texi (Backwards Compatibility): Mention friend
6419 injection. Note for-scope is deprecated.
6420 * doc/invoke.texi (-ffriend-injection): Deprecate.
6422 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6424 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6425 that moved to I2, also allow destinations that are a paradoxical
6426 subreg (instead of a normal reg).
6428 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6431 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6434 2018-02-16 Richard Biener <rguenther@suse.de>
6436 PR tree-optimization/84037
6437 PR tree-optimization/84016
6439 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6440 Adjust vec_construct for the fact we need additional higher latency
6441 128bit inserts for AVX256 and AVX512 vector builds.
6442 (ix86_add_stmt_cost): Scale vector construction cost for
6445 2018-02-16 Richard Biener <rguenther@suse.de>
6447 PR tree-optimization/84417
6448 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6449 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6450 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6452 2018-02-16 Martin Liska <mliska@suse.cz>
6455 * internal-fn.def (ASAN_CHECK): Set proper flags.
6456 (ASAN_MARK): Likewise.
6458 2018-02-16 Julia Koval <julia.koval@intel.com>
6460 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6461 from PTA_CANNONLAKE.
6463 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6466 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6467 Use ++iter rather than iter++ for std::list iterators.
6468 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6469 defer deleting them until all nodes in the forest are processed. Do
6470 free even leaf nodes. Change to_process into auto_vec.
6473 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6474 * vec.h (vec_default_construct): Use memset instead of placement new
6475 if BROKEN_VALUE_INITIALIZATION is defined.
6476 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6477 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6480 PR rtl-optimization/83723
6481 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6482 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6483 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6485 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6487 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6489 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6491 PR rtl-optimization/81443
6492 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6493 from inner REGs to paradoxical SUBREGs.
6495 2018-02-16 Richard Biener <rguenther@suse.de>
6497 PR tree-optimization/84399
6498 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6499 For operands we can analyze at their definition make sure we can
6500 analyze them at each use as well.
6502 2018-02-16 Richard Biener <rguenther@suse.de>
6504 PR tree-optimization/84190
6505 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6506 volatile accesses if the decl isn't volatile.
6508 2018-02-15 Jason Merrill <jason@redhat.com>
6510 PR c++/84314 - ICE with templates and fastcall attribute.
6511 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6512 TYPE_CANONICAL on an existing type.
6514 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6516 PR tree-optimization/84383
6517 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6518 dstoff nor call operand_equal_p if dstbase is NULL.
6520 PR tree-optimization/84334
6521 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6522 also a CONSTANT_CLASS_P, punt.
6524 2018-02-14 Jim Wilson <jimw@sifive.com>
6526 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6527 first SMALL_OPERAND check. New local min_second_step. Move assert
6528 to where locals are set. Add TARGET_RVC support.
6529 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6531 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6533 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6535 2018-02-14 Martin Sebor <msebor@redhat.com>
6537 PR tree-optimization/83698
6538 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6539 arrays constrain the offset range to their bounds.
6540 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6541 (builtin_access::overlap): Avoid setting the size of overlap if it's
6543 (maybe_diag_overlap): Also consider arrays when deciding what values
6544 of offsets to include in diagnostics.
6546 2018-02-14 Martin Sebor <msebor@redhat.com>
6549 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6550 that correspond to the kind of a declaration.
6552 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6555 * config/pa/pa.md: Load address of PIC label using the linkage table
6556 if the label is nonlocal.
6558 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6560 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6561 warning message if user requests -maltivec=be.
6562 * doc/invoke.texi: Document deprecation of -maltivec=be.
6564 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6567 * config/rs6000/rs6000-c.c: Update definitions for
6568 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6569 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6571 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6574 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6575 add _get_ssp intrinsics. Remove argument from
6576 __builtin_ia32_rdssp[d|q].
6577 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6578 * config/i386/i386-builtin.def: Remove argument from
6579 __builtin_ia32_rdssp[d|q].
6580 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6581 ix86_expand_special_args_builtin for _rdssp[d|q].
6582 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6583 Clear register before usage.
6584 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6585 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6587 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6589 PR tree-optimization/84357
6590 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6591 operand 1 of an ARRAY_REF too.
6593 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6596 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6597 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6599 (set_of_reg): New struct.
6600 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6601 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6602 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6604 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6605 Split into bitclr, bitset, bitinvert patterns if appropriate.
6606 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6607 use rx_fuse_in_memory_bitop.
6608 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6609 to named insn, correct maximum insn length.
6611 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6614 * machmode.def: Define a complex mode for PARTIAL_INT.
6615 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6617 * doc/rtl.texi: Document CSPImode.
6618 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6620 (msp430_hard_regno_nregs_with_padding): Likewise.
6622 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6625 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6627 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6629 PR rtl-optimization/84169
6630 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6631 we generated a parallel as new i3 and we split that to new i2 and i3
6632 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6633 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6634 those to i2, not i1. Partially rewrite this scan code.
6636 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6639 * stor-layout.c (place_field): For variable length fields, adjust
6640 offset_align afterwards not just based on the field's alignment,
6641 but also on the size.
6644 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6645 of exps and logs in the use_exp2 case.
6647 2018-02-13 Jeff Law <law@redhat.com>
6649 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6652 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6655 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6659 * common.opt (gas-loc-support, gas-locview-support): New.
6660 (ginline-points, ginternal-reset-location-views): New.
6661 * doc/invoke.texi: Document them. Use @itemx where intended.
6662 (gvariable-location-views): Adjust.
6663 * target.def (reset_location_view): New.
6664 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6665 (TARGET_RESET_LOCATION_VIEW): New.
6666 * doc/tm.texi: Rebuilt.
6667 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6668 (dwarf2out_default_as_locview_support): New.
6669 (output_asm_line_debug_info): Use option variables.
6670 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6671 (output_loc_list): Likewise.
6672 (add_high_low_attributes): Check option variables.
6673 Don't output entry view attribute in strict mode.
6674 (gen_inlined_subroutine_die): Check option variables.
6675 (dwarf2out_inline_entry): Likewise.
6676 (init_sections_and_labels): Likewise.
6677 (dwarf2out_early_finish): Likewise.
6678 (maybe_reset_location_view): New, from...
6679 (dwarf2out_var_location): ... here. Call it.
6680 * debug.h (dwarf2out_default_as_loc_support): Declare.
6681 (dwarf2out_default_as_locview_support): Declare.
6682 * hooks.c (hook_int_rtx_insn_0): New.
6683 * hooks.h (hook_int_rtx_insn_0): Declare.
6684 * toplev.c (process_options): Take -gas-loc-support and
6685 -gas-locview-support from dwarf2out. Enable
6686 -gvariable-location-views by default only with locview
6687 assembler support. Enable -ginternal-reset-location-views by
6688 default only if the target defines the corresponding hook.
6689 Enable -ginline-points by default if location views are
6690 enabled; force it disabled if statement frontiers are
6692 * tree-inline.c (expand_call_inline): Check option variables.
6693 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6695 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6697 PR tree-optimization/84321
6698 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6699 handling. Also check whether the anti-range contains any values
6700 that satisfy the mask; switch to a VR_RANGE if not.
6702 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6705 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6707 2018-02-13 Martin Jambor <mjambor@suse.cz>
6710 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6711 of call statements, also set location of a load to a temporary.
6713 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6715 * config/rl78/rl78.c (add_vector_labels): New function.
6716 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6717 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6718 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6719 which checks that no arguments are passed.
6720 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6721 * doc/extend.texi: Documentation for the new attribute.
6723 2018-02-13 Andreas Schwab <schwab@suse.de>
6725 * config/riscv/linux.h (CPP_SPEC): Define.
6727 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6730 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6731 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6732 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6733 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6734 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6735 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6736 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6737 temporarily for AES and PCLMUL builtins.
6739 PR tree-optimization/84339
6740 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6741 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6745 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6746 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6747 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6749 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6751 * omp-simd-clone.h: New file.
6752 * omp-simd-clone.c: Include omp-simd-clone.h.
6753 (expand_simd_clones): No longer static.
6754 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6755 cgraph.h and omp-simd-clone.h.
6756 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6757 (vect_recog_widen_shift_pattern): Formatting fix.
6758 (vect_pattern_recog_1): Don't check optab for calls.
6761 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6762 operands[2] into a REG before using gen_lowpart on it.
6764 2018-02-12 Jeff Law <law@redhat.com>
6767 * config/sh/sh.c (find_barrier): Consider a sibling call
6770 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6771 successfully back substituting a reg.
6773 2018-02-12 Richard Biener <rguenther@suse.de>
6775 PR tree-optimization/84037
6776 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6777 parameter, move visited init to caller.
6778 (vect_slp_analyze_operations): Separate cost from validity
6779 check, initialize visited once for all instances.
6780 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6782 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6784 (vect_model_promotion_demotion_cost): Likewise.
6785 (vectorizable_bswap): Guard cost modeling with !slp_node
6786 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6788 (vectorizable_call): Likewise.
6789 (vectorizable_conversion): Likewise.
6790 (vectorizable_assignment): Likewise.
6791 (vectorizable_shift): Likewise.
6792 (vectorizable_operation): Likewise.
6793 (vectorizable_store): Likewise.
6794 (vectorizable_load): Likewise.
6795 (vectorizable_condition): Likewise.
6796 (vectorizable_comparison): Likewise.
6798 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6801 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6802 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6803 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6805 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6808 * params.def (inline-min-speedup): Increase from 8 to 15.
6809 (max-inline-insns-auto): Decrease from 40 to 30.
6810 * ipa-split.c (consider_split): Add some buffer for function to
6811 be considered inlining candidate.
6812 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6815 2018-02-12 Richard Biener <rguenther@suse.de>
6817 PR tree-optimization/84037
6818 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6819 matched stmts if we cannot swap the non-matched ones.
6821 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6823 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6824 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6825 _mm_maskz_scalef_round_ss): New intrinsics.
6826 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6827 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6828 __builtin_ia32_scalefss_round): Remove.
6829 (__builtin_ia32_scalefsd_mask_round,
6830 __builtin_ia32_scalefss_mask_round): New intrinsics.
6831 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6832 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6833 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6834 "<round_constraint>")): Changed to ...
6835 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6836 "<round_scalar_constraint>")): ... this.
6837 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6838 %0, %1, %2<round_op3>}"): Changed to ...
6839 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6840 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6841 %2<round_scalar_mask_op3>}"): ... this.
6842 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6844 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6846 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6847 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6848 (_mm_maskz_sqrt_round_ss): New intrinsics.
6849 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6850 (__builtin_ia32_sqrtsd_mask_round)
6851 (__builtin_ia32_sqrtss_mask_round): New builtins.
6852 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6853 (__builtin_ia32_sqrtss_round): Remove.
6854 (__builtin_ia32_sqrtsd_mask_round)
6855 (__builtin_ia32_sqrtss_mask_round): New builtins.
6856 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6857 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6858 ((match_operand:VF_128 1 "vector_operand"
6859 "xBm,<round_constraint>")): Changed to ...
6860 ((match_operand:VF_128 1 "vector_operand"
6861 "xBm,<round_scalar_constraint>")): ... this.
6862 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6863 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6864 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6865 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6866 %<iptr>1<round_scalar_mask_op3>}): ... this.
6867 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6868 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6870 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6873 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6874 Cast vec_cmpeq result to correct type.
6875 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6876 Cast vec_cmpgt result to correct type.
6878 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6880 * final.c (final_scan_insn_1): Renamed from...
6881 (final_scan_insn): ... this. New wrapper, to recover
6882 seen from the outermost call in recursive ones.
6883 * config/sparc/sparc.c (output_return): Drop seen from call.
6884 (output_sibcall): Likewise.
6885 * config/visium/visium.c (output_branch): Likewise.
6887 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6889 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6892 2018-02-10 Alan Modra <amodra@gmail.com>
6895 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6896 Specify LR as an input.
6898 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6901 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6902 remove_member_access_dummy_vars): New functions.
6903 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6904 lower_omp_1, execute_lower_omp): Use them.
6906 PR rtl-optimization/84308
6907 * shrink-wrap.c (spread_components): Release todo vector.
6909 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6911 PR rtl-optimization/57193
6912 * ira-color.c (struct allocno_color_data): Add member
6913 conflict_allocno_hard_prefs.
6914 (update_conflict_allocno_hard_prefs): New.
6915 (bucket_allocno_compare_func): Add a preference based on
6916 conflict_allocno_hard_prefs.
6917 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6918 (color_allocnos): Remove a dead code. Initiate
6919 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6921 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6924 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6925 constraint from =wa to wa. Avoid a subreg on the output operand,
6926 instead use a pseudo and subreg it in a move.
6927 (p9_xxbrd_<mode>): Changed to ...
6928 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6929 (p9_xxbrd_v2df): New expander.
6930 (p9_xxbrw_<mode>): Changed to ...
6931 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6932 (p9_xxbrw_v4sf): New expander.
6934 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6936 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6938 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6941 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6942 multiply in 32-bit mode.
6943 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6944 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6947 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6949 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6950 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6951 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6952 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6954 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6957 * dwarf2out.c (is_trivial_indirect_ref): New function.
6958 (dwarf2out_late_global_decl): Do not generate a location
6959 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6960 and that are not defined in the current unit.
6962 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6964 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6965 instead of a libcall for UNORDERED.
6967 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6970 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6971 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6973 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6976 * config/s390/s390.c (s390_set_current_function): Invoke
6977 s390_indirect_branch_settings also if fndecl didn't change.
6979 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6981 * config/rs6000/rs6000.md (blockage): Set length to zero.
6983 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6985 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6987 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6990 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6991 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6995 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6996 PARALLEL incoming that failed vt_get_decl_and_offset check.
6999 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7000 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7002 (get_variable_section): For decls in named .bss* sections pass true as
7003 second argument to bss_initializer_p.
7005 2018-02-09 Marek Polacek <polacek@redhat.com>
7006 Jakub Jelinek <jakub@redhat.com>
7009 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7010 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7011 Sync some changes from cxx_fold_indirect_ref.
7013 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7015 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7017 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7018 (BLOCK_INLINE_ENTRY_LABEL): New.
7019 (dwarf2out_var_location): Disregard inline entry markers.
7020 (inline_entry_data): New struct.
7021 (inline_entry_data_hasher): New hashtable type.
7022 (inline_entry_data_hasher::hash): New.
7023 (inline_entry_data_hasher::equal): New.
7024 (inline_entry_data_table): New variable.
7025 (add_high_low_attributes): Add DW_AT_entry_pc and
7026 DW_AT_GNU_entry_view attributes if a pending entry is found
7027 in inline_entry_data_table. Add old entry_pc attribute only
7028 if debug nonbinding markers are disabled.
7029 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7030 markers are enabled.
7031 (block_within_block_p, dwarf2out_inline_entry): New.
7032 (dwarf2out_finish): Check that no entries remained in
7033 inline_entry_data_table.
7034 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7035 (final_scan_insn, notice_source_line): Likewise.
7036 (rest_of_clean_state): Skip inline entry markers.
7037 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7039 * gimple.c (gimple_build_debug_inline_entry): New.
7040 * gimple.h (enum gimple_debug_subcode): Add
7041 GIMPLE_DEBUG_INLINE_ENTRY.
7042 (gimple_build_debug_inline_entry): Declare.
7043 (gimple_debug_inline_entry_p): New.
7044 (gimple_debug_nonbind_marker_p): Adjust.
7045 * insn-notes.def (INLINE_ENTRY): New.
7046 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7047 inline entry marker notes.
7048 (print_insn): Likewise.
7049 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7050 (INSN_DEBUG_MARKER_KIND): Likewise.
7051 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7052 * tree-inline.c (expand_call_inline): Build and insert
7053 debug_inline_entry stmt.
7054 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7055 inline entry blocks early, if nonbind markers are enabled.
7056 (dump_scope_block): Dump fragment info.
7057 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7058 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7059 (gimple_build_debug_inline_entry): New.
7060 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7061 Enable/disable inline entry points too.
7062 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7063 (DEBUG_INSN): Describe inline entry markers.
7065 * common.opt (gvariable-location-views): New.
7066 (gvariable-location-views=incompat5): New.
7067 * config.in: Rebuilt.
7068 * configure: Rebuilt.
7069 * configure.ac: Test assembler for view support.
7070 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7071 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7072 * dwarf2out.c (var_loc_view): New typedef.
7073 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7074 (dwarf2out_locviews_in_attribute): New.
7075 (dwarf2out_locviews_in_loclist): New.
7076 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7077 (enum dw_line_info_opcode): Add LI_adv_address.
7078 (struct dw_line_info_table): Add view.
7079 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7080 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7081 (zero_view_p): New variable.
7082 (ZERO_VIEW_P): New macro.
7083 (output_asm_line_debug_info): New.
7084 (struct var_loc_node): Add view.
7085 (add_AT_view_list, AT_loc_list): New.
7086 (add_var_loc_to_decl): Add view param. Test it against last.
7087 (new_loc_list): Add view params. Record them.
7088 (AT_loc_list_ptr): Handle loc and view lists.
7089 (view_list_to_loc_list_val_node): New.
7090 (print_dw_val): Handle dw_val_class_view_list.
7091 (size_of_die): Likewise.
7092 (value_format): Likewise.
7093 (loc_list_has_views): New.
7094 (gen_llsym): Set vl_symbol too.
7095 (maybe_gen_llsym, skip_loc_list_entry): New.
7096 (dwarf2out_maybe_output_loclist_view_pair): New.
7097 (output_loc_list): Output view list or entries too.
7098 (output_view_list_offset): New.
7099 (output_die): Handle dw_val_class_view_list.
7100 (output_dwarf_version): New.
7101 (output_compilation_unit_header): Use it.
7102 (output_skeleton_debug_sections): Likewise.
7103 (output_rnglists, output_line_info): Likewise.
7104 (output_pubnames, output_aranges): Update version comments.
7105 (output_one_line_info_table): Output view numbers in asm comments.
7106 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7107 Call maybe_gen_llsym.
7108 (loc_list_from_tree_1): Adjust.
7109 (add_AT_location_description): Create view list attribute if
7110 needed, check it's absent otherwise.
7111 (convert_cfa_to_fb_loc_list): Adjust.
7112 (maybe_emit_file): Call output_asm_line_debug_info for test.
7113 (dwarf2out_var_location): Reset views as needed. Precompute
7114 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7115 attribute. Set view.
7116 (new_line_info_table): Reset next view.
7117 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7118 (dwarf2out_source_line): Likewise. Output view resets and labels to
7119 the assembler, or select appropriate line info opcodes.
7120 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7121 (optimize_string_length): Catch it. Adjust.
7122 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7123 dw_val_class_view_list, and remove it if no longer needed.
7124 (hash_loc_list): Hash view numbers.
7125 (loc_list_hasher::equal): Compare them.
7126 (optimize_location_lists): Check whether a view list symbol is
7127 needed, and whether the locview attribute is present, and
7128 whether they match. Remove the locview attribute if no longer
7130 (index_location_lists): Call skip_loc_list_entry for test.
7131 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7132 Use output_dwarf_version.
7133 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7134 (struct dw_val_node): Add val_view_list.
7135 * final.c (SEEN_NEXT_VIEW): New.
7136 (set_next_view_needed): New.
7137 (clear_next_view_needed): New.
7138 (maybe_output_next_view): New.
7139 (final_start_function): Rename to...
7140 (final_start_function_1): ... this. Take pointer to FIRST,
7141 add SEEN parameter. Emit param bindings in the initial view.
7142 (final_start_function): Reintroduce SEEN-less interface.
7143 (final): Rename to...
7144 (final_1): ... this. Take SEEN parameter. Output final pending
7145 next view at the end.
7146 (final): Reintroduce seen-less interface.
7147 (final_scan_insn): Output pending next view before switching
7148 sections or ending a block. Mark the next view as needed when
7149 outputting variable locations. Notify debug backend of section
7150 changes, and of location view changes.
7151 (rest_of_handle_final): Adjust.
7152 * toplev.c (process_options): Autodetect value for debug variable
7153 location views option. Warn on incompat5 without -gdwarf-5.
7154 * doc/invoke.texi (gvariable-location-views): New.
7155 (gvariable-location-views=incompat5): New.
7156 (gno-variable-location-views): New.
7158 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7160 PR tree-optimization/84136
7161 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7162 that the result of find_edge is non-NULL.
7164 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7167 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7168 storing integer register in SImode. Fix cost of 256 and 512
7169 byte aligned SSE register store.
7171 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7173 * config/i386/i386.c (ix86_multiplication_cost): Fix
7174 multiplication cost for TARGET_AVX512DQ.
7176 2018-02-08 Marek Polacek <polacek@redhat.com>
7178 PR tree-optimization/84238
7179 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7182 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7184 PR tree-optimization/84265
7185 * tree-vect-stmts.c (vectorizable_store): Don't treat
7186 VMAT_CONTIGUOUS accesses as grouped.
7187 (vectorizable_load): Likewise.
7189 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7191 PR tree-optimization/81635
7192 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7193 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7194 (test_round_for_mask): New functions.
7195 (wide_int_cc_tests): Call test_round_for_mask.
7196 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7197 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7198 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7199 range returned by get_range_info.
7201 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7204 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7205 * symtab.c: Include builtins.h
7206 (symtab_node::output_to_lto_symbol_table_p): Move here
7207 from lto-streamer-out.c:output_symbol_p.
7208 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7209 (output_symbol_p): Move all logic to symtab.c
7210 (produce_symtab): Update.
7212 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7214 * config/s390/s390-opts.h (enum indirect_branch): Define.
7215 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7216 (s390_indirect_branch_via_thunk)
7217 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7218 (enum s390_indirect_branch_type): Define.
7219 * config/s390/s390.c (struct s390_frame_layout, struct
7220 machine_function): Remove.
7221 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7222 (indirect_branch_table_label_no, indirect_branch_table_name):
7224 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7225 (enum s390_indirect_branch_option): Define.
7226 (s390_return_addr_from_memory): New function.
7227 (s390_handle_string_attribute): New function.
7228 (s390_attribute_table): Add new attribute handler.
7229 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7230 (s390_indirect_branch_via_thunk): New function.
7231 (s390_indirect_branch_via_inline_thunk): New function.
7232 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7233 sibling call optimization for non z10 compiles.
7234 (s390_emit_call): Force indirect branch target to be a single
7235 register. Add r1 clobber for non-z10 compiles.
7236 (s390_emit_epilogue): Emit return jump via return_use expander.
7237 (s390_reorg): Handle JUMP_INSNs as execute targets.
7238 (s390_option_override_internal): Perform validity checks for the
7239 new command line options.
7240 (s390_indirect_branch_attrvalue): New function.
7241 (s390_indirect_branch_settings): New function.
7242 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7243 (s390_output_indirect_thunk_function): New function.
7244 (s390_code_end): Implement target hook.
7245 (s390_case_values_threshold): Implement target hook.
7246 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7248 * config/s390/s390.h (struct s390_frame_layout)
7249 (struct machine_function): Move here from s390.c.
7250 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7251 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7252 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7253 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7254 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7255 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7256 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7257 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7258 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7259 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7260 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7261 (mnemonic attribute): Add values which aren't recognized
7263 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7264 pattern for branch conversion. Fix mnemonic attribute.
7265 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7266 indirect branch via thunk if requested.
7267 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7268 ("*indirect_jump"): Disable for branch conversion using out of
7270 ("indirect_jump_via_thunk<mode>_z10")
7271 ("indirect_jump_via_thunk<mode>")
7272 ("indirect_jump_via_inlinethunk<mode>_z10")
7273 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7274 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7275 ("casesi_jump_via_inlinethunk<mode>_z10")
7276 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7277 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7278 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7279 ("*indirect2_jump"): Disable for branch conversion.
7280 ("casesi_jump"): Turn into expander and expand patterns for branch
7282 ("return_use"): New expander.
7283 ("*return"): Emit return via thunk and rename it to ...
7284 ("*return<mode>"): ... this one.
7285 * config/s390/s390.opt: Add new options and and enum for the
7288 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7290 * lra-constraints.c (match_reload): Unconditionally use
7291 gen_lowpart_SUBREG, rather than selecting between that
7292 and equivalent gen_rtx_SUBREG code.
7294 2018-02-08 Richard Biener <rguenther@suse.de>
7296 PR tree-optimization/84233
7297 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7298 changed flag instead of boguously re-using phi_inserted.
7300 2018-02-08 Martin Jambor <mjambor@suse.cz>
7302 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7303 static local variables.
7305 2018-02-08 Richard Biener <rguenther@suse.de>
7307 PR tree-optimization/84278
7308 * tree-vect-stmts.c (vectorizable_store): When looking for
7309 smaller vector types to perform grouped strided loads/stores
7310 make sure the mode is supported by the target.
7311 (vectorizable_load): Likewise.
7313 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7315 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7316 Increase LDP/STP opportunities by adding adjacent callee-saves.
7318 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7320 PR rtl-optimization/84068
7321 PR rtl-optimization/83459
7322 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7324 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7326 PR tree-optimization/84224
7327 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7328 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7331 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7334 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7335 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7336 count, remove one USE.
7338 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7340 * doc/install.texi (Configuration): Document the
7341 --with-long-double-format={ibm,ieee} PowerPC configuration
7345 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7346 Convert from define_expand to be define_insn_and_split. Rework
7347 float/double/_Float128 conversions to QI/HI/SImode to work with
7348 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7349 conversions to QI/HImode types did a store and then a load to
7350 truncate the value. For conversions to VSX registers, don't split
7351 the insn, instead emit the code directly. Use the code iterator
7352 any_fix to combine signed and unsigned conversions.
7353 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7354 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7355 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7356 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7357 (fix_<mode>di2_hw): Likewise.
7358 (fixuns_<mode>di2_hw): Likewise.
7359 (fix_<mode>si2_hw): Likewise.
7360 (fixuns_<mode>si2_hw): Likewise.
7361 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7362 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7363 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7364 fix<uns>_trunc<SFDF:mode>si2_p8.
7365 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7367 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7368 (fix<uns>_<mode>_mem): Likewise.
7369 (fctiw<u>z_<mode>_mem): Likewise.
7370 (fix<uns>_<mode>_mem): Likewise.
7371 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7372 the register allocator from doing a direct move to the GPRs to do
7373 a store, and instead use the ISA 3.0 store byte/half-word from
7374 vector register instruction. For IEEE 128-bit floating point,
7375 also optimize stores of 32-bit ints.
7376 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7378 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7380 * genextract.c (push_pathstr_operand): New function to support
7382 (walk_rtx): Call push_pathstr_operand.
7383 (print_path): Support [a-zA-Z].
7385 2018-02-07 Richard Biener <rguenther@suse.de>
7387 PR tree-optimization/84037
7388 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7389 (cse_and_gimplify_to_preheader): Declare.
7390 (vect_get_place_in_interleaving_chain): Likewise.
7391 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7393 (_loop_vec_info::~_loop_vec_info): Delete it.
7394 (cse_and_gimplify_to_preheader): New function.
7395 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7396 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7397 (vectorizable_load): Likewise. For grouped stores always base
7398 the IV on the first element.
7399 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7400 condition before gimplifying.
7402 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7404 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7405 *DIV_EXPR and *MOD_EXPR.
7407 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7410 * config/i386/i386.c (ix86_option_override_internal): Mask out
7411 the CF_SET bit when checking -fcf-protection.
7413 2018-02-07 Tom de Vries <tom@codesourcery.com>
7416 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7419 2018-02-07 Richard Biener <rguenther@suse.de>
7421 PR tree-optimization/84204
7422 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7425 PR tree-optimization/84205
7426 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7427 special-case isl_ast_op_zdiv_r.
7429 PR tree-optimization/84223
7430 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7431 Only add conditions from within the region.
7432 (gather_bbs::after_dom_children): Adjust.
7434 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7437 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7438 * config/avr/avr.md: Only post-reload split REG-REG moves if
7439 either register is GENERAL_REG_P.
7441 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7443 PR tree-optimization/84235
7444 * tree-ssa-scopedtables.c
7445 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7446 if the subtraction is performed in floating point type where NaNs are
7447 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7448 build 1. Formatting fix.
7450 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7453 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7454 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7455 and skip it regardless of bb boundaries. Use CALL_P macro,
7456 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7457 unnecessarily, formatting fix.
7459 2018-02-06 Michael Collison <michael.collison@arm.com>
7461 * config/arm/thumb2.md:
7462 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7463 (*thumb_mov_notscc): Ditto.
7465 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7468 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7469 unsigned_fix, not "s".
7471 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7473 * configure.ac (gcc_fn_eh_frame_ro): New function.
7474 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7475 correct .eh_frame permissions.
7476 * configure: Regenerate.
7478 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7480 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7483 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7485 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7486 Display warning message for -mno-speculate-indirect-jumps.
7488 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7490 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7492 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7494 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7496 PR tree-optimization/84225
7497 * tree-eh.c (find_trapping_overflow): Only call
7498 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7500 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7503 * config/i386/i386.c: Reimplement the check of possible options
7504 -mibt/-mshstk conbination. Change error messages.
7505 * doc/invoke.texi: Fix a typo: remove extra '='.
7507 2018-02-06 Marek Polacek <polacek@redhat.com>
7509 PR tree-optimization/84228
7510 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7512 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7515 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7516 emitted arch directives.
7517 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7518 __ARM_FEATURE_COPROC before changing architectures.
7520 2018-02-06 Richard Biener <rguenther@suse.de>
7522 * config/i386/i386.c (print_reg): Fix typo.
7523 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7525 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7527 * configure: Regenerate.
7529 2018-02-05 Martin Sebor <msebor@redhat.com>
7531 PR tree-optimization/83369
7532 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7535 2018-02-05 Martin Liska <mliska@suse.cz>
7537 * doc/invoke.texi: Cherry-pick upstream r323995.
7539 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7541 * ira.c (ira_init_register_move_cost): Adjust comment.
7543 2018-02-05 Martin Liska <mliska@suse.cz>
7545 PR gcov-profile/84137
7546 * doc/gcov.texi: Fix typo in documentation.
7548 2018-02-05 Martin Liska <mliska@suse.cz>
7550 PR gcov-profile/83879
7551 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7552 using dlopen functionality.
7554 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7556 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7557 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7558 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7559 _mm_maskz_range_round_ss): New intrinsics.
7560 (__builtin_ia32_rangesd128_round)
7561 (__builtin_ia32_rangess128_round): Remove.
7562 (__builtin_ia32_rangesd128_mask_round,
7563 __builtin_ia32_rangess128_mask_round): New builtins.
7564 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7565 __builtin_ia32_rangess128_round): Remove.
7566 (__builtin_ia32_rangesd128_mask_round,
7567 __builtin_ia32_rangess128_mask_round): New builtins.
7568 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7569 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7570 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7571 "<round_saeonly_constraint>")): Changed to ...
7572 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7573 "<round_saeonly_scalar_constraint>")): ... this.
7574 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7575 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7576 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7577 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7578 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7580 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7582 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7584 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7585 Remove all values except native, 8540 and 8548.
7587 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7589 * config/i386/i386.c (ix86_output_function_return): Pass
7590 INVALID_REGNUM, instead of -1, as invalid register number to
7591 indirect_thunk_name and output_indirect_thunk.
7593 2018-02-02 Julia Koval <julia.koval@intel.com>
7595 * config.gcc: Add -march=icelake.
7596 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7597 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7598 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7599 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7600 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7601 (processor_target_table): Add icelake.
7602 (ix86_option_override_internal): Handle new PTAs.
7603 (get_builtin_code_for_version): Handle icelake.
7604 (M_INTEL_COREI7_ICELAKE): New.
7605 (fold_builtin_cpu): Handle icelake.
7606 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7607 * doc/invoke.texi: Add -march=icelake.
7609 2018-02-02 Julia Koval <julia.koval@intel.com>
7611 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7612 to wide_int_bitmask.
7613 * wide-int-bitmask.h: New.
7615 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7618 * config/i386/i386.md: Replace Pmode with word_mode in
7619 builtin_setjmp_setup and builtin_longjmp to support x32.
7621 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7625 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7627 (rs6000_supported_cpu_names): New static variable.
7628 (linux_cpu_translation_table): Likewise.
7629 (elf_platform) <cpu>: Define new static variable and use it.
7630 Translate kernel AT_PLATFORM name to canonical name if needed.
7631 Error if platform name is unknown.
7633 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7636 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7638 2018-02-01 Jeff Law <law@redhat.com>
7641 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7642 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7643 the scratch if RELEASE_VIA_POP is false.
7644 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7645 If we have to save a temporary register, decrement SIZE appropriately.
7646 Pass new arguments to release_scratch_register_on_entry.
7647 (ix86_adjust_stack_and_probe): Likewise.
7648 (ix86_emit_probe_stack_range): Pass new arguments to
7649 release_scratch_register_on_entry.
7651 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7653 PR rtl-optimization/84157
7654 * combine.c (change_zero_ext): Use REG_P predicate in
7655 front of HARD_REGISTER_P predicate.
7657 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7659 * config/avr/avr.c (avr_option_override): Move disabling of
7660 -fdelete-null-pointer-checks to...
7661 * common/config/avr/avr-common.c (avr_option_optimization_table):
7664 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7666 PR tree-optimization/81635
7667 * tree-data-ref.c (split_constant_offset_1): For types that
7668 wrap on overflow, try to use range info to prove that wrapping
7671 2018-02-01 Renlin Li <renlin.li@arm.com>
7674 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7676 (aarch64_register_move_cost): Likewise.
7677 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7679 (REG_CLASS_NAMES): Likewise.
7680 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7681 TAILCALL_ADDR_REGS. Remove IP registers.
7682 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7684 2018-02-01 Richard Biener <rguenther@suse.de>
7686 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7687 for specifying RPO order and allow NULL for that.
7688 * domwalk.c (dom_walker::dom_walker): Likewise.
7689 (dom_walker::walk): Handle NULL RPO order.
7690 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7692 (rewrite_update_dom_walker): Likewise.
7693 (mark_def_dom_walker): Likewise.
7695 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7697 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7698 (aarch64_maybe_expand_sve_subreg_move): Declare.
7699 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7700 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7702 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7703 that are semantically a reverse operation.
7704 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7705 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7706 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7708 (aarch64_can_change_mode_class): For big-endian, forbid changes
7709 between two SVE modes if they have different element sizes.
7711 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7713 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7714 the TImode handling for big-endian targets.
7716 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7718 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7719 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7721 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7722 Remove BSWAP handing for big-endian targets and use the form of
7723 LD1RQ appropariate for the mode.
7725 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7727 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7728 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7731 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7734 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7735 check for operands that need to go through aarch64_sve_reload_be.
7737 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7739 PR tree-optimization/81661
7740 PR tree-optimization/84117
7741 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7742 * tree-eh.c: Include gimplify.h.
7743 (find_trapping_overflow, replace_trapping_overflow,
7744 rewrite_to_non_trapping_overflow): New functions.
7745 * tree-vect-loop.c: Include tree-eh.h.
7746 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7747 * tree-data-ref.c: Include tree-eh.h.
7748 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7750 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7752 PR rtl-optimization/84123
7753 * combine.c (change_zero_ext): Check if hard register satisfies
7754 can_change_dest_mode before calling gen_lowpart_SUBREG.
7756 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7759 * ira.c (ira_init_register_move_cost): Remove assert.
7761 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7763 PR rtl-optimization/84071
7764 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7765 * doc/tm.texi: Regenerate.
7767 2018-01-31 Richard Biener <rguenther@suse.de>
7769 PR tree-optimization/84132
7770 * tree-data-ref.c (analyze_miv_subscript): Properly
7771 check whether evolution_function_is_affine_multivariate_p
7772 before calling gcd_of_steps_may_divide_p.
7774 2018-01-31 Julia Koval <julia.koval@intel.com>
7777 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7778 * config/i386/i386.md (rdpid_rex64) New.
7779 (rdpid): Make 32bit only.
7781 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7784 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7785 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7787 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7790 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7792 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7794 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7796 PR rtl-optimization/84071
7797 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7798 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7800 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7802 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7803 (arc_attribute_table): Add 'aux' attribute.
7804 (arc_in_small_data_p): Consider aux like variables.
7805 (arc_is_aux_reg_p): New function.
7806 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7807 (arc_get_aux_arg): New function.
7808 (prepare_move_operands): Handle aux-register access.
7809 (arc_handle_aux_attribute): New function.
7810 * doc/extend.texi (ARC Variable attributes): Add subsection.
7812 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7814 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7815 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7816 (arc_attribute_table): Add 'uncached' attribute.
7817 (arc_print_operand): Print '.di' flag for uncached memory
7819 (arc_in_small_data_p): Do not consider for small data the uncached
7821 (arc_is_uncached_mem_p): New function.
7822 * config/arc/predicates.md (compact_store_memory_operand): Check
7823 for uncached memory accesses.
7824 (nonvol_nonimm_operand): Likewise.
7825 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7827 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7830 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7831 falign-loops=): Add Optimization flag.
7833 2018-01-30 Jeff Law <law@redhat.com>
7836 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7837 INT_REGISTERS_SAVED. Check it prior to calling
7838 get_scratch_register_on_entry.
7839 (ix86_adjust_stack_and_probe): Similarly.
7840 (ix86_emit_probe_stack_range): Similarly.
7841 (ix86_expand_prologue): Corresponding changes.
7843 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7846 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7847 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7849 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7852 * lra-constraints.c (curr_insn_transform): Process AND in the
7855 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7857 PR rtl-optimization/83986
7858 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7859 dependence against last_pending_memory_flush in addition to
7862 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7864 PR tree-optimization/81611
7865 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7868 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7871 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7874 2018-01-30 Richard Biener <rguenther@suse.de>
7875 Jakub Jelinek <jakub@redhat.com>
7877 PR tree-optimization/84111
7878 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7879 inner loops added during recursion, as they don't have up-to-date
7882 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7885 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7886 (can_inline_edge_by_limits_p): ... here.
7887 (can_early_inline_edge_p, check_callers,
7888 update_caller_keys, update_callee_keys, recursive_inlining,
7889 add_new_edges_to_heap, speculation_useful_p,
7890 inline_small_functions,
7891 inline_small_functions, flatten_function,
7892 inline_to_all_callers_1): Update.
7894 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7896 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7899 2018-01-30 Richard Biener <rguenther@suse.de>
7901 PR tree-optimization/83008
7902 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7903 invariant and constant vector uses in stmts when they need
7906 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7909 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7910 * configure: Regenerate.
7912 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7914 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7916 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7917 Use gen_rtx_REG rather than gen_lowpart.
7919 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7921 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7922 rather than 0 when creating partial subregs.
7924 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7926 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7929 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7932 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7933 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7934 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7935 flags. This restores the settings used before the 2017-07-24.
7936 Turning off pre increment/decrement/modify allows IVOPTS to
7937 optimize DF/SF loops where the index is an int.
7939 2018-01-29 Richard Biener <rguenther@suse.de>
7940 Kelvin Nilsen <kelvin@gcc.gnu.org>
7943 * tree-vect-stmts.c (vectorizable_call): Don't call
7944 targetm.vectorize_builtin_md_vectorized_function if callee is
7947 2018-01-22 Carl Love <cel@us.ibm.com>
7949 * doc/extend.tex: Fix typo in second arg in
7950 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7952 2018-01-29 Richard Biener <rguenther@suse.de>
7954 PR tree-optimization/84086
7955 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7956 (flush_ssaname_freelist): When SSA names were released reset
7957 the SCEV hash table.
7959 2018-01-29 Richard Biener <rguenther@suse.de>
7961 PR tree-optimization/84057
7962 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7963 removed paths when removing edges.
7965 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7967 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7968 -mfunction-return=@var{choice}.
7970 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7973 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7975 (layout::print_source_line): Likewise.
7976 (test_get_line_width_without_trailing_whitespace): Add test cases.
7978 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7981 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7984 2018-01-26 Jim Wilson <jimw@sifive.com>
7986 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7988 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7991 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7993 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7994 and CMP + SUB-immediate -> SUBS.
7996 2018-01-26 Martin Sebor <msebor@redhat.com>
7998 PR tree-optimization/83896
7999 * tree-ssa-strlen.c (get_string_len): Rename...
8000 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8001 Avoid assuming length is constant.
8002 (handle_char_store): Use HOST_WIDE_INT for string length.
8004 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8007 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8008 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8010 2018-01-26 Richard Biener <rguenther@suse.de>
8012 PR rtl-optimization/84003
8013 * dse.c (record_store): Only record redundant stores when
8014 the earlier store aliases at least all accesses the later one does.
8016 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8018 PR rtl-optimization/83985
8019 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8020 REG_CFA_RESTORE insns.
8021 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8022 REG_CFA_RESTORE insns here.
8025 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8026 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8028 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8030 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8031 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8032 (arc_init): Likewise.
8033 (arc_override_options): Likewise.
8034 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8036 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8038 * config/arc/arc.h (TARGET_DBNZ): Define.
8039 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8040 properly set the tune attribute.
8041 (dbnz): Use TARGET_DBNZ guard.
8042 * config/arc/arc.opt (mtune): Add core3 option.
8044 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8046 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8047 recognize new pic like addresses.
8048 (arc_delegitimize_address): Clean up.
8050 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8052 * config/arc/arc-arches.def: Option mrf16 valid for all
8054 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8055 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8056 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8057 * config/arc/arc-tables.opt: Regenerate.
8058 * config/arc/arc.c (arc_conditional_register_usage): Handle
8059 reduced register file case.
8060 (arc_file_start): Set must have build attributes.
8061 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8063 * config/arc/arc.opt (mrf16): Add new option.
8064 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8065 * config/arc/genmultilib.awk: Handle new mrf16 option.
8066 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8067 * config/arc/t-multilib: Regenerate.
8068 * doc/invoke.texi (ARC Options): Document mrf16 option.
8070 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8072 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8073 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8074 (arc_attribute_table): Add 'secure_call' attribute.
8075 (arc_print_operand): Print secure call operand.
8076 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8078 (arc_is_secure_call_p): New function. * config/arc/arc.md
8079 (call_i): Add support for sjli instruction.
8080 (call_value_i): Likewise.
8081 * config/arc/constraints.md (Csc): New constraint.
8083 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8084 John Eric Martin <John.Martin@emmicro-us.com>
8086 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8087 * config/arc/arc.c (_arc_jli_section): New struct.
8088 (arc_jli_section): New type.
8089 (rc_jli_sections): New static variable.
8090 (arc_handle_jli_attribute): New function.
8091 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8092 (arc_file_end): New function.
8093 (TARGET_ASM_FILE_END): Define.
8094 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8095 (arc_add_jli_section): New function.
8096 (jli_call_scan): Likewise.
8097 (arc_reorg): Call jli_call_scan.
8098 (arc_output_addsi): Remove 'S' from printing asm operand.
8099 (arc_is_jli_call_p): New function.
8100 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8102 (movhi_insn): Likewise.
8103 (movsi_insn): Likewise.
8104 (movsi_set_cc_insn): Likewise.
8105 (loadqi_update): Likewise.
8106 (load_zeroextendqisi_update): Likewise.
8107 (load_signextendqisi_update): Likewise.
8108 (loadhi_update): Likewise.
8109 (load_zeroextendhisi_update): Likewise.
8110 (load_signextendhisi_update): Likewise.
8111 (loadsi_update): Likewise.
8112 (loadsf_update): Likewise.
8113 (movsicc_insn): Likewise.
8114 (bset_insn): Likewise.
8115 (bxor_insn): Likewise.
8116 (bclr_insn): Likewise.
8117 (bmsk_insn): Likewise.
8118 (bicsi3_insn): Likewise.
8119 (cmpsi_cc_c_insn): Likewise.
8120 (movsi_ne): Likewise.
8121 (movsi_cond_exec): Likewise.
8122 (clrsbsi2): Likewise.
8133 (call_i): Remove 'S' asm letter, add jli instruction.
8134 (call_value_i): Likewise.
8135 * config/arc/arc.op (mjli-always): New option.
8136 * config/arc/constraints.md (Cji): New constraint.
8137 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8139 (subsf3_fpx): Likewise.
8140 (mulsf3_fpx): Likewise.
8141 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8143 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8145 * doc/invoke.texi (ARC): Document mjli-always option.
8147 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8149 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8150 avoid addition with 0 and use incw and decw where possible.
8152 2018-01-26 Richard Biener <rguenther@suse.de>
8154 PR tree-optimization/81082
8155 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8156 association if it requires casting to unsigned.
8157 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8158 from fold_plusminus_mult_expr to catch important cases late when
8159 range info is available.
8161 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8163 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8164 * configure.ac (hidden_linkonce): New test.
8165 * configure: Regenerate.
8166 * config.in: Regenerate.
8168 2018-01-26 Julia Koval <julia.koval@intel.com>
8170 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8171 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8172 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8173 _mm_mask_bitshuffle_epi64_mask): Fix type.
8174 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8175 USI_FTYPE_V4DI_V4DI_USI): Remove.
8176 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8177 __builtin_ia32_vpshufbitqmb256_mask,
8178 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8179 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8180 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8182 2018-01-26 Alan Modra <amodra@gmail.com>
8185 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8186 UNSPEC_VBPERMQ. Sort other unspecs.
8188 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8190 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8192 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8195 * predict.c (drop_profile): Do not push/pop cfun; update also
8197 (handle_missing_profiles): Fix logic looking for zero profiles.
8199 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8202 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8203 on functions with #pragma omp declare simd or functions with simd
8205 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8206 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8207 Remove trailing \n from warning_at calls.
8209 2018-01-25 Tom de Vries <tom@codesourcery.com>
8212 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8213 for neutered workers.
8215 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8218 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8219 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8221 2018-01-24 Jeff Law <law@redhat.com>
8224 * i386.c (get_probe_interval): Move to earlier point.
8225 (ix86_compute_frame_layout): If -fstack-clash-protection and
8226 the frame is larger than the probe interval, then use pushes
8227 to save registers rather than reg->mem moves.
8228 (ix86_expand_prologue): Remove conditional for int_registers_saved
8231 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8234 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8235 min/max for never referenced object.
8237 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8240 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8242 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8243 attributes from DECL_ATTRIBUTES (decl) without affecting
8244 DECL_ATTRIBUTES (current_function_decl).
8245 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8246 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8248 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8250 PR tree-optimization/83979
8251 * fold-const.c (fold_comparison): Use constant_boolean_node
8252 instead of boolean_{true,false}_node.
8254 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8256 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8259 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8261 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8262 Simplify the clause that sets the length attribute.
8263 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8264 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8265 clause that sets the length attribute.
8266 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8268 2018-01-24 Tom de Vries <tom@codesourcery.com>
8271 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8272 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8273 Add strict parameter.
8274 (prevent_branch_around_nothing): Insert dummy insn between branch to
8275 label and label with no ptx insn inbetween.
8276 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8278 2018-01-24 Tom de Vries <tom@codesourcery.com>
8281 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8282 for neutered threads in warp.
8283 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8285 2018-01-24 Richard Biener <rguenther@suse.de>
8287 PR tree-optimization/83176
8288 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8291 2018-01-24 Richard Biener <rguenther@suse.de>
8293 PR tree-optimization/82819
8294 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8295 code generating pluses that are no-ops in the target precision.
8297 2018-01-24 Richard Biener <rguenther@suse.de>
8300 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8302 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8304 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8305 to merge probabilities.
8306 * predict.c (probably_never_executed): Also mark as cold functions
8307 with global 0 profile and guessed local profile.
8308 * profile-count.c (profile_probability::combine_with_count): New
8310 * profile-count.h (profile_probability::operator*,
8311 profile_probability::operator*=, profile_probability::operator/,
8312 profile_probability::operator/=): Reduce precision to adjusted
8313 and set value to guessed on contradictory divisions.
8314 (profile_probability::combine_with_freq): Remove.
8315 (profile_probability::combine_wiht_count): Declare.
8316 (profile_count::force_nonzero):: Set to adjusted.
8317 (profile_count::probability_in):: Set quality to adjusted.
8318 * tree-ssa-tail-merge.c (replace_block_by): Use
8321 2018-01-23 Andrew Waterman <andrew@sifive.com>
8322 Jim Wilson <jimw@sifive.com>
8324 * config/riscv/riscv.c (riscv_stack_boundary): New.
8325 (riscv_option_override): Set riscv_stack_boundary. Handle
8326 riscv_preferred_stack_boundary_arg.
8327 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8328 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8329 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8330 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8331 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8332 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8334 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8337 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8338 of struct ix86_frame.
8339 (ix86_expand_epilogue): Likewise. Add a local variable for
8340 the reg_save_offset field in struct ix86_frame.
8342 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8344 PR tree-optimization/82604
8345 * tree-loop-distribution.c (enum partition_kind): New enum item
8346 PKIND_PARTIAL_MEMSET.
8347 (partition_builtin_p): Support above new enum item.
8348 (generate_code_for_partition): Ditto.
8349 (compute_access_range): Differentiate cases that equality can be
8350 proven at all loops, the innermost loops or no loops.
8351 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8352 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8353 (finalize_partitions, distribute_loop): Don't fuse partition of
8354 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8355 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8358 2018-01-23 Martin Liska <mliska@suse.cz>
8360 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8361 order to ignore the predictor.
8362 (PRED_POLYMORPHIC_CALL): Likewise.
8363 (PRED_RECURSIVE_CALL): Likewise.
8365 2018-01-23 Martin Liska <mliska@suse.cz>
8367 * tree-profile.c (tree_profiling): Print function header to
8368 aware reader which function we are working on.
8369 * value-prof.c (gimple_find_values_to_profile): Do not print
8370 not interesting value histograms.
8372 2018-01-23 Martin Liska <mliska@suse.cz>
8374 * profile-count.h (enum profile_quality): Add
8375 profile_uninitialized as the first value. Do not number values
8376 as they are zero based.
8377 (profile_count::verify): Update sanity check.
8378 (profile_probability::verify): Likewise.
8380 2018-01-23 Nathan Sidwell <nathan@acm.org>
8382 * doc/invoke.texi (ffor-scope): Deprecate.
8384 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8386 PR tree-optimization/83510
8387 * domwalk.c (set_all_edges_as_executable): New function.
8388 (dom_walker::dom_walker): Convert bool param
8389 "skip_unreachable_blocks" to enum reachability. Move setup of
8390 edge flags to set_all_edges_as_executable and only do it when
8391 reachability is REACHABLE_BLOCKS.
8392 * domwalk.h (enum dom_walker::reachability): New enum.
8393 (dom_walker::dom_walker): Convert bool param
8394 "skip_unreachable_blocks" to enum reachability.
8395 (set_all_edges_as_executable): New decl.
8396 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8397 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8399 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8400 but converting true to REACHABLE_BLOCKS.
8401 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8403 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8404 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8405 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8407 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8408 if check_all_array_refs will be called.
8410 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8412 * tree.c (selftest::test_location_wrappers): Add more test
8415 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8417 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8418 (selftest::test_bit_in_range): Likewise.
8420 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8423 * doc/sourcebuild.texi (vect_float): Say that the selector
8424 only describes the situation when -funsafe-math-optimizations is on.
8425 (vect_float_strict): Document.
8427 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8429 PR tree-optimization/83965
8430 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8431 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8432 instead of checking only for a reduction.
8433 (vect_recog_widen_sum_pattern): Likewise.
8435 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8437 * predict.c (probably_never_executed): Only use precise profile info.
8438 (compute_function_frequency): Skip after inlining hack since we now
8439 have quality checking.
8441 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8443 * profile-count.h (profile_probability::very_unlikely,
8444 profile_probability::unlikely, profile_probability::even): Set
8445 precision to guessed.
8447 2018-01-23 Richard Biener <rguenther@suse.de>
8449 PR tree-optimization/83963
8450 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8451 Properly terminate dominator walk when crossing the exit edge not
8452 when visiting its source block.
8454 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8457 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8458 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8460 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8462 PR tree-optimization/83957
8463 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8464 semicolon after for body surrounded by braces.
8466 PR tree-optimization/83081
8467 * profile-count.h (profile_probability::split): New method.
8468 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8469 Use profile_probability::split.
8470 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8471 when splitting a single conditional jump into 2.
8473 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8475 PR tree-optimization/69452
8476 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8479 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8481 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8482 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8483 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8485 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8487 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8488 * config/rl78/rl78.md (movdi): New define_expand.
8489 * config/rl78/rl78.c (rl78_split_movdi): New function.
8491 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8494 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8496 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8497 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8498 128-bit to produce an UNSPEC move to get the double word with the
8499 signbit and then a shift directly to do signbit.
8500 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8501 implementation with a new version that just does either a direct
8502 move or a regular move. Move memory interface to separate insns.
8503 Move insns so they are next to the expander.
8504 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8505 with signbit move. Split big and little endian case.
8506 (signbit<mode>2_dm_mem_le): Likewise.
8507 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8508 (signbit<mode>2_dm2): Likewise.
8510 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8512 * config/rl78/rl78.md (anddi3): New define_expand.
8514 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8516 * config/rl78/rl78.md (umindi3): New define_expand.
8518 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8520 * config/rl78/rl78.md (smindi3): New define_expand.
8522 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8524 * config/rl78/rl78.md (smaxdi3): New define_expand.
8526 2018-01-22 Carl Love <cel@us.ibm.com>
8528 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8529 LVX_V1TI): Add macro expansion.
8530 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8531 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8532 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8533 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8534 Change check to determine if the instruction is a byte reversing
8535 entry. Fix typo in comment.
8536 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8537 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8538 Add def_builtin calls for new builtins.
8539 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8540 Add define_insn expansion.
8542 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8544 * config/rl78/rl78.md (umaxdi3): New define_expand.
8546 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8548 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8549 for non-QImode registers.
8551 2018-01-22 Richard Biener <rguenther@suse.de>
8553 PR tree-optimization/83963
8554 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8555 including the loop exit block.
8556 (scop_detection::merge_sese): Likewise.
8557 (scop_detection::add_scop): Do it here instead.
8559 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8561 * doc/sourcebuild.texi (arm_softfloat): Document.
8563 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8566 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8567 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8568 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8570 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8571 David Edelsohn <dje.gcc@gmail.com>
8574 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8575 Change "crset eq" to "crset 2".
8576 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8577 (*call_indirect_aix<mode>_nospec): Likewise.
8578 (*call_value_indirect_aix<mode>_nospec): Likewise.
8579 (*call_indirect_elfv2<mode>_nospec): Likewise.
8580 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8581 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8582 change assembly output from . to $.
8583 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8584 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8585 (*tablejump<mode>_internal1_nospec): Likewise.
8587 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8590 * config/sh/sh_optimize_sett_clrt.cc:
8591 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8593 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8595 PR tree-optimization/83940
8596 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8597 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8598 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8599 use it to pass back the definition type.
8600 (vect_check_store_rhs): Likewise rhs_dt_out.
8601 (vect_build_gather_load_calls): Add a mask_dt argument and use
8602 it instead of a call to vect_is_simple_use.
8603 (vectorizable_store): Update calls to vect_check_load_store_mask
8604 and vect_check_store_rhs. Use the dt returned by the latter instead
8605 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8606 instead of calls to vect_is_simple_use. Pass the scalar rather
8607 than the vector operand to vect_is_simple_use when handling
8608 second and subsequent copies of an rhs value.
8609 (vectorizable_load): Update calls to vect_check_load_store_mask
8610 and vect_build_gather_load_calls. Use the cached mask_dt and
8611 gs_info.offset_dt instead of calls to vect_is_simple_use.
8613 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8616 * tree-emutls.c: Include gimplify.h.
8617 (lower_emutls_2): New function.
8618 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8619 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8620 it before further processing.
8623 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8624 UINTVAL (trueop1) instead of INTVAL (op1).
8626 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8630 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8631 INCOMING_FRAME_SP_OFFSET if not defined.
8632 (scan_trace): Add ENTRY argument. If true and
8633 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8634 emit a note to adjust the CFA offset.
8635 (create_cfi_notes): Adjust scan_trace callers.
8636 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8637 INCOMING_FRAME_SP_OFFSET in the CIE.
8638 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8639 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8641 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8642 * doc/tm.texi: Regenerated.
8644 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8646 PR rtl-optimization/83147
8647 * lra-constraints.c (remove_inheritance_pseudos): Use
8648 lra_substitute_pseudo_within_insn.
8650 2018-01-19 Tom de Vries <tom@codesourcery.com>
8651 Cesar Philippidis <cesar@codesourcery.com>
8654 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8656 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8659 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8660 spaces for function labels.
8662 2018-01-19 Martin Liska <mliska@suse.cz>
8664 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8665 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8666 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8667 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8668 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8669 (PRED_CONST_RETURN): Change from 69 to 65.
8670 (PRED_NULL_RETURN): Change from 91 to 71.
8671 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8672 (PRED_LOOP_GUARD): Change from 66 to 73.
8674 2018-01-19 Martin Liska <mliska@suse.cz>
8676 * predict.c (predict_insn_def): Add new assert.
8677 (struct branch_predictor): Change type to signed integer.
8678 (test_prediction_value_range): Amend test to cover
8680 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8681 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8682 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8683 (PRED_LOOP_IV_COMPARE): Likewise.
8684 * predict.h (PROB_UNINITIALIZED): Define new constant.
8686 2018-01-19 Martin Liska <mliska@suse.cz>
8688 * predict.c (dump_prediction): Add new format for
8689 analyze_brprob.py script which is enabled with -details
8691 * profile-count.h (precise_p): New function.
8693 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8695 PR tree-optimization/83922
8696 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8697 there are no statements that need masking.
8698 (vect_active_double_reduction_p): New function.
8699 (vect_analyze_loop_operations): Use it when handling phis that
8700 are not in the loop header.
8702 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8704 PR tree-optimization/83914
8705 * tree-vect-loop.c (vectorizable_induction): Don't convert
8706 init_expr or apply the peeling adjustment for inductions
8707 that are nested within the vectorized loop.
8709 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8711 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8714 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8718 * function.h (gimplify_parameters): Add gimple_seq * argument.
8719 * function.c: Include gimple.h and options.h.
8720 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8721 for the added local temporaries if needed.
8722 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8723 if there are any parameter cleanups, wrap whole body into a
8724 try/finally with the cleanups.
8726 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8729 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8730 Use GET_MODE_CLASS for scalar floating point.
8732 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8736 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8737 Fix call of call_cgraph_insertion_hooks.
8739 2018-01-18 Martin Sebor <msebor@redhat.com>
8741 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8743 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8746 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8749 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8752 * common.opt: (-ffile-prefix-map): New option.
8753 * opts.c (common_handle_option): Defer it.
8754 * opts-global.c (handle_common_deferred_options): Handle it.
8755 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8756 * file-prefix-map.h: New file.
8757 (remap_debug_filename, add_debug_prefix_map): ...here.
8758 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8759 * final.c (debug_prefix_map, add_debug_prefix_map
8760 remap_debug_filename): Move to...
8761 * file-prefix-map.c: New file.
8762 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8763 generalize, get rid of alloca(), use strrchr() instead of strchr().
8764 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8765 Implement in terms of add_prefix_map().
8766 (remap_macro_filename, remap_debug_filename): Implement in term of
8768 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8769 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8770 * dbxout.c: Include file-prefix-map.h.
8771 * varasm.c: Likewise.
8772 * vmsdbgout.c: Likewise.
8773 * xcoffout.c: Likewise.
8774 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8775 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8776 * doc/invoke.texi (-ffile-prefix-map): Document.
8777 (-fdebug-prefix-map): Update description.
8779 2018-01-18 Martin Liska <mliska@suse.cz>
8781 * config/i386/i386.c (indirect_thunk_name): Document that also
8783 (output_indirect_thunk): Document why both instructions
8784 (pause and lfence) are generated.
8786 2018-01-18 Richard Biener <rguenther@suse.de>
8788 PR tree-optimization/83887
8789 * graphite-scop-detection.c
8790 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8791 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8792 (scop_detection::merge_sese): Re-implement with a flood-fill
8793 algorithm that properly finds a SESE region if it exists.
8795 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8798 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8799 pointer_diff optimizations use view_convert instead of convert.
8801 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8803 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8804 Generate different code for -mno-speculate-indirect-jumps.
8805 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8806 (*call_indirect_aix<mode>): Disable for
8807 -mno-speculate-indirect-jumps.
8808 (*call_indirect_aix<mode>_nospec): New define_insn.
8809 (*call_value_indirect_aix<mode>): Disable for
8810 -mno-speculate-indirect-jumps.
8811 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8812 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8813 -mno-speculate-indirect-jumps.
8814 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8816 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8818 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8819 long double type, set the flags for noting the default long double
8820 type, even if we don't pass or return a long double type.
8822 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8825 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8828 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8830 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8831 support for merge[hl].
8832 (fold_mergehl_helper): New helper function.
8833 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8834 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8835 (altivec_vmrglw_direct): Add xxmrglw insn.
8837 2018-01-17 Andrew Waterman <andrew@sifive.com>
8839 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8840 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8842 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8845 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8846 call the lto_location_cache before reading the
8847 DECL_SOURCE_LOCATION of the types.
8849 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8850 Richard Sandiford <richard.sandiford@linaro.org>
8852 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8853 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8854 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8855 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8856 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8858 * config/aarch64/constraints.md (aarch64_movti_operand):
8860 * config/aarch64/predicates.md (Uti): Add new constraint.
8862 2018-01-17 Carl Love <cel@us.ibm.com>
8864 * config/rs6000/vsx.md (define_expand xl_len_r,
8865 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8866 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8868 (define_expand, define_insn): Move the shift left from the
8869 define_insn to the define_expand for lxvl and stxvl instructions.
8870 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8871 and XL_LEN_R definitions to PURE.
8873 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8875 * config/i386/i386.c (indirect_thunk_name): Declare regno
8876 as unsigned int. Compare regno with INVALID_REGNUM.
8877 (output_indirect_thunk): Ditto.
8878 (output_indirect_thunk_function): Ditto.
8879 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8880 in the call to output_indirect_thunk_function.
8882 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8885 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8886 rather than the size of inner_type to determine the stack slot size
8887 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8889 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8892 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8895 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8897 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8898 endian Linux systems to optionally enable multilibs for selecting
8899 the long double type if the user configured an explicit type.
8900 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8901 have no long double multilibs if not defined.
8902 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8903 warn if the user used -mabi={ieee,ibm}longdouble and we built
8904 multilibs for long double.
8905 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8906 appropriate multilib option.
8907 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8909 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8910 for building long double multilibs.
8911 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8913 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8915 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8918 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8920 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8923 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8926 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8929 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8931 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8932 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8934 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8936 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8937 different rtl trees depending on TARGET_64BIT.
8938 (rs6000_gen_lvx): Likewise.
8940 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8942 * config/visium/visium.md (nop): Tweak comment.
8943 (hazard_nop): Likewise.
8945 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8947 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8948 -mspeculate-indirect-jumps.
8949 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8950 for -mno-speculate-indirect-jumps.
8951 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8952 (*call_value_indirect_elfv2<mode>): Disable for
8953 -mno-speculate-indirect-jumps.
8954 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8955 (indirect_jump): Emit different RTL for
8956 -mno-speculate-indirect-jumps.
8957 (*indirect_jump<mode>): Disable for
8958 -mno-speculate-indirect-jumps.
8959 (*indirect_jump<mode>_nospec): New define_insn.
8960 (tablejump): Emit different RTL for
8961 -mno-speculate-indirect-jumps.
8962 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8963 (tablejumpsi_nospec): New define_expand.
8964 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8965 (tablejumpdi_nospec): New define_expand.
8966 (*tablejump<mode>_internal1): Disable for
8967 -mno-speculate-indirect-jumps.
8968 (*tablejump<mode>_internal1_nospec): New define_insn.
8969 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8972 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8974 * caller-save.c (insert_save): Drop unnecessary parameter. All
8977 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8978 Richard Biener <rguenth@suse.de>
8981 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8982 return early, inline manually is_gimple_sizepos. Make sure if we
8983 call gimplify_expr we don't end up with a gimple constant.
8984 * tree.c (variably_modified_type_p): Don't return true for
8985 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8986 * gimplify.h (is_gimple_sizepos): Remove.
8988 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8990 PR tree-optimization/83857
8991 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8992 vectorizable_live_operation for pure SLP statements.
8993 (vectorizable_live_operation): Handle PHIs.
8995 2018-01-16 Richard Biener <rguenther@suse.de>
8997 PR tree-optimization/83867
8998 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8999 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9001 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9004 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9005 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9006 If off is not INTEGER_CST, issue a may not be aligned warning
9007 rather than isn't aligned. Use isn%'t rather than isn't.
9008 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9010 <case MULT_EXPR>: Improve the case when bottom and one of the
9011 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9012 operand, in that case check if the other operand is multiple of
9013 bottom divided by the INTEGER_CST operand.
9015 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9018 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9019 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9020 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9021 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9022 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9023 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9024 (pa_function_arg_size): New function.
9026 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9028 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9029 in a separate statement.
9031 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9033 PR tree-optimization/83847
9034 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9035 group gathers and scatters.
9037 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9039 PR rtl-optimization/86620
9040 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9042 PR rtl-optimization/83213
9043 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9044 to last if both are JUMP_INSNs.
9046 PR tree-optimization/83843
9047 * gimple-ssa-store-merging.c
9048 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9049 store_immediate_info for bswap/nop orig_stores.
9051 2018-01-15 Andrew Waterman <andrew@sifive.com>
9053 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9055 <UDIV>: Increase cost if !TARGET_DIV.
9057 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9059 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9060 (define_attr "cr_logical_3op"): New.
9061 (cceq_ior_compare): Adjust.
9062 (cceq_ior_compare_complement): Adjust.
9063 (*cceq_rev_compare): Adjust.
9064 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9065 (is_cracked_insn): Adjust.
9066 (insn_must_be_first_in_group): Adjust.
9067 * config/rs6000/40x.md: Adjust.
9068 * config/rs6000/440.md: Adjust.
9069 * config/rs6000/476.md: Adjust.
9070 * config/rs6000/601.md: Adjust.
9071 * config/rs6000/603.md: Adjust.
9072 * config/rs6000/6xx.md: Adjust.
9073 * config/rs6000/7450.md: Adjust.
9074 * config/rs6000/7xx.md: Adjust.
9075 * config/rs6000/8540.md: Adjust.
9076 * config/rs6000/cell.md: Adjust.
9077 * config/rs6000/e300c2c3.md: Adjust.
9078 * config/rs6000/e500mc.md: Adjust.
9079 * config/rs6000/e500mc64.md: Adjust.
9080 * config/rs6000/e5500.md: Adjust.
9081 * config/rs6000/e6500.md: Adjust.
9082 * config/rs6000/mpc.md: Adjust.
9083 * config/rs6000/power4.md: Adjust.
9084 * config/rs6000/power5.md: Adjust.
9085 * config/rs6000/power6.md: Adjust.
9086 * config/rs6000/power7.md: Adjust.
9087 * config/rs6000/power8.md: Adjust.
9088 * config/rs6000/power9.md: Adjust.
9089 * config/rs6000/rs64.md: Adjust.
9090 * config/rs6000/titan.md: Adjust.
9092 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9094 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9095 ix86_indirect_branch_register logic.
9097 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9099 * config/i386/constraints.md (Bs): Update
9100 ix86_indirect_branch_register check. Don't check
9101 ix86_indirect_branch_register with GOT_memory_operand.
9103 * config/i386/predicates.md (GOT_memory_operand): Don't check
9104 ix86_indirect_branch_register here.
9105 (GOT32_symbol_operand): Likewise.
9107 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9109 * config/i386/predicates.md (constant_call_address_operand):
9110 Rewrite ix86_indirect_branch_register logic.
9111 (sibcall_insn_operand): Likewise.
9113 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9115 * config/i386/constraints.md (Bs): Replace
9116 ix86_indirect_branch_thunk_register with
9117 ix86_indirect_branch_register.
9119 * config/i386/i386.md (indirect_jump): Likewise.
9120 (tablejump): Likewise.
9121 (*sibcall_memory): Likewise.
9122 (*sibcall_value_memory): Likewise.
9123 Peepholes of indirect call and jump via memory: Likewise.
9124 * config/i386/i386.opt: Likewise.
9125 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9126 (GOT_memory_operand): Likewise.
9127 (call_insn_operand): Likewise.
9128 (sibcall_insn_operand): Likewise.
9129 (GOT32_symbol_operand): Likewise.
9131 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9134 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9135 type rather than type addr's type points to.
9136 (expand_omp_atomic_mutex): Likewise.
9137 (expand_omp_atomic): Likewise.
9139 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9142 * config/i386/i386.c (output_indirect_thunk_function): Use
9143 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9144 for __x86_return_thunk.
9146 2018-01-15 Richard Biener <rguenther@suse.de>
9149 * expmed.c (extract_bit_field_1): Fix typo.
9151 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9154 * config/arm/iterators.md (VF): New mode iterator.
9155 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9156 Remove integer-related logic from pattern.
9157 (neon_vabd<mode>_3): Likewise.
9159 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9162 * common.opt (fstrict-overflow): No longer an alias.
9163 (fwrapv-pointer): New option.
9164 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9165 also for pointer types based on flag_wrapv_pointer.
9166 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9167 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9168 opts->x_flag_wrapv got set.
9169 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9170 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9171 POINTER_TYPE_OVERFLOW_UNDEFINED.
9172 * match.pd: Likewise in address comparison pattern.
9173 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9175 2018-01-15 Richard Biener <rguenther@suse.de>
9178 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9179 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9180 Reset type names to their identifier if their TYPE_DECL doesn't
9181 have linkage (and thus is used for ODR and devirt).
9182 (save_debug_info_for_decl): Remove.
9183 (save_debug_info_for_type): Likewise.
9184 (add_tree_to_fld_list): Adjust.
9185 * tree-pretty-print.c (dump_generic_node): Make dumping of
9186 type names more robust.
9188 2018-01-15 Richard Biener <rguenther@suse.de>
9190 * BASE-VER: Bump to 8.0.1.
9192 2018-01-14 Martin Sebor <msebor@redhat.com>
9195 * builtins.c (check_access): Avoid warning when the no-warning bit
9198 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9200 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9201 * ira-color (allocno_hard_regs_compare): Likewise.
9203 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9206 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9207 Use .pushsection/.popsection.
9209 2018-01-14 Martin Sebor <msebor@redhat.com>
9212 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9214 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9216 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9217 entry from extra_headers.
9218 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9219 extra_headers, make the list bitwise identical to the i?86-*-* one.
9221 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9223 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9224 -mcmodel=large with -mindirect-branch=thunk,
9225 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9226 -mfunction-return=thunk-extern.
9227 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9228 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9229 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9231 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9233 * config/i386/i386.c (print_reg): Print the name of the full
9234 integer register without '%'.
9235 (ix86_print_operand): Handle 'V'.
9236 * doc/extend.texi: Document 'V' modifier.
9238 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9240 * config/i386/constraints.md (Bs): Disallow memory operand for
9241 -mindirect-branch-register.
9243 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9244 (GOT_memory_operand): Likewise.
9245 (call_insn_operand): Likewise.
9246 (sibcall_insn_operand): Likewise.
9247 (GOT32_symbol_operand): Likewise.
9248 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9249 for -mindirect-branch-register.
9250 (tablejump): Likewise.
9251 (*sibcall_memory): Likewise.
9252 (*sibcall_value_memory): Likewise.
9253 Disallow peepholes of indirect call and jump via memory for
9254 -mindirect-branch-register.
9255 (*call_pop): Replace m with Bw.
9256 (*call_value_pop): Likewise.
9257 (*sibcall_pop_memory): Replace m with Bs.
9258 * config/i386/i386.opt (mindirect-branch-register): New option.
9259 * doc/invoke.texi: Document -mindirect-branch-register option.
9261 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9263 * config/i386/i386-protos.h (ix86_output_function_return): New.
9264 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9265 set function_return_type.
9266 (indirect_thunk_name): Add ret_p to indicate thunk for function
9268 (output_indirect_thunk_function): Pass false to
9269 indirect_thunk_name.
9270 (ix86_output_indirect_branch_via_reg): Likewise.
9271 (ix86_output_indirect_branch_via_push): Likewise.
9272 (output_indirect_thunk_function): Create alias for function
9273 return thunk if regno < 0.
9274 (ix86_output_function_return): New function.
9275 (ix86_handle_fndecl_attribute): Handle function_return.
9276 (ix86_attribute_table): Add function_return.
9277 * config/i386/i386.h (machine_function): Add
9278 function_return_type.
9279 * config/i386/i386.md (simple_return_internal): Use
9280 ix86_output_function_return.
9281 (simple_return_internal_long): Likewise.
9282 * config/i386/i386.opt (mfunction-return=): New option.
9283 (indirect_branch): Mention -mfunction-return=.
9284 * doc/extend.texi: Document function_return function attribute.
9285 * doc/invoke.texi: Document -mfunction-return= option.
9287 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9289 * config/i386/i386-opts.h (indirect_branch): New.
9290 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9291 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9292 with local indirect jump when converting indirect call and jump.
9293 (ix86_set_indirect_branch_type): New.
9294 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9295 (indirectlabelno): New.
9296 (indirect_thunk_needed): Likewise.
9297 (indirect_thunk_bnd_needed): Likewise.
9298 (indirect_thunks_used): Likewise.
9299 (indirect_thunks_bnd_used): Likewise.
9300 (INDIRECT_LABEL): Likewise.
9301 (indirect_thunk_name): Likewise.
9302 (output_indirect_thunk): Likewise.
9303 (output_indirect_thunk_function): Likewise.
9304 (ix86_output_indirect_branch_via_reg): Likewise.
9305 (ix86_output_indirect_branch_via_push): Likewise.
9306 (ix86_output_indirect_branch): Likewise.
9307 (ix86_output_indirect_jmp): Likewise.
9308 (ix86_code_end): Call output_indirect_thunk_function if needed.
9309 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9311 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9312 (ix86_attribute_table): Add indirect_branch.
9313 * config/i386/i386.h (machine_function): Add indirect_branch_type
9314 and has_local_indirect_jump.
9315 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9317 (tablejump): Likewise.
9318 (*indirect_jump): Use ix86_output_indirect_jmp.
9319 (*tablejump_1): Likewise.
9320 (simple_return_indirect_internal): Likewise.
9321 * config/i386/i386.opt (mindirect-branch=): New option.
9322 (indirect_branch): New.
9325 (thunk-inline): Likewise.
9326 (thunk-extern): Likewise.
9327 * doc/extend.texi: Document indirect_branch function attribute.
9328 * doc/invoke.texi: Document -mindirect-branch= option.
9330 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9333 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9335 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9337 * ipa-inline.c (want_inline_small_function_p): Return false if
9338 inlining has already failed with CIF_FINAL_ERROR.
9339 (update_caller_keys): Call want_inline_small_function_p before
9341 (update_callee_keys): Likewise.
9343 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9345 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9347 (rs6000_quadword_masked_address_p): Likewise.
9348 (quad_aligned_load_p): Likewise.
9349 (quad_aligned_store_p): Likewise.
9350 (const_load_sequence_p): Add comment to describe the outer-most loop.
9351 (mimic_memory_attributes_and_flags): New function.
9352 (rs6000_gen_stvx): Likewise.
9353 (replace_swapped_aligned_store): Likewise.
9354 (rs6000_gen_lvx): Likewise.
9355 (replace_swapped_aligned_load): Likewise.
9356 (replace_swapped_load_constant): Capitalize argument name in
9357 comment describing this function.
9358 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9359 and stores that access quad-word aligned addresses and replace
9360 with stvx or lvx instructions when appropriate.
9361 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9362 New function prototype.
9363 (rs6000_quadword_masked_address_p): Likewise.
9364 (rs6000_gen_lvx): Likewise.
9365 (rs6000_gen_stvx): Likewise.
9366 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9367 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9368 when memory address is aligned.
9369 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9370 this split to select lvx instruction when memory address is aligned.
9371 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9372 instruction when memory address is aligned.
9373 (*vsx_le_perm_load_v16qi): Likewise.
9374 (four unnamed splitters): Modify to select the stvx instruction
9375 when memory is aligned.
9377 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9379 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9380 which appears in the queue multiple times.
9382 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9383 Alan Hayward <alan.hayward@arm.com>
9384 David Sherwood <david.sherwood@arm.com>
9386 * tree-vectorizer.h (vec_lower_bound): New structure.
9387 (_loop_vec_info): Add check_nonzero and lower_bounds.
9388 (LOOP_VINFO_CHECK_NONZERO): New macro.
9389 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9390 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9391 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9392 fields. Make seg_len the distance travelled, not including the
9394 (dr_direction_indicator): Declare.
9395 (dr_zero_step_indicator): Likewise.
9396 (dr_known_forward_stride_p): Likewise.
9397 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9399 (runtime_alias_check_p): Allow runtime alias checks with
9401 (operator ==): Compare access_size and align.
9402 (prune_runtime_alias_test_list): Rework for new distinction between
9403 the access_size and seg_len.
9404 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9406 (get_segment_min_max): New function.
9407 (create_intersect_range_checks): Use it.
9408 (dr_step_indicator): New function.
9409 (dr_direction_indicator): Likewise.
9410 (dr_zero_step_indicator): Likewise.
9411 (dr_known_forward_stride_p): Likewise.
9412 * tree-loop-distribution.c (data_ref_segment_size): Return
9413 DR_STEP * (niters - 1).
9414 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9416 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9417 (vect_preserves_scalar_order_p): New function, split out from...
9418 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9419 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9420 (vect_vfa_access_size): New function.
9421 (vect_vfa_align): Likewise.
9422 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9423 (dump_lower_bound): New function.
9424 (vect_check_lower_bound): Likewise.
9425 (vect_small_gap_p): Likewise.
9426 (vectorizable_with_step_bound_p): Likewise.
9427 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9428 depencies if the vectorization factor is 1. Convert the checks
9429 for nonzero steps into checks on the bounds of DR_STEP. Try using
9430 a bunds check for variable steps if the minimum required step is
9431 relatively small. Update calls to the dr_with_seg_len
9432 constructor and to vect_compile_time_alias.
9433 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9435 (vect_loop_versioning): Call it.
9436 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9438 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9440 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9441 Alan Hayward <alan.hayward@arm.com>
9442 David Sherwood <david.sherwood@arm.com>
9444 * doc/sourcebuild.texi (vect_scatter_store): Document.
9445 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9447 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9449 * genopinit.c (main): Add supports_vec_scatter_store and
9450 supports_vec_scatter_store_cached to target_optabs.
9451 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9452 IFN_MASK_SCATTER_STORE.
9453 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9455 * internal-fn.h (internal_store_fn_p): Declare.
9456 (internal_fn_stored_value_index): Likewise.
9457 * internal-fn.c (scatter_store_direct): New macro.
9458 (expand_scatter_store_optab_fn): New function.
9459 (direct_scatter_store_optab_supported_p): New macro.
9460 (internal_store_fn_p): New function.
9461 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9462 IFN_MASK_SCATTER_STORE.
9463 (internal_fn_mask_index): Likewise.
9464 (internal_fn_stored_value_index): New function.
9465 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9467 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9468 * optabs-query.c (supports_vec_scatter_store_p): New function.
9469 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9470 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9471 true for scatter stores.
9472 (vect_gather_scatter_fn_p): Handle scatter stores too.
9473 (vect_check_gather_scatter): Consider using scatter stores if
9474 supports_vec_scatter_store_p.
9475 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9477 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9478 internal_fn_stored_value_index.
9479 (check_load_store_masking): Handle scatter stores too.
9480 (vect_get_store_rhs): Make public.
9481 (vectorizable_call): Use internal_store_fn_p.
9482 (vectorizable_store): Handle scatter store internal functions.
9483 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9484 when deciding whether the end of the group has been reached.
9485 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9486 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9487 (mask_scatter_store<mode>): New insns.
9489 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9490 Alan Hayward <alan.hayward@arm.com>
9491 David Sherwood <david.sherwood@arm.com>
9493 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9494 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9495 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9497 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9498 Use vect_truncate_gather_scatter_offset if we can't treat the
9499 operation as a normal gather load or scatter store.
9500 (get_group_load_store_type): Take the gather_scatter_info
9501 as argument. Try using a gather load or scatter store for
9502 single-element groups.
9503 (get_load_store_type): Update calls to get_group_load_store_type
9504 and vect_use_strided_gather_scatters_p.
9506 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9507 Alan Hayward <alan.hayward@arm.com>
9508 David Sherwood <david.sherwood@arm.com>
9510 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9511 optional tree argument.
9512 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9514 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9515 but continue to use the current value as a fallback.
9516 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9517 to compare the updates.
9518 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9519 (get_load_store_type): Use it when handling a strided access.
9520 (vect_get_strided_load_store_ops): New function.
9521 (vect_get_data_ptr_increment): Likewise.
9522 (vectorizable_load): Handle strided gather loads. Always pass
9523 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9525 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9526 Alan Hayward <alan.hayward@arm.com>
9527 David Sherwood <david.sherwood@arm.com>
9529 * doc/md.texi (gather_load@var{m}): Document.
9530 (mask_gather_load@var{m}): Likewise.
9531 * genopinit.c (main): Add supports_vec_gather_load and
9532 supports_vec_gather_load_cached to target_optabs.
9533 * optabs-tree.c (init_tree_optimization_optabs): Use
9534 ggc_cleared_alloc to allocate target_optabs.
9535 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9536 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9538 * internal-fn.h (internal_load_fn_p): Declare.
9539 (internal_gather_scatter_fn_p): Likewise.
9540 (internal_fn_mask_index): Likewise.
9541 (internal_gather_scatter_fn_supported_p): Likewise.
9542 * internal-fn.c (gather_load_direct): New macro.
9543 (expand_gather_load_optab_fn): New function.
9544 (direct_gather_load_optab_supported_p): New macro.
9545 (direct_internal_fn_optab): New function.
9546 (internal_load_fn_p): Likewise.
9547 (internal_gather_scatter_fn_p): Likewise.
9548 (internal_fn_mask_index): Likewise.
9549 (internal_gather_scatter_fn_supported_p): Likewise.
9550 * optabs-query.c (supports_at_least_one_mode_p): New function.
9551 (supports_vec_gather_load_p): Likewise.
9552 * optabs-query.h (supports_vec_gather_load_p): Declare.
9553 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9554 and memory_type field.
9555 (NUM_PATTERNS): Bump to 15.
9556 * tree-vect-data-refs.c: Include internal-fn.h.
9557 (vect_gather_scatter_fn_p): New function.
9558 (vect_describe_gather_scatter_call): Likewise.
9559 (vect_check_gather_scatter): Try using internal functions for
9560 gather loads. Recognize existing calls to a gather load function.
9561 (vect_analyze_data_refs): Consider using gather loads if
9562 supports_vec_gather_load_p.
9563 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9564 (vect_get_gather_scatter_offset_type): Likewise.
9565 (vect_convert_mask_for_vectype): Likewise.
9566 (vect_add_conversion_to_patterm): Likewise.
9567 (vect_try_gather_scatter_pattern): Likewise.
9568 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9569 (vect_vect_recog_func_ptrs): Add it.
9570 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9571 internal_fn_mask_index and internal_gather_scatter_fn_p.
9572 (check_load_store_masking): Take the gather_scatter_info as an
9573 argument and handle gather loads.
9574 (vect_get_gather_scatter_ops): New function.
9575 (vectorizable_call): Check internal_load_fn_p.
9576 (vectorizable_load): Likewise. Handle gather load internal
9578 (vectorizable_store): Update call to check_load_store_masking.
9579 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9580 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9581 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9582 (aarch64_gather_scale_operand_d): New predicates.
9583 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9584 (mask_gather_load<mode>): New insns.
9586 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9587 Alan Hayward <alan.hayward@arm.com>
9588 David Sherwood <david.sherwood@arm.com>
9590 * optabs.def (fold_left_plus_optab): New optab.
9591 * doc/md.texi (fold_left_plus_@var{m}): Document.
9592 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9593 * internal-fn.c (fold_left_direct): Define.
9594 (expand_fold_left_optab_fn): Likewise.
9595 (direct_fold_left_optab_supported_p): Likewise.
9596 * fold-const-call.c (fold_const_fold_left): New function.
9597 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9598 * tree-parloops.c (valid_reduction_p): New function.
9599 (gather_scalar_reductions): Use it.
9600 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9601 (vect_finish_replace_stmt): Declare.
9602 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9603 (needs_fold_left_reduction_p): New function, split out from...
9604 (vect_is_simple_reduction): ...here. Accept reductions that
9605 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9606 (vect_force_simple_reduction): Also store the reduction type in
9607 the assignment's STMT_VINFO_REDUC_TYPE.
9608 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9609 (merge_with_identity): New function.
9610 (vect_expand_fold_left): Likewise.
9611 (vectorize_fold_left_reduction): Likewise.
9612 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9613 scalar phi in place for it. Check for target support and reject
9614 cases that would reassociate the operation. Defer the transform
9615 phase to vectorize_fold_left_reduction.
9616 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9617 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9618 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9620 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9622 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9623 call to ifc_temp_var.
9625 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9626 Alan Hayward <alan.hayward@arm.com>
9627 David Sherwood <david.sherwood@arm.com>
9629 * target.def (legitimize_address_displacement): Take the original
9630 offset as a poly_int.
9631 * targhooks.h (default_legitimize_address_displacement): Update
9633 * targhooks.c (default_legitimize_address_displacement): Likewise.
9634 * doc/tm.texi: Regenerate.
9635 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9636 as an argument, moving assert of ad->disp == ad->disp_term to...
9637 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9638 Try calling targetm.legitimize_address_displacement before expanding
9639 the address rather than afterwards, and adjust for the new interface.
9640 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9641 Match the new hook interface. Handle SVE addresses.
9642 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9645 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9647 * Makefile.in (OBJS): Add early-remat.o.
9648 * target.def (select_early_remat_modes): New hook.
9649 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9650 * doc/tm.texi: Regenerate.
9651 * targhooks.h (default_select_early_remat_modes): Declare.
9652 * targhooks.c (default_select_early_remat_modes): New function.
9653 * timevar.def (TV_EARLY_REMAT): New timevar.
9654 * passes.def (pass_early_remat): New pass.
9655 * tree-pass.h (make_pass_early_remat): Declare.
9656 * early-remat.c: New file.
9657 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9659 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9661 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9662 Alan Hayward <alan.hayward@arm.com>
9663 David Sherwood <david.sherwood@arm.com>
9665 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9666 vfm1 with a bound_epilog parameter.
9667 (vect_do_peeling): Update calls accordingly, and move the prologue
9668 call earlier in the function. Treat the base bound_epilog as 0 for
9669 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9670 this base when peeling for gaps.
9671 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9672 with fully-masked loops.
9673 (vect_estimate_min_profitable_iters): Handle the single peeled
9674 iteration in that case.
9676 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9677 Alan Hayward <alan.hayward@arm.com>
9678 David Sherwood <david.sherwood@arm.com>
9680 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9681 single-element interleaving even if the size is not a power of 2.
9682 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9683 accesses for single-element interleaving if the group size is
9686 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9687 Alan Hayward <alan.hayward@arm.com>
9688 David Sherwood <david.sherwood@arm.com>
9690 * doc/md.texi (fold_extract_last_@var{m}): Document.
9691 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9692 * optabs.def (fold_extract_last_optab): New optab.
9693 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9694 * internal-fn.c (fold_extract_direct): New macro.
9695 (expand_fold_extract_optab_fn): Likewise.
9696 (direct_fold_extract_optab_supported_p): Likewise.
9697 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9698 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9699 EXTRACT_LAST_REDUCTION.
9700 (get_initial_def_for_reduction): Do not create an initial vector
9701 for EXTRACT_LAST_REDUCTION reductions.
9702 (vectorizable_reduction): Leave the scalar phi in place for
9703 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9704 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9705 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9706 transform phase to vectorizable_condition.
9707 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9709 (vect_finish_stmt_generation): ...here.
9710 (vect_finish_replace_stmt): New function.
9711 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9712 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9714 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9716 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9717 Alan Hayward <alan.hayward@arm.com>
9718 David Sherwood <david.sherwood@arm.com>
9720 * doc/md.texi (extract_last_@var{m}): Document.
9721 * optabs.def (extract_last_optab): New optab.
9722 * internal-fn.def (EXTRACT_LAST): New internal function.
9723 * internal-fn.c (cond_unary_direct): New macro.
9724 (expand_cond_unary_optab_fn): Likewise.
9725 (direct_cond_unary_optab_supported_p): Likewise.
9726 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9727 loops using EXTRACT_LAST.
9728 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9729 (extract_last_<mode>): ...this optab.
9730 (vec_extract<mode><Vel>): Update accordingly.
9732 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9733 Alan Hayward <alan.hayward@arm.com>
9734 David Sherwood <david.sherwood@arm.com>
9736 * target.def (empty_mask_is_expensive): New hook.
9737 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9738 * doc/tm.texi: Regenerate.
9739 * targhooks.h (default_empty_mask_is_expensive): Declare.
9740 * targhooks.c (default_empty_mask_is_expensive): New function.
9741 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9742 if the target says that empty masks are expensive.
9743 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9745 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9747 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9748 Alan Hayward <alan.hayward@arm.com>
9749 David Sherwood <david.sherwood@arm.com>
9751 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9752 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9753 (vect_use_loop_mask_for_alignment_p): New function.
9754 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9755 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9756 niters_skip argument. Make sure that the first niters_skip elements
9757 of the first iteration are inactive.
9758 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9759 Update call to vect_set_loop_masks_directly.
9760 (get_misalign_in_elems): New function, split out from...
9761 (vect_gen_prolog_loop_niters): ...here.
9762 (vect_update_init_of_dr): Take a code argument that specifies whether
9763 the adjustment should be added or subtracted.
9764 (vect_update_init_of_drs): Likewise.
9765 (vect_prepare_for_masked_peels): New function.
9766 (vect_do_peeling): Skip prologue peeling if we're using a mask
9767 instead. Update call to vect_update_inits_of_drs.
9768 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9770 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9771 alignment. Do not include the number of peeled iterations in
9772 the minimum threshold in that case.
9773 (vectorizable_induction): Adjust the start value down by
9774 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9775 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9776 Take the number of skipped iterations into account when calculating
9778 * tree-vect-stmts.c (vect_gen_while_not): New function.
9780 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9781 Alan Hayward <alan.hayward@arm.com>
9782 David Sherwood <david.sherwood@arm.com>
9784 * doc/sourcebuild.texi (vect_fully_masked): Document.
9785 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9787 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9789 (vect_analyze_loop_2): ...here. Don't check the vectorization
9790 factor against the number of loop iterations if the loop is
9793 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9794 Alan Hayward <alan.hayward@arm.com>
9795 David Sherwood <david.sherwood@arm.com>
9797 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9798 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9799 (dump_groups): Update accordingly.
9800 (iv_use::mem_type): New member variable.
9801 (address_p): New function.
9802 (record_use): Add a mem_type argument and initialize the new
9804 (record_group_use): Add a mem_type argument. Use address_p.
9805 Remove obsolete null checks of base_object. Update call to record_use.
9806 (find_interesting_uses_op): Update call to record_group_use.
9807 (find_interesting_uses_cond): Likewise.
9808 (find_interesting_uses_address): Likewise.
9809 (get_mem_type_for_internal_fn): New function.
9810 (find_address_like_use): Likewise.
9811 (find_interesting_uses_stmt): Try find_address_like_use before
9812 calling find_interesting_uses_op.
9813 (addr_offset_valid_p): Use the iv mem_type field as the type
9814 of the addressed memory.
9815 (add_autoinc_candidates): Likewise.
9816 (get_address_cost): Likewise.
9817 (split_small_address_groups_p): Use address_p.
9818 (split_address_groups): Likewise.
9819 (add_iv_candidate_for_use): Likewise.
9820 (autoinc_possible_for_pair): Likewise.
9821 (rewrite_groups): Likewise.
9822 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9823 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9824 (get_alias_ptr_type_for_ptr_address): New function.
9825 (rewrite_use_address): Rewrite address uses in calls that were
9826 identified by find_address_like_use.
9828 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9829 Alan Hayward <alan.hayward@arm.com>
9830 David Sherwood <david.sherwood@arm.com>
9832 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9834 * gimple-expr.h (is_gimple_addressable: Likewise.
9835 * gimple-expr.c (is_gimple_address): Likewise.
9836 * internal-fn.c (expand_call_mem_ref): New function.
9837 (expand_mask_load_optab_fn): Use it.
9838 (expand_mask_store_optab_fn): Likewise.
9840 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9841 Alan Hayward <alan.hayward@arm.com>
9842 David Sherwood <david.sherwood@arm.com>
9844 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9845 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9846 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9847 (cond_umax@var{mode}): Document.
9848 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9849 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9850 (cond_umin_optab, cond_umax_optab): New optabs.
9851 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9852 (COND_IOR, COND_XOR): New internal functions.
9853 * internal-fn.h (get_conditional_internal_fn): Declare.
9854 * internal-fn.c (cond_binary_direct): New macro.
9855 (expand_cond_binary_optab_fn): Likewise.
9856 (direct_cond_binary_optab_supported_p): Likewise.
9857 (get_conditional_internal_fn): New function.
9858 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9859 Cope with reduction statements that are vectorized as calls rather
9861 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9862 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9863 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9864 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9865 (UNSPEC_COND_EOR): New unspecs.
9866 (optab): Add mappings for them.
9867 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9868 (sve_int_op, sve_fp_op): New int attributes.
9870 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9871 Alan Hayward <alan.hayward@arm.com>
9872 David Sherwood <david.sherwood@arm.com>
9874 * optabs.def (while_ult_optab): New optab.
9875 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9876 * internal-fn.def (WHILE_ULT): New internal function.
9877 * internal-fn.h (direct_internal_fn_supported_p): New override
9878 that takes two types as argument.
9879 * internal-fn.c (while_direct): New macro.
9880 (expand_while_optab_fn): New function.
9881 (convert_optab_supported_p): Likewise.
9882 (direct_while_optab_supported_p): New macro.
9883 * wide-int.h (wi::udiv_ceil): New function.
9884 * tree-vectorizer.h (rgroup_masks): New structure.
9885 (vec_loop_masks): New typedef.
9886 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9888 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9889 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9890 (vect_max_vf): New function.
9891 (slpeel_make_loop_iterate_ntimes): Delete.
9892 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9893 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9894 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9895 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9896 internal-fn.h, stor-layout.h and optabs-query.h.
9897 (vect_set_loop_mask): New function.
9898 (add_preheader_seq): Likewise.
9899 (add_header_seq): Likewise.
9900 (interleave_supported_p): Likewise.
9901 (vect_maybe_permute_loop_masks): Likewise.
9902 (vect_set_loop_masks_directly): Likewise.
9903 (vect_set_loop_condition_masked): Likewise.
9904 (vect_set_loop_condition_unmasked): New function, split out from
9905 slpeel_make_loop_iterate_ntimes.
9906 (slpeel_make_loop_iterate_ntimes): Rename to..
9907 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9908 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9909 (vect_do_peeling): Update call accordingly.
9910 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9912 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9913 mask_compare_type, can_fully_mask_p and fully_masked_p.
9914 (release_vec_loop_masks): New function.
9915 (_loop_vec_info): Use it to free the loop masks.
9916 (can_produce_all_loop_masks_p): New function.
9917 (vect_get_max_nscalars_per_iter): Likewise.
9918 (vect_verify_full_masking): Likewise.
9919 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9920 retries, and free the mask rgroups before retrying. Check loop-wide
9921 reasons for disallowing fully-masked loops. Make the final decision
9922 about whether use a fully-masked loop or not.
9923 (vect_estimate_min_profitable_iters): Do not assume that peeling
9924 for the number of iterations will be needed for fully-masked loops.
9925 (vectorizable_reduction): Disable fully-masked loops.
9926 (vectorizable_live_operation): Likewise.
9927 (vect_halve_mask_nunits): New function.
9928 (vect_double_mask_nunits): Likewise.
9929 (vect_record_loop_mask): Likewise.
9930 (vect_get_loop_mask): Likewise.
9931 (vect_transform_loop): Handle the case in which the final loop
9932 iteration might handle a partial vector. Call vect_set_loop_condition
9933 instead of slpeel_make_loop_iterate_ntimes.
9934 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9935 (check_load_store_masking): New function.
9936 (prepare_load_store_mask): Likewise.
9937 (vectorizable_store): Handle fully-masked loops.
9938 (vectorizable_load): Likewise.
9939 (supportable_widening_operation): Use vect_halve_mask_nunits for
9941 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9942 (vect_gen_while): New function.
9943 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9944 (aarch64_uqdec<mode>): New insn.
9946 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9947 Alan Hayward <alan.hayward@arm.com>
9948 David Sherwood <david.sherwood@arm.com>
9950 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9951 (reduc_xor_scal_optab): New optabs.
9952 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9953 (reduc_xor_scal_@var{m}): Document.
9954 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9955 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9957 * fold-const-call.c (fold_const_call): Handle them.
9958 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9959 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9960 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9961 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9962 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9963 (UNSPEC_XORV): New unspecs.
9964 (optab): Add entries for them.
9965 (BITWISEV): New int iterator.
9966 (bit_reduc_op): New int attributes.
9968 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9969 Alan Hayward <alan.hayward@arm.com>
9970 David Sherwood <david.sherwood@arm.com>
9972 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9973 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9974 * optabs.def (vec_shl_insert_optab): New optab.
9975 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9976 (duplicate_and_interleave): Likewise.
9977 * tree-vect-loop.c: Include internal-fn.h.
9978 (neutral_op_for_slp_reduction): New function, split out from
9979 get_initial_defs_for_reduction.
9980 (get_initial_def_for_reduction): Handle option 2 for variable-length
9981 vectors by loading the neutral value into a vector and then shifting
9982 the initial value into element 0.
9983 (get_initial_defs_for_reduction): Replace the code argument with
9984 the neutral value calculated by neutral_op_for_slp_reduction.
9985 Use gimple_build_vector for constant-length vectors.
9986 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9987 but the first group_size elements have a neutral value.
9988 Use duplicate_and_interleave otherwise.
9989 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9990 Update call to get_initial_defs_for_reduction. Handle SLP
9991 reductions for variable-length vectors by creating one vector
9992 result for each scalar result, with the elements associated
9993 with other scalar results stubbed out with the neutral value.
9994 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9995 Require IFN_VEC_SHL_INSERT for double reductions on
9996 variable-length vectors, or SLP reductions that have
9997 a neutral value. Require can_duplicate_and_interleave_p
9998 support for variable-length unchained SLP reductions if there
9999 is no neutral value, such as for MIN/MAX reductions. Also require
10000 the number of vector elements to be a multiple of the number of
10001 SLP statements when doing variable-length unchained SLP reductions.
10002 Update call to vect_create_epilog_for_reduction.
10003 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10004 and remove initial values.
10005 (duplicate_and_interleave): Make public.
10006 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10007 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10009 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10010 Alan Hayward <alan.hayward@arm.com>
10011 David Sherwood <david.sherwood@arm.com>
10013 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10014 (can_duplicate_and_interleave_p): New function.
10015 (vect_get_and_check_slp_defs): Take the vector of statements
10016 rather than just the current one. Remove excess parentheses.
10017 Restriction rejectinon of vect_constant_def and vect_external_def
10018 for variable-length vectors to boolean types, or types for which
10019 can_duplicate_and_interleave_p is false.
10020 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10021 (duplicate_and_interleave): New function.
10022 (vect_get_constant_vectors): Use gimple_build_vector for
10023 constant-length vectors and suitable variable-length constant
10024 vectors. Use duplicate_and_interleave for other variable-length
10025 vectors. Don't defer the update when inserting new statements.
10027 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10028 Alan Hayward <alan.hayward@arm.com>
10029 David Sherwood <david.sherwood@arm.com>
10031 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10032 min_profitable_iters doesn't go negative.
10034 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10035 Alan Hayward <alan.hayward@arm.com>
10036 David Sherwood <david.sherwood@arm.com>
10038 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10039 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10040 * optabs.def (vec_mask_load_lanes_optab): New optab.
10041 (vec_mask_store_lanes_optab): Likewise.
10042 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10043 (MASK_STORE_LANES): Likewise.
10044 * internal-fn.c (mask_load_lanes_direct): New macro.
10045 (mask_store_lanes_direct): Likewise.
10046 (expand_mask_load_optab_fn): Handle masked operations.
10047 (expand_mask_load_lanes_optab_fn): New macro.
10048 (expand_mask_store_optab_fn): Handle masked operations.
10049 (expand_mask_store_lanes_optab_fn): New macro.
10050 (direct_mask_load_lanes_optab_supported_p): Likewise.
10051 (direct_mask_store_lanes_optab_supported_p): Likewise.
10052 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10054 (vect_load_lanes_supported): Likewise.
10055 * tree-vect-data-refs.c (strip_conversion): New function.
10056 (can_group_stmts_p): Likewise.
10057 (vect_analyze_data_ref_accesses): Use it instead of checking
10058 for a pair of assignments.
10059 (vect_store_lanes_supported): Take a masked_p parameter.
10060 (vect_load_lanes_supported): Likewise.
10061 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10062 vect_store_lanes_supported and vect_load_lanes_supported.
10063 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10064 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10065 parameter. Don't allow gaps for masked accesses.
10066 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10067 and vect_load_lanes_supported.
10068 (get_load_store_type): Take a masked_p parameter and update
10069 call to get_group_load_store_type.
10070 (vectorizable_store): Update call to get_load_store_type.
10071 Handle IFN_MASK_STORE_LANES.
10072 (vectorizable_load): Update call to get_load_store_type.
10073 Handle IFN_MASK_LOAD_LANES.
10075 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10076 Alan Hayward <alan.hayward@arm.com>
10077 David Sherwood <david.sherwood@arm.com>
10079 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10081 * config/aarch64/aarch64-protos.h
10082 (aarch64_sve_struct_memory_operand_p): Declare.
10083 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10084 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10085 (VPRED, vpred): Handle SVE structure modes.
10086 * config/aarch64/constraints.md (Utx): New constraint.
10087 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10088 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10089 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10090 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10091 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10092 structure modes. Split into pieces after RA.
10093 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10094 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10096 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10097 SVE structure modes.
10098 (aarch64_classify_address): Likewise.
10099 (sizetochar): Move earlier in file.
10100 (aarch64_print_operand): Handle SVE register lists.
10101 (aarch64_array_mode): New function.
10102 (aarch64_sve_struct_memory_operand_p): Likewise.
10103 (TARGET_ARRAY_MODE): Redefine.
10105 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10106 Alan Hayward <alan.hayward@arm.com>
10107 David Sherwood <david.sherwood@arm.com>
10109 * target.def (array_mode): New target hook.
10110 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10111 * doc/tm.texi: Regenerate.
10112 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10113 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10114 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10115 targetm.array_mode.
10116 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10119 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10120 Alan Hayward <alan.hayward@arm.com>
10121 David Sherwood <david.sherwood@arm.com>
10123 * fold-const.c (fold_binary_loc): Check the argument types
10124 rather than the result type when testing for a vector operation.
10126 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10128 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10129 * doc/tm.texi: Regenerate.
10131 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10132 Alan Hayward <alan.hayward@arm.com>
10133 David Sherwood <david.sherwood@arm.com>
10135 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10136 (sve): Document new AArch64 extension.
10137 * doc/md.texi (w): Extend the description of the AArch64
10138 constraint to include SVE vectors.
10139 (Upl, Upa): Document new AArch64 predicate constraints.
10140 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10142 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10143 (msve-vector-bits=): New option.
10144 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10145 SVE when these are disabled.
10146 (sve): New extension.
10147 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10148 modes. Adjust their number of units based on aarch64_sve_vg.
10149 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10150 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10151 aarch64_addr_query_type.
10152 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10153 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10154 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10155 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10156 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10157 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10158 (aarch64_simd_imm_zero_p): Delete.
10159 (aarch64_check_zero_based_sve_index_immediate): Declare.
10160 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10161 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10162 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10163 (aarch64_sve_float_mul_immediate_p): Likewise.
10164 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10165 rather than an rtx.
10166 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10167 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10168 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10169 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10170 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10171 (aarch64_regmode_natural_size): Likewise.
10172 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10173 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10175 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10176 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10177 for VG and the SVE predicate registers.
10178 (V_ALIASES): Add a "z"-prefixed alias.
10179 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10180 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10181 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10182 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10183 (REG_CLASS_NAMES): Add entries for them.
10184 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10185 and the predicate registers.
10186 (aarch64_sve_vg): Declare.
10187 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10188 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10189 (REGMODE_NATURAL_SIZE): Define.
10190 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10192 * config/aarch64/aarch64.c: Include cfgrtl.h.
10193 (simd_immediate_info): Add a constructor for series vectors,
10194 and an associated step field.
10195 (aarch64_sve_vg): New variable.
10196 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10197 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10198 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10199 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10200 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10201 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10202 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10203 (aarch64_get_mask_mode): New functions.
10204 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10205 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10206 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10207 predicate modes and predicate registers. Explicitly restrict
10208 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10209 to store a vector mode if it is recognized by
10210 aarch64_classify_vector_mode.
10211 (aarch64_regmode_natural_size): New function.
10212 (aarch64_hard_regno_caller_save_mode): Return the original mode
10214 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10215 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10216 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10217 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10219 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10220 does not overlap dest if the function is frame-related. Handle
10222 (aarch64_split_add_offset): New function.
10223 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10224 them aarch64_add_offset.
10225 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10226 and update call to aarch64_sub_sp.
10227 (aarch64_add_cfa_expression): New function.
10228 (aarch64_expand_prologue): Pass extra temporary registers to the
10229 functions above. Handle the case in which we need to emit new
10230 DW_CFA_expressions for registers that were originally saved
10231 relative to the stack pointer, but now have to be expressed
10232 relative to the frame pointer.
10233 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10235 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10236 IP0 and IP1 values for SVE frames.
10237 (aarch64_expand_vec_series): New function.
10238 (aarch64_expand_sve_widened_duplicate): Likewise.
10239 (aarch64_expand_sve_const_vector): Likewise.
10240 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10241 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10242 into the register, rather than emitting a SET directly.
10243 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10244 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10245 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10246 (offset_9bit_signed_scaled_p): New functions.
10247 (aarch64_replicate_bitmask_imm): New function.
10248 (aarch64_bitmask_imm): Use it.
10249 (aarch64_cannot_force_const_mem): Reject expressions involving
10250 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10251 (aarch64_classify_index): Handle SVE indices, by requiring
10252 a plain register index with a scale that matches the element size.
10253 (aarch64_classify_address): Handle SVE addresses. Assert that
10254 the mode of the address is VOIDmode or an integer mode.
10255 Update call to aarch64_classify_symbol.
10256 (aarch64_classify_symbolic_expression): Update call to
10257 aarch64_classify_symbol.
10258 (aarch64_const_vec_all_in_range_p): New function.
10259 (aarch64_print_vector_float_operand): Likewise.
10260 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10261 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10262 and the FP immediates 1.0 and 0.5.
10263 (aarch64_print_address_internal): Handle SVE addresses.
10264 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10265 (aarch64_regno_regclass): Handle predicate registers.
10266 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10268 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10269 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10270 (aarch64_convert_sve_vector_bits): New function.
10271 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10272 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10273 rather than an rtx.
10274 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10275 Handle SVE vector and predicate modes. Accept VL-based constants
10276 that need only one temporary register, and VL offsets that require
10277 no temporary registers.
10278 (aarch64_conditional_register_usage): Mark the predicate registers
10279 as fixed if SVE isn't available.
10280 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10281 Return true for SVE vector and predicate modes.
10282 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10283 rather than an unsigned int. Handle SVE modes.
10284 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10286 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10288 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10289 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10290 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10291 (aarch64_sve_float_mul_immediate_p): New functions.
10292 (aarch64_sve_valid_immediate): New function.
10293 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10294 Explicitly reject structure modes. Check for INDEX constants.
10295 Handle PTRUE and PFALSE constants.
10296 (aarch64_check_zero_based_sve_index_immediate): New function.
10297 (aarch64_simd_imm_zero_p): Delete.
10298 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10299 vector modes. Accept constants in the range of CNT[BHWD].
10300 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10301 ask for an Advanced SIMD mode.
10302 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10303 (aarch64_simd_vector_alignment): Handle SVE predicates.
10304 (aarch64_vectorize_preferred_vector_alignment): New function.
10305 (aarch64_simd_vector_alignment_reachable): Use it instead of
10307 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10308 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10310 (MAX_VECT_LEN): Delete.
10311 (expand_vec_perm_d): Add a vec_flags field.
10312 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10313 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10314 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10316 (aarch64_evpc_rev): Rename to...
10317 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10318 (aarch64_evpc_rev_global): New function.
10319 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10320 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10322 (aarch64_evpc_sve_tbl): New function.
10323 (aarch64_expand_vec_perm_const_1): Update after rename of
10324 aarch64_evpc_rev. Handle SVE permutes too, trying
10325 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10326 than aarch64_evpc_tbl.
10327 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10328 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10329 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10330 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10331 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10332 (aarch64_expand_sve_vcond): New functions.
10333 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10334 of aarch64_vector_mode_p.
10335 (aarch64_dwarf_poly_indeterminate_value): New function.
10336 (aarch64_compute_pressure_classes): Likewise.
10337 (aarch64_can_change_mode_class): Likewise.
10338 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10339 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10340 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10341 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10342 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10343 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10344 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10345 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10347 (Dn, Dl, Dr): Accept const as well as const_vector.
10348 (Dz): Likewise. Compare against CONST0_RTX.
10349 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10350 of "vector" where appropriate.
10351 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10352 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10353 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10354 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10355 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10356 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10357 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10358 (v_int_equiv): Extend to SVE modes.
10359 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10361 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10362 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10363 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10364 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10365 (SVE_COND_FP_CMP): New int iterators.
10366 (perm_hilo): Handle the new unpack unspecs.
10367 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10369 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10370 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10371 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10372 (aarch64_equality_operator, aarch64_constant_vector_operand)
10373 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10374 (aarch64_sve_nonimmediate_operand): Likewise.
10375 (aarch64_sve_general_operand): Likewise.
10376 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10377 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10378 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10379 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10380 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10381 (aarch64_sve_float_arith_immediate): Likewise.
10382 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10383 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10384 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10385 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10386 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10387 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10388 (aarch64_sve_float_arith_operand): Likewise.
10389 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10390 (aarch64_sve_float_mul_operand): Likewise.
10391 (aarch64_sve_vec_perm_operand): Likewise.
10392 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10393 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10394 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10395 as well as const_vector.
10396 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10397 in file. Use CONST0_RTX and CONSTM1_RTX.
10398 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10399 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10400 Use aarch64_simd_imm_zero.
10401 * config/aarch64/aarch64-sve.md: New file.
10402 * config/aarch64/aarch64.md: Include it.
10403 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10404 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10405 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10406 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10407 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10408 (sve): New attribute.
10409 (enabled): Disable instructions with the sve attribute unless
10411 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10412 aarch64_expand_mov_immediate.
10413 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10414 CNT[BHSD] immediates.
10415 (movti): Split CONST_POLY_INT moves into two halves.
10416 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10417 Split additions that need a temporary here if the destination
10418 is the stack pointer.
10419 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10420 (*add<mode>3_poly_1): New instruction.
10421 (set_clobber_cc): New expander.
10423 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10425 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10426 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10427 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10428 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10429 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10430 Change innermode from fixed_mode_size to machine_mode.
10431 (simplify_subreg): Update call accordingly. Handle a constant-sized
10432 subreg of a variable-length CONST_VECTOR.
10434 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10435 Alan Hayward <alan.hayward@arm.com>
10436 David Sherwood <david.sherwood@arm.com>
10438 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10439 (add_offset_to_base): New function, split out from...
10440 (create_mem_ref): ...here. When handling a scale other than 1,
10441 check first whether the address is valid without the offset.
10442 Add it into the base if so, leaving the index and scale as-is.
10444 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10447 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10448 fold_for_warn before checking if arg2 is INTEGER_CST.
10450 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10452 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10453 (store_multiple_operation): Delete.
10454 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10455 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10456 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10457 guarded by TARGET_STRING.
10458 (rs6000_output_load_multiple): Delete.
10459 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10460 OPTION_MASK_STRING / TARGET_STRING handling.
10461 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10462 (const rs6000_opt_masks) <"string">: Change mask to 0.
10463 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10464 (MASK_STRING): Delete.
10465 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10467 (load_multiple): Delete.
10474 (store_multiple): Delete.
10481 (movmemsi_8reg): Delete.
10482 (corresponding unnamed define_insn): Delete.
10483 (movmemsi_6reg): Delete.
10484 (corresponding unnamed define_insn): Delete.
10485 (movmemsi_4reg): Delete.
10486 (corresponding unnamed define_insn): Delete.
10487 (movmemsi_2reg): Delete.
10488 (corresponding unnamed define_insn): Delete.
10489 (movmemsi_1reg): Delete.
10490 (corresponding unnamed define_insn): Delete.
10491 * config/rs6000/rs6000.opt (mno-string): New.
10492 (mstring): Replace by deprecation warning stub.
10493 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10495 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10497 * regrename.c (regrename_do_replace): If replacing the same
10498 reg multiple times, try to reuse last created gen_raw_REG.
10501 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10502 main to workaround a bug in GDB.
10504 2018-01-12 Tom de Vries <tom@codesourcery.com>
10507 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10509 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10511 PR rtl-optimization/80481
10512 * ira-color.c (get_cap_member): New function.
10513 (allocnos_conflict_by_live_ranges_p): Use it.
10514 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10515 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10517 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10520 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10521 (*saddl_se_1): Ditto.
10522 (*ssubsi_1): Ditto.
10523 (*ssubl_se_1): Ditto.
10525 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10527 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10528 rather than wi::to_widest for DR_INITs.
10529 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10530 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10531 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10533 (vect_analyze_group_access_1): Note that here.
10535 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10537 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10538 polynomial type sizes.
10540 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10542 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10543 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10544 (gimple_add_tmp_var): Likewise.
10546 2018-01-12 Martin Liska <mliska@suse.cz>
10548 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10549 (gimple_alloc_sizes): Likewise.
10550 (dump_gimple_statistics): Use PRIu64 in printf format.
10551 * gimple.h: Change uint64_t to int.
10553 2018-01-12 Martin Liska <mliska@suse.cz>
10555 * tree-core.h: Use uint64_t instead of int.
10556 * tree.c (tree_node_counts): Likewise.
10557 (tree_node_sizes): Likewise.
10558 (dump_tree_statistics): Use PRIu64 in printf format.
10560 2018-01-12 Martin Liska <mliska@suse.cz>
10562 * Makefile.in: As qsort_chk is implemented in vec.c, add
10563 vec.o to linkage of gencfn-macros.
10564 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10565 passing the info to record_node_allocation_statistics.
10566 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10568 * ggc-common.c (struct ggc_usage): Add operator== and use
10569 it in operator< and compare function.
10570 * mem-stats.h (struct mem_usage): Likewise.
10571 * vec.c (struct vec_usage): Remove operator< and compare
10572 function. Can be simply inherited.
10574 2018-01-12 Martin Jambor <mjambor@suse.cz>
10577 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10578 * tree-ssa-math-opts.c: Include domwalk.h.
10579 (convert_mult_to_fma_1): New function.
10580 (fma_transformation_info): New type.
10581 (fma_deferring_state): Likewise.
10582 (cancel_fma_deferring): New function.
10583 (result_of_phi): Likewise.
10584 (last_fma_candidate_feeds_initial_phi): Likewise.
10585 (convert_mult_to_fma): Added deferring logic, split actual
10586 transformation to convert_mult_to_fma_1.
10587 (math_opts_dom_walker): New type.
10588 (math_opts_dom_walker::after_dom_children): New method, body moved
10589 here from pass_optimize_widening_mul::execute, added deferring logic
10591 (pass_optimize_widening_mul::execute): Moved most of code to
10592 math_opts_dom_walker::after_dom_children.
10593 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10594 * config/i386/i386.c (ix86_option_override_internal): Added
10595 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10597 2018-01-12 Richard Biener <rguenther@suse.de>
10600 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10601 inline instance vars.
10603 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10606 * config/rx/rx.c (rx_is_restricted_memory_address):
10607 Handle SUBREG case.
10609 2018-01-12 Richard Biener <rguenther@suse.de>
10611 PR tree-optimization/80846
10612 * target.def (split_reduction): New target hook.
10613 * targhooks.c (default_split_reduction): New function.
10614 * targhooks.h (default_split_reduction): Declare.
10615 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10616 target requests first reduce vectors by combining low and high
10618 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10619 (get_vectype_for_scalar_type_and_size): Export.
10620 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10621 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10622 * doc/tm.texi: Regenerate.
10623 * config/i386/i386.c (ix86_split_reduction): Implement
10624 TARGET_VECTORIZE_SPLIT_REDUCTION.
10626 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10629 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10630 in PIC mode except for TARGET_VXWORKS_RTP.
10631 * config/sparc/sparc.c: Include cfgrtl.h.
10632 (TARGET_INIT_PIC_REG): Define.
10633 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10634 (sparc_pic_register_p): New predicate.
10635 (sparc_legitimate_address_p): Use it.
10636 (sparc_legitimize_pic_address): Likewise.
10637 (sparc_delegitimize_address): Likewise.
10638 (sparc_mode_dependent_address_p): Likewise.
10639 (gen_load_pcrel_sym): Remove 4th parameter.
10640 (load_got_register): Adjust call to above. Remove obsolete stuff.
10641 (sparc_expand_prologue): Do not call load_got_register here.
10642 (sparc_flat_expand_prologue): Likewise.
10643 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10644 (sparc_use_pseudo_pic_reg): New function.
10645 (sparc_init_pic_reg): Likewise.
10646 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10647 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10649 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10651 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10652 Add item for branch_cost.
10654 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10656 PR rtl-optimization/83565
10657 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10658 not extend the result to a larger mode for rotate operations.
10659 (num_sign_bit_copies1): Likewise.
10661 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10664 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10666 Use values-Xc.o for -pedantic.
10667 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10669 2018-01-12 Martin Liska <mliska@suse.cz>
10672 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10674 (possible_polymorphic_call_targets): Use it.
10675 (ipa_devirt): Likewise.
10677 2018-01-12 Martin Liska <mliska@suse.cz>
10679 * profile-count.h (enum profile_quality): Use 0 as invalid
10680 enum value of profile_quality.
10682 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10684 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10685 -mext-string options.
10687 2018-01-12 Richard Biener <rguenther@suse.de>
10689 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10690 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10691 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10693 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10695 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10697 * configure.ac (--with-long-double-format): Add support for the
10698 configuration option to change the default long double format on
10700 * config.gcc (powerpc*-linux*-*): Likewise.
10701 * configure: Regenerate.
10702 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10703 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10704 used without modification.
10706 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10708 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10709 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10710 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10711 MISC_BUILTIN_SPEC_BARRIER.
10712 (rs6000_init_builtins): Likewise.
10713 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10715 (speculation_barrier): New define_insn.
10716 * doc/extend.texi: Document __builtin_speculation_barrier.
10718 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10721 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10722 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10723 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10725 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10726 integral modes instead of "ss" and "sd".
10727 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10728 vectors with 32-bit and 64-bit elements.
10729 (vecdupssescalarmodesuffix): New mode attribute.
10730 (vec_dup<mode>): Use it.
10732 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10735 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10736 frame if argument is passed on stack.
10738 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10741 * ree.c (combine_reaching_defs): Optimize also
10742 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10743 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10745 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10747 PR middle-end/83189
10748 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10750 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10752 PR middle-end/83718
10753 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10754 after they are computed.
10756 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10758 PR tree-optimization/83695
10759 * gimple-loop-linterchange.cc
10760 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10761 reset cached scev information after interchange.
10762 (pass_linterchange::execute): Remove call to scev_reset_htab.
10764 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10766 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10767 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10768 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10769 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10770 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10771 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10772 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10773 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10774 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10775 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10776 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10777 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10778 (V_lane_reg): Likewise.
10779 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10781 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10782 (vfmal_lane_low<mode>_intrinsic,
10783 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10784 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10785 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10786 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10787 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10788 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10790 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10792 * config/arm/arm-cpus.in (fp16fml): New feature.
10793 (ALL_SIMD): Add fp16fml.
10794 (armv8.2-a): Add fp16fml as an option.
10795 (armv8.3-a): Likewise.
10796 (armv8.4-a): Add fp16fml as part of fp16.
10797 * config/arm/arm.h (TARGET_FP16FML): Define.
10798 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10800 * config/arm/arm-modes.def (V2HF): Define.
10801 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10802 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10803 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10804 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10805 vfmsl_low, vfmsl_high): New set of builtins.
10806 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10807 (vfml_op): New code attribute.
10808 (VFMLHALVES): New int iterator.
10809 (VFML, VFMLSEL): New mode attributes.
10810 (V_reg): Define mapping for V2HF.
10811 (V_hi, V_lo): New mode attributes.
10812 (VF_constraint): Likewise.
10813 (vfml_half, vfml_half_selector): New int attributes.
10814 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10816 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10817 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10819 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10820 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10821 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10822 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10824 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10825 Document new effective target and option set.
10827 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10829 * config/arm/arm-cpus.in (armv8_4): New feature.
10830 (ARMv8_4a): New fgroup.
10831 (armv8.4-a): New arch.
10832 * config/arm/arm-tables.opt: Regenerate.
10833 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10834 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10835 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10836 Add matching rules for -march=armv8.4-a and extensions.
10837 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10839 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10842 * config/rx/rx.md (BW): New mode attribute.
10843 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10845 2018-01-11 Richard Biener <rguenther@suse.de>
10847 PR tree-optimization/83435
10848 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10849 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10850 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
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.c (aarch64_address_info): Add a const_offset
10858 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10859 (aarch64_print_address_internal): Use it to check for a zero offset.
10861 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10862 Alan Hayward <alan.hayward@arm.com>
10863 David Sherwood <david.sherwood@arm.com>
10865 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10866 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10867 Return a poly_int64 rather than a HOST_WIDE_INT.
10868 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10869 rather than a HOST_WIDE_INT.
10870 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10871 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10872 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10873 final_offset from HOST_WIDE_INT to poly_int64.
10874 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10875 to_constant when getting the number of units in an Advanced SIMD
10877 (aarch64_builtin_vectorized_function): Check for a constant number
10879 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10881 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10882 attribute instead of GET_MODE_NUNITS.
10883 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10884 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10885 GET_MODE_SIZE for fixed-size registers.
10886 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10887 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10888 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10889 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10890 (aarch64_print_operand, aarch64_print_address_internal)
10891 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10892 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10893 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10894 Handle polynomial GET_MODE_SIZE.
10895 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10896 wider than SImode without modification.
10897 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10898 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10899 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10900 passing and returning SVE modes.
10901 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10902 rather than GEN_INT.
10903 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10904 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10905 (aarch64_allocate_and_probe_stack_space): Likewise.
10906 (aarch64_layout_frame): Cope with polynomial offsets.
10907 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10908 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10909 polynomial offsets.
10910 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10911 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10912 poly_int64 rather than a HOST_WIDE_INT.
10913 (aarch64_get_separate_components, aarch64_process_components)
10914 (aarch64_expand_prologue, aarch64_expand_epilogue)
10915 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10916 (aarch64_anchor_offset): New function, split out from...
10917 (aarch64_legitimize_address): ...here.
10918 (aarch64_builtin_vectorization_cost): Handle polynomial
10919 TYPE_VECTOR_SUBPARTS.
10920 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10922 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10923 number of elements from the PARALLEL rather than the mode.
10924 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10925 rather than GET_MODE_BITSIZE.
10926 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10927 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10928 (aarch64_expand_vec_perm_const_1): Handle polynomial
10929 d->perm.length () and d->perm elements.
10930 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10931 Apply to_constant to d->perm elements.
10932 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10933 polynomial CONST_VECTOR_NUNITS.
10934 (aarch64_move_pointer): Take amount as a poly_int64 rather
10936 (aarch64_progress_pointer): Avoid temporary variable.
10937 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10938 the mode attribute instead of GET_MODE.
10940 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10941 Alan Hayward <alan.hayward@arm.com>
10942 David Sherwood <david.sherwood@arm.com>
10944 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10945 x exists before using it.
10946 (aarch64_add_constant_internal): Rename to...
10947 (aarch64_add_offset_1): ...this. Replace regnum with separate
10948 src and dest rtxes. Handle the case in which they're different,
10949 including when the offset is zero. Replace scratchreg with an rtx.
10950 Use 2 additions if there is no spare register into which we can
10951 move a 16-bit constant.
10952 (aarch64_add_constant): Delete.
10953 (aarch64_add_offset): Replace reg with separate src and dest
10954 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10955 Use aarch64_add_offset_1.
10956 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10957 an rtx rather than an int. Take the delta as a poly_int64
10958 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10959 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10960 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10961 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10962 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10963 and aarch64_add_sp.
10964 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10965 aarch64_add_constant.
10967 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10969 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10970 Use scalar_float_mode.
10972 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10974 * config/aarch64/aarch64-simd.md
10975 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10976 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10977 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10978 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10979 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10980 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10981 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10982 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10983 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10984 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10986 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10989 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10990 targ_options->x_arm_arch_string is non NULL.
10992 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10994 * config/aarch64/aarch64.h
10995 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10997 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11000 * expmed.c (emit_store_flag_force): Swap if const op0
11001 and change VOIDmode to mode of op0.
11003 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11005 PR rtl-optimization/83761
11006 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11007 than bytes to mode_for_size.
11009 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11011 PR middle-end/83189
11012 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11013 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11016 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11018 PR middle-end/83575
11019 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11020 when in layout mode.
11021 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11022 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11025 2018-01-10 Michael Collison <michael.collison@arm.com>
11027 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11028 * config/aarch64/aarch64-option-extension.def: Add
11029 AARCH64_OPT_EXTENSION of 'fp16fml'.
11030 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11031 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11032 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11033 * config/aarch64/constraints.md (Ui7): New constraint.
11034 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11035 (VFMLA_SEL_W): Ditto.
11038 (VFMLA16_LOW): New int iterator.
11039 (VFMLA16_HIGH): Ditto.
11040 (UNSPEC_FMLAL): New unspec.
11041 (UNSPEC_FMLSL): Ditto.
11042 (UNSPEC_FMLAL2): Ditto.
11043 (UNSPEC_FMLSL2): Ditto.
11044 (f16mac): New code attribute.
11045 * config/aarch64/aarch64-simd-builtins.def
11046 (aarch64_fmlal_lowv2sf): Ditto.
11047 (aarch64_fmlsl_lowv2sf): Ditto.
11048 (aarch64_fmlalq_lowv4sf): Ditto.
11049 (aarch64_fmlslq_lowv4sf): Ditto.
11050 (aarch64_fmlal_highv2sf): Ditto.
11051 (aarch64_fmlsl_highv2sf): Ditto.
11052 (aarch64_fmlalq_highv4sf): Ditto.
11053 (aarch64_fmlslq_highv4sf): Ditto.
11054 (aarch64_fmlal_lane_lowv2sf): Ditto.
11055 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11056 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11057 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11058 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11059 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11060 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11061 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11062 (aarch64_fmlal_lane_highv2sf): Ditto.
11063 (aarch64_fmlsl_lane_highv2sf): Ditto.
11064 (aarch64_fmlal_laneq_highv2sf): Ditto.
11065 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11066 (aarch64_fmlalq_lane_highv4sf): Ditto.
11067 (aarch64_fmlsl_lane_highv4sf): Ditto.
11068 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11069 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11070 * config/aarch64/aarch64-simd.md:
11071 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11072 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11073 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11074 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11075 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11076 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11077 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11078 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11079 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11080 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11081 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11082 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11083 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11084 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11085 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11086 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11087 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11088 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11089 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11090 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11091 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11092 (vfmlsl_low_u32): Ditto.
11093 (vfmlalq_low_u32): Ditto.
11094 (vfmlslq_low_u32): Ditto.
11095 (vfmlal_high_u32): Ditto.
11096 (vfmlsl_high_u32): Ditto.
11097 (vfmlalq_high_u32): Ditto.
11098 (vfmlslq_high_u32): Ditto.
11099 (vfmlal_lane_low_u32): Ditto.
11100 (vfmlsl_lane_low_u32): Ditto.
11101 (vfmlal_laneq_low_u32): Ditto.
11102 (vfmlsl_laneq_low_u32): Ditto.
11103 (vfmlalq_lane_low_u32): Ditto.
11104 (vfmlslq_lane_low_u32): Ditto.
11105 (vfmlalq_laneq_low_u32): Ditto.
11106 (vfmlslq_laneq_low_u32): Ditto.
11107 (vfmlal_lane_high_u32): Ditto.
11108 (vfmlsl_lane_high_u32): Ditto.
11109 (vfmlal_laneq_high_u32): Ditto.
11110 (vfmlsl_laneq_high_u32): Ditto.
11111 (vfmlalq_lane_high_u32): Ditto.
11112 (vfmlslq_lane_high_u32): Ditto.
11113 (vfmlalq_laneq_high_u32): Ditto.
11114 (vfmlslq_laneq_high_u32): Ditto.
11115 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11116 (AARCH64_FL_FOR_ARCH8_4): New.
11117 (AARCH64_ISA_F16FML): New ISA flag.
11118 (TARGET_F16FML): New feature flag for fp16fml.
11119 (doc/invoke.texi): Document new fp16fml option.
11121 2018-01-10 Michael Collison <michael.collison@arm.com>
11123 * config/aarch64/aarch64-builtins.c:
11124 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11125 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11126 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11127 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11128 (AARCH64_ISA_SHA3): New ISA flag.
11129 (TARGET_SHA3): New feature flag for sha3.
11130 * config/aarch64/iterators.md (sha512_op): New int attribute.
11131 (CRYPTO_SHA512): New int iterator.
11132 (UNSPEC_SHA512H): New unspec.
11133 (UNSPEC_SHA512H2): Ditto.
11134 (UNSPEC_SHA512SU0): Ditto.
11135 (UNSPEC_SHA512SU1): Ditto.
11136 * config/aarch64/aarch64-simd-builtins.def
11137 (aarch64_crypto_sha512hqv2di): New builtin.
11138 (aarch64_crypto_sha512h2qv2di): Ditto.
11139 (aarch64_crypto_sha512su0qv2di): Ditto.
11140 (aarch64_crypto_sha512su1qv2di): Ditto.
11141 (aarch64_eor3qv8hi): Ditto.
11142 (aarch64_rax1qv2di): Ditto.
11143 (aarch64_xarqv2di): Ditto.
11144 (aarch64_bcaxqv8hi): Ditto.
11145 * config/aarch64/aarch64-simd.md:
11146 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11147 (aarch64_crypto_sha512su0qv2di): Ditto.
11148 (aarch64_crypto_sha512su1qv2di): Ditto.
11149 (aarch64_eor3qv8hi): Ditto.
11150 (aarch64_rax1qv2di): Ditto.
11151 (aarch64_xarqv2di): Ditto.
11152 (aarch64_bcaxqv8hi): Ditto.
11153 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11154 (vsha512h2q_u64): Ditto.
11155 (vsha512su0q_u64): Ditto.
11156 (vsha512su1q_u64): Ditto.
11157 (veor3q_u16): Ditto.
11158 (vrax1q_u64): Ditto.
11159 (vxarq_u64): Ditto.
11160 (vbcaxq_u16): Ditto.
11161 * config/arm/types.md (crypto_sha512): New type attribute.
11162 (crypto_sha3): Ditto.
11163 (doc/invoke.texi): Document new sha3 option.
11165 2018-01-10 Michael Collison <michael.collison@arm.com>
11167 * config/aarch64/aarch64-builtins.c:
11168 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11169 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11170 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11171 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11172 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11173 (AARCH64_ISA_SM4): New ISA flag.
11174 (TARGET_SM4): New feature flag for sm4.
11175 * config/aarch64/aarch64-simd-builtins.def
11176 (aarch64_sm3ss1qv4si): Ditto.
11177 (aarch64_sm3tt1aq4si): Ditto.
11178 (aarch64_sm3tt1bq4si): Ditto.
11179 (aarch64_sm3tt2aq4si): Ditto.
11180 (aarch64_sm3tt2bq4si): Ditto.
11181 (aarch64_sm3partw1qv4si): Ditto.
11182 (aarch64_sm3partw2qv4si): Ditto.
11183 (aarch64_sm4eqv4si): Ditto.
11184 (aarch64_sm4ekeyqv4si): Ditto.
11185 * config/aarch64/aarch64-simd.md:
11186 (aarch64_sm3ss1qv4si): Ditto.
11187 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11188 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11189 (aarch64_sm4eqv4si): Ditto.
11190 (aarch64_sm4ekeyqv4si): Ditto.
11191 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11192 (sm3part_op): Ditto.
11193 (CRYPTO_SM3TT): Ditto.
11194 (CRYPTO_SM3PART): Ditto.
11195 (UNSPEC_SM3SS1): New unspec.
11196 (UNSPEC_SM3TT1A): Ditto.
11197 (UNSPEC_SM3TT1B): Ditto.
11198 (UNSPEC_SM3TT2A): Ditto.
11199 (UNSPEC_SM3TT2B): Ditto.
11200 (UNSPEC_SM3PARTW1): Ditto.
11201 (UNSPEC_SM3PARTW2): Ditto.
11202 (UNSPEC_SM4E): Ditto.
11203 (UNSPEC_SM4EKEY): Ditto.
11204 * config/aarch64/constraints.md (Ui2): New constraint.
11205 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11206 * config/arm/types.md (crypto_sm3): New type attribute.
11207 (crypto_sm4): Ditto.
11208 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11209 (vsm3tt1aq_u32): Ditto.
11210 (vsm3tt1bq_u32): Ditto.
11211 (vsm3tt2aq_u32): Ditto.
11212 (vsm3tt2bq_u32): Ditto.
11213 (vsm3partw1q_u32): Ditto.
11214 (vsm3partw2q_u32): Ditto.
11215 (vsm4eq_u32): Ditto.
11216 (vsm4ekeyq_u32): Ditto.
11217 (doc/invoke.texi): Document new sm4 option.
11219 2018-01-10 Michael Collison <michael.collison@arm.com>
11221 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11222 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11223 (AARCH64_FL_FOR_ARCH8_4): New.
11224 (AARCH64_FL_V8_4): New flag.
11225 (doc/invoke.texi): Document new armv8.4-a option.
11227 2018-01-10 Michael Collison <michael.collison@arm.com>
11229 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11230 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11231 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11232 * config/aarch64/aarch64-option-extension.def: Add
11233 AARCH64_OPT_EXTENSION of 'sha2'.
11234 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11235 (crypto): Disable sha2 and aes if crypto disabled.
11236 (crypto): Enable aes and sha2 if enabled.
11237 (simd): Disable sha2 and aes if simd disabled.
11238 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11240 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11241 (TARGET_SHA2): New feature flag for sha2.
11242 (TARGET_AES): New feature flag for aes.
11243 * config/aarch64/aarch64-simd.md:
11244 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11245 conditional on TARGET_AES.
11246 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11247 (aarch64_crypto_sha1hsi): Make pattern conditional
11249 (aarch64_crypto_sha1hv4si): Ditto.
11250 (aarch64_be_crypto_sha1hv4si): Ditto.
11251 (aarch64_crypto_sha1su1v4si): Ditto.
11252 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11253 (aarch64_crypto_sha1su0v4si): Ditto.
11254 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11255 (aarch64_crypto_sha256su0v4si): Ditto.
11256 (aarch64_crypto_sha256su1v4si): Ditto.
11257 (doc/invoke.texi): Document new aes and sha2 options.
11259 2018-01-10 Martin Sebor <msebor@redhat.com>
11261 PR tree-optimization/83781
11262 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11265 2018-01-11 Martin Sebor <msebor@gmail.com>
11266 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11268 PR tree-optimization/83501
11269 PR tree-optimization/81703
11271 * tree-ssa-strlen.c (get_string_cst): Rename...
11272 (get_string_len): ...to this. Handle global constants.
11273 (handle_char_store): Adjust.
11275 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11276 Jim Wilson <jimw@sifive.com>
11278 * config/riscv/riscv-protos.h (riscv_output_return): New.
11279 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11280 (riscv_attribute_table, riscv_output_return),
11281 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11282 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11283 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11285 (riscv_expand_prologue): Add early return for naked function.
11286 (riscv_expand_epilogue): Likewise.
11287 (riscv_function_ok_for_sibcall): Return false for naked function.
11288 (riscv_set_current_function): New.
11289 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11290 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11291 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11292 * doc/extend.texi (RISC-V Function Attributes): New.
11294 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11296 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11297 check for 128-bit long double before checking TCmode.
11298 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11299 128-bit long doubles before checking TFmode or TCmode.
11300 (FLOAT128_IBM_P): Likewise.
11302 2018-01-10 Martin Sebor <msebor@redhat.com>
11304 PR tree-optimization/83671
11305 * builtins.c (c_strlen): Unconditionally return zero for the empty
11307 Use -Warray-bounds for warnings.
11308 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11309 for non-constant array indices with COMPONENT_REF, arrays of
11310 arrays, and pointers to arrays.
11311 (gimple_fold_builtin_strlen): Determine and set length range for
11312 non-constant character arrays.
11314 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11316 PR middle-end/81897
11317 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11320 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11322 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11324 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11327 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11328 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11329 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11330 indexed_or_indirect_operand predicate.
11331 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11332 (*vsx_le_perm_load_v8hi): Likewise.
11333 (*vsx_le_perm_load_v16qi): Likewise.
11334 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11335 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11336 (*vsx_le_perm_store_v8hi): Likewise.
11337 (*vsx_le_perm_store_v16qi): Likewise.
11338 (eight unnamed splitters): Likewise.
11340 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11342 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11343 * config/rs6000/emmintrin.h: Likewise.
11344 * config/rs6000/mmintrin.h: Likewise.
11345 * config/rs6000/xmmintrin.h: Likewise.
11347 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11350 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11352 * tree.c (tree_nop_conversion): Return true for location wrapper
11354 (maybe_wrap_with_location): New function.
11355 (selftest::check_strip_nops): New function.
11356 (selftest::test_location_wrappers): New function.
11357 (selftest::tree_c_tests): Call it.
11358 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11359 (maybe_wrap_with_location): New decl.
11360 (EXPR_LOCATION_WRAPPER_P): New macro.
11361 (location_wrapper_p): New inline function.
11362 (tree_strip_any_location_wrapper): New inline function.
11364 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11367 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11368 stack_realign_offset for the largest alignment of stack slot
11370 (ix86_find_max_used_stack_alignment): New function.
11371 (ix86_finalize_stack_frame_flags): Use it. Set
11372 max_used_stack_alignment if we don't realign stack.
11373 * config/i386/i386.h (machine_function): Add
11374 max_used_stack_alignment.
11376 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11378 * config/arm/arm.opt (-mbranch-cost): New option.
11379 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11382 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11385 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11386 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11388 2018-01-10 Richard Biener <rguenther@suse.de>
11391 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11392 early out so it also covers the case where we have a non-NULL
11395 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11397 PR tree-optimization/83753
11398 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11399 for non-strided grouped accesses if the number of elements is 1.
11401 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11404 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11405 * i386.h (TARGET_USE_GATHER): Define.
11406 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11408 2018-01-10 Martin Liska <mliska@suse.cz>
11411 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11412 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11414 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11415 CLEANUP_NO_PARTITIONING is not set.
11417 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11419 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11420 for vectors, as a partial revert of r254296.
11421 * rtl.h (const_vec_p): Delete.
11422 (const_vec_duplicate_p): Don't test for vector CONSTs.
11423 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11424 * expmed.c (make_tree): Likewise.
11427 * common.md (E, F): Use CONSTANT_P instead of checking for
11429 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11430 checking for CONST_VECTOR.
11432 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11434 PR middle-end/83575
11435 * predict.c (force_edge_cold): Handle in more sane way edges
11436 with no prediction.
11438 2018-01-09 Carl Love <cel@us.ibm.com>
11440 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11442 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11443 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11444 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11445 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11446 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11447 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11448 * config/rs6000/rs6000-protos.h: Add extern defition for
11449 rs6000_generate_float2_double_code.
11450 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11452 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11453 (float2_v2df): Add define_expand.
11455 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11458 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11459 op_mode in the force_to_mode call.
11461 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11463 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11464 instead of checking each element individually.
11465 (aarch64_evpc_uzp): Likewise.
11466 (aarch64_evpc_zip): Likewise.
11467 (aarch64_evpc_ext): Likewise.
11468 (aarch64_evpc_rev): Likewise.
11469 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11470 instead of checking each element individually. Return true without
11472 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11473 whether all selected elements come from the same input, instead of
11474 checking each element individually. Remove calls to gen_rtx_REG,
11475 start_sequence and end_sequence and instead assert that no rtl is
11478 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11480 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11481 order of HIGH and CONST checks.
11483 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11485 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11486 if the destination isn't an SSA_NAME.
11488 2018-01-09 Richard Biener <rguenther@suse.de>
11490 PR tree-optimization/83668
11491 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11493 (canonicalize_loop_form): ... here, renamed from ...
11494 (canonicalize_loop_closed_ssa_form): ... this and amended to
11495 swap successor edges for loop exit blocks to make us use
11496 the RPO order we need for initial schedule generation.
11498 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11500 PR tree-optimization/64811
11501 * match.pd: When optimizing comparisons with Inf, avoid
11502 introducing or losing exceptions from comparisons with NaN.
11504 2018-01-09 Martin Liska <mliska@suse.cz>
11507 * asan.c (shadow_mem_size): Add gcc_assert.
11509 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11511 Don't save registers in main().
11514 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11515 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11516 * config/avr/avr.c (avr_set_current_function): Don't error if
11517 naked, OS_task or OS_main are specified at the same time.
11518 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11520 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11522 * common/config/avr/avr-common.c (avr_option_optimization_table):
11523 Switch on -mmain-is-OS_task for optimizing compilations.
11525 2018-01-09 Richard Biener <rguenther@suse.de>
11527 PR tree-optimization/83572
11528 * graphite.c: Include cfganal.h.
11529 (graphite_transform_loops): Connect infinite loops to exit
11530 and remove fake edges at the end.
11532 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11534 * ipa-inline.c (edge_badness): Revert accidental checkin.
11536 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11539 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11540 symbols; not inline clones.
11542 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11545 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11546 hard registers. Formatting fixes.
11548 PR preprocessor/83722
11549 * gcc.c (try_generate_repro): Pass
11550 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11551 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11554 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11555 Kito Cheng <kito.cheng@gmail.com>
11557 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11558 (riscv_leaf_function_p): Delete.
11559 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11561 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11563 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11565 (do_ifelse): New function.
11566 (do_isel): New function.
11567 (do_sub3): New function.
11568 (do_add3): New function.
11569 (do_load_mask_compare): New function.
11570 (do_overlap_load_compare): New function.
11571 (expand_compare_loop): New function.
11572 (expand_block_compare): Call expand_compare_loop() when appropriate.
11573 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11574 option description.
11575 (-mblock-compare-inline-loop-limit): New option.
11577 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11580 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11581 Reverse order of second and third operands in first alternative.
11582 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11583 of first and second elements in UNSPEC_VPERMR vector.
11584 (altivec_expand_vec_perm_le): Likewise.
11586 2018-01-08 Jeff Law <law@redhat.com>
11588 PR rtl-optimizatin/81308
11589 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11590 (process_switch): If group_case_labels makes a change, then set
11592 (pass_convert_switch::execute): If a switch is converted, then
11593 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11595 PR rtl-optimization/81308
11596 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11599 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11601 PR target/83663 - Revert r255946
11602 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11603 generation for cases where splatting a value is not useful.
11604 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11605 across a vec_duplicate and a paradoxical subreg forming a vector
11606 mode to a vec_concat.
11608 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11610 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11611 -march=armv8.3-a variants.
11612 * config/arm/t-multilib: Likewise.
11613 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11615 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11617 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11619 (cceq_ior_compare_complement): Give it a name so I can use it, and
11620 change boolean_or_operator predicate to boolean_operator so it can
11621 be used to generate a crand.
11622 (eqne): New code iterator.
11623 (bd/bd_neg): New code_attrs.
11624 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11625 a single define_insn.
11626 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11627 decrement (bdnzt/bdnzf/bdzt/bdzf).
11628 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11629 with the new names of the branch decrement patterns, and added the
11630 names of the branch decrement conditional patterns.
11632 2018-01-08 Richard Biener <rguenther@suse.de>
11634 PR tree-optimization/83563
11635 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11638 2018-01-08 Richard Biener <rguenther@suse.de>
11640 PR middle-end/83713
11641 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11643 2018-01-08 Richard Biener <rguenther@suse.de>
11645 PR tree-optimization/83685
11646 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11647 references to abnormals.
11649 2018-01-08 Richard Biener <rguenther@suse.de>
11652 * dwarf2out.c (output_indirect_strings): Handle empty
11653 skeleton_debug_str_hash.
11654 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11656 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11658 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11659 (emit_store_direct): Likewise.
11660 (arc_trampoline_adjust_address): Likewise.
11661 (arc_asm_trampoline_template): New function.
11662 (arc_initialize_trampoline): Use asm_trampoline_template.
11663 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11664 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11665 * config/arc/arc.md (flush_icache): Delete pattern.
11667 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11669 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11670 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11673 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11676 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11677 by not USED_FOR_TARGET.
11678 (make_pass_resolve_sw_modes): Likewise.
11680 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11682 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11685 2018-01-08 Richard Biener <rguenther@suse.de>
11687 PR middle-end/83580
11688 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11690 2018-01-08 Richard Biener <rguenther@suse.de>
11692 PR middle-end/83517
11693 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11695 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11697 PR middle-end/81897
11698 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11699 basic blocks with a small number of successors.
11700 (convert_control_dep_chain_into_preds): Improve handling of
11702 (dump_predicates): Split apart into...
11703 (dump_pred_chain): ...here...
11704 (dump_pred_info): ...and here.
11705 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11706 (can_chain_union_be_invalidated_p): Improve check for invalidation
11708 (uninit_uses_cannot_happen): Avoid unnecessary if
11709 convert_control_dep_chain_into_preds yielded nothing.
11711 2018-01-06 Martin Sebor <msebor@redhat.com>
11713 PR tree-optimization/83640
11714 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11715 subtracting negative offset from size.
11716 (builtin_access::overlap): Adjust offset bounds of the access to fall
11717 within the size of the object if possible.
11719 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11721 PR rtl-optimization/83699
11722 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11723 extract_bit_field_as_subreg to cases in which the extracted
11724 value is also a vector.
11726 * lra-constraints.c (process_alt_operands): Test for the equivalence
11727 substitutions when detecting a possible reload cycle.
11729 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11732 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11733 by default if flag_selective_schedling{,2}. Formatting fixes.
11735 PR rtl-optimization/83682
11736 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11737 if it has non-VECTOR_MODE element mode.
11738 (vec_duplicate_p): Likewise.
11740 PR middle-end/83694
11741 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11742 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11744 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11747 * config/i386/i386-builtin.def
11748 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11749 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11750 Require also OPTION_MASK_ISA_AVX512F in addition to
11751 OPTION_MASK_ISA_GFNI.
11752 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11753 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11754 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11755 to OPTION_MASK_ISA_GFNI.
11756 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11757 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11758 OPTION_MASK_ISA_AVX512BW.
11759 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11760 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11761 addition to OPTION_MASK_ISA_GFNI.
11762 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11763 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11764 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11765 to OPTION_MASK_ISA_GFNI.
11766 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11767 a requirement for all ISAs rather than any of them with a few
11769 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11771 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11772 bitmasks to be enabled with 3 exceptions, instead of requiring any
11773 enabled ISA with lots of exceptions.
11774 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11775 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11776 Change avx512bw in isa attribute to avx512f.
11777 * config/i386/sgxintrin.h: Add license boilerplate.
11778 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11779 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11780 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11781 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11783 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11784 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11785 temporarily sse2 rather than sse if not enabled already.
11788 * config/i386/sse.md (VI248_VLBW): Rename to ...
11789 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11790 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11791 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11792 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11793 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11794 mode iterator instead of VI248_VLBW.
11796 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11798 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11799 (record_modified): Skip clobbers; add debug output.
11800 (param_change_prob): Use sreal frequencies.
11802 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11804 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11805 punt for user-aligned variables.
11807 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11809 * tree-chrec.c (chrec_contains_symbols): Return true for
11812 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11815 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11816 of (x|y) == x for BICS pattern.
11818 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11820 PR tree-optimization/83605
11821 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11822 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11825 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11827 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11828 * config/epiphany/rtems.h: New file.
11830 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11831 Uros Bizjak <ubizjak@gmail.com>
11834 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11835 QIreg_operand instead of register_operand predicate.
11836 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11837 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11838 comments instead of -fmitigate[-_]rop.
11840 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11843 * cgraphunit.c (symbol_table::compile): Switch to text_section
11844 before calling assembly_start debug hook.
11845 * run-rtl-passes.c (run_rtl_passes): Likewise.
11848 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11850 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11851 range_int_cst_p rather than !symbolic_range_p before calling
11852 extract_range_from_multiplicative_op_1.
11854 2018-01-04 Jeff Law <law@redhat.com>
11856 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11857 redundant test in assertion.
11859 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11861 * doc/rtl.texi: Document machine_mode wrapper classes.
11863 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11865 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11866 using tree_to_uhwi.
11868 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11870 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11871 the VEC_PERM_EXPR fold to fail.
11873 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11876 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11877 to switched_sections.
11879 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11882 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11883 test for d.testing.
11885 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11888 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11889 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11891 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11894 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11895 is BLKmode and bitpos not zero or mode change is needed.
11897 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11900 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11903 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11906 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11907 instead of MULT rtx. Update all corresponding splitters.
11908 (*saddl_se): Ditto.
11909 (*ssub<modesuffix>): Ditto.
11910 (*ssubl_se): Ditto.
11911 (*cmp_sadd_di): Update split patterns.
11912 (*cmp_sadd_si): Ditto.
11913 (*cmp_sadd_sidi): Ditto.
11914 (*cmp_ssub_di): Ditto.
11915 (*cmp_ssub_si): Ditto.
11916 (*cmp_ssub_sidi): Ditto.
11917 * config/alpha/predicates.md (const23_operand): New predicate.
11918 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11919 Look for ASHIFT, not MULT inner operand.
11920 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11922 2018-01-04 Martin Liska <mliska@suse.cz>
11924 PR gcov-profile/83669
11925 * gcov.c (output_intermediate_file): Add version to intermediate
11927 * doc/gcov.texi: Document new field 'version' in intermediate
11928 file format. Fix location of '-k' option of gcov command.
11930 2018-01-04 Martin Liska <mliska@suse.cz>
11933 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11935 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11937 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11939 2018-01-03 Martin Sebor <msebor@redhat.com>
11941 PR tree-optimization/83655
11942 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11943 checking calls with invalid arguments.
11945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11947 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11948 (vectorizable_mask_load_store): Delete.
11949 (vectorizable_call): Return false for masked loads and stores.
11950 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11951 instead of gimple_assign_rhs1.
11952 (vectorizable_load): Handle IFN_MASK_LOAD.
11953 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11957 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11959 (vectorizable_mask_load_store): ...here.
11960 (vectorizable_load): ...and here.
11962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11964 * tree-vect-stmts.c (vect_build_all_ones_mask)
11965 (vect_build_zero_merge_argument): New functions, split out from...
11966 (vectorizable_load): ...here.
11968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11970 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11972 (vectorizable_mask_load_store): ...here.
11973 (vectorizable_store): ...and here.
11975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11977 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11979 (vectorizable_mask_load_store): ...here.
11981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11983 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11984 (vect_model_store_cost): Take a vec_load_store_type instead of a
11986 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11987 (vect_model_store_cost): Take a vec_load_store_type instead of a
11989 (vectorizable_mask_load_store): Update accordingly.
11990 (vectorizable_store): Likewise.
11991 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11995 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11996 IFN_MASK_LOAD calls here rather than...
11997 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12000 Alan Hayward <alan.hayward@arm.com>
12001 David Sherwood <david.sherwood@arm.com>
12003 * expmed.c (extract_bit_field_1): For vector extracts,
12004 fall back to extract_bit_field_as_subreg if vec_extract
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 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12012 they are variable or constant sized.
12013 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12014 slots for constant-sized data.
12016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12017 Alan Hayward <alan.hayward@arm.com>
12018 David Sherwood <david.sherwood@arm.com>
12020 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12021 handling COND_EXPRs with boolean comparisons, try to find a better
12022 basis for the mask type than the boolean itself.
12024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12026 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12027 is calculated and how it can be overridden.
12028 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12029 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12031 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12034 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12035 Alan Hayward <alan.hayward@arm.com>
12036 David Sherwood <david.sherwood@arm.com>
12038 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12039 Remove the mode argument.
12040 (aarch64_simd_valid_immediate): Remove the mode and inverse
12042 * config/aarch64/iterators.md (bitsize): New iterator.
12043 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12044 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12045 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12046 aarch64_simd_valid_immediate.
12047 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12048 (aarch64_reg_or_bic_imm): Likewise.
12049 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12050 with an insn_type enum and msl with a modifier_type enum.
12051 Replace element_width with a scalar_mode. Change the shift
12052 to unsigned int. Add constructors for scalar_float_mode and
12053 scalar_int_mode elements.
12054 (aarch64_vect_float_const_representable_p): Delete.
12055 (aarch64_can_const_movi_rtx_p)
12056 (aarch64_simd_scalar_immediate_valid_for_move)
12057 (aarch64_simd_make_constant): Update call to
12058 aarch64_simd_valid_immediate.
12059 (aarch64_advsimd_valid_immediate_hs): New function.
12060 (aarch64_advsimd_valid_immediate): Likewise.
12061 (aarch64_simd_valid_immediate): Remove mode and inverse
12062 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12063 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12064 and aarch64_float_const_representable_p on the result.
12065 (aarch64_output_simd_mov_immediate): Remove mode argument.
12066 Update call to aarch64_simd_valid_immediate and use of
12067 simd_immediate_info.
12068 (aarch64_output_scalar_simd_mov_immediate): Update call
12071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12072 Alan Hayward <alan.hayward@arm.com>
12073 David Sherwood <david.sherwood@arm.com>
12075 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12076 (mode_nunits): Likewise CONST_MODE_NUNITS.
12077 * machmode.def (ADJUST_NUNITS): Document.
12078 * genmodes.c (mode_data::need_nunits_adj): New field.
12079 (blank_mode): Update accordingly.
12080 (adj_nunits): New variable.
12081 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12083 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12084 listed in adj_nunits.
12085 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12086 listed in adj_nunits. Don't emit case statements for such modes.
12087 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12088 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12089 nothing if adj_nunits is nonnull.
12090 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12091 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12092 (emit_mode_fbit): Update use of print_maybe_const_decl.
12093 (emit_move_size): Likewise. Treat the array as non-const
12095 (emit_mode_adjustments): Handle adj_nunits.
12097 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12099 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12100 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12101 (VECTOR_MODES): Use it.
12102 (make_vector_modes): Take the prefix as an argument.
12104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12105 Alan Hayward <alan.hayward@arm.com>
12106 David Sherwood <david.sherwood@arm.com>
12108 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12109 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12110 for MODE_VECTOR_BOOL.
12111 * machmode.def (VECTOR_BOOL_MODE): Document.
12112 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12113 (make_vector_bool_mode): New function.
12114 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12116 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12117 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12119 * stor-layout.c (int_mode_for_mode): Likewise.
12120 * tree.c (build_vector_type_for_mode): Likewise.
12121 * varasm.c (output_constant_pool_2): Likewise.
12122 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12123 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12124 for MODE_VECTOR_BOOL.
12125 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12126 of mode class checks.
12127 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12128 instead of a list of mode class checks.
12129 (expand_vector_scalar_condition): Likewise.
12130 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12133 Alan Hayward <alan.hayward@arm.com>
12134 David Sherwood <david.sherwood@arm.com>
12136 * machmode.h (mode_size): Change from unsigned short to
12138 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12139 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12140 or if measurement_type is not polynomial.
12141 (fixed_size_mode::includes_p): Check for constant-sized modes.
12142 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12143 return a poly_uint16 rather than an unsigned short.
12144 (emit_mode_size): Change the type of mode_size from unsigned short
12145 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12146 (emit_mode_adjustments): Cope with polynomial vector sizes.
12147 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12149 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12151 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12152 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12153 * caller-save.c (setup_save_areas): Likewise.
12154 (replace_reg_with_saved_mem): Likewise.
12155 * calls.c (emit_library_call_value_1): Likewise.
12156 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12157 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12158 (gen_lowpart_for_combine): Likewise.
12159 * convert.c (convert_to_integer_1): Likewise.
12160 * cse.c (equiv_constant, cse_insn): Likewise.
12161 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12162 (cselib_subst_to_values): Likewise.
12163 * dce.c (word_dce_process_block): Likewise.
12164 * df-problems.c (df_word_lr_mark_ref): Likewise.
12165 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12166 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12167 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12168 (rtl_for_decl_location): Likewise.
12169 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12170 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12171 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12172 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12173 (expand_expr_real_1): Likewise.
12174 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12175 (pad_below): Likewise.
12176 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12177 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12178 * ira.c (get_subreg_tracking_sizes): Likewise.
12179 * ira-build.c (ira_create_allocno_objects): Likewise.
12180 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12181 (ira_sort_regnos_for_alter_reg): Likewise.
12182 * ira-costs.c (record_operand_costs): Likewise.
12183 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12184 (resolve_simple_move): Likewise.
12185 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12186 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12187 (lra_constraints): Likewise.
12188 (CONST_POOL_OK_P): Reject variable-sized modes.
12189 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12190 (add_pseudo_to_slot, lra_spill): Likewise.
12191 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12192 * optabs-query.c (get_best_extraction_insn): Likewise.
12193 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12194 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12195 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12196 * recog.c (offsettable_address_addr_space_p): Likewise.
12197 * regcprop.c (maybe_mode_change): Likewise.
12198 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12199 * regrename.c (build_def_use): Likewise.
12200 * regstat.c (dump_reg_info): Likewise.
12201 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12202 (find_reloads, find_reloads_subreg_address): Likewise.
12203 * reload1.c (eliminate_regs_1): Likewise.
12204 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12205 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12206 (simplify_binary_operation_1, simplify_subreg): Likewise.
12207 * targhooks.c (default_function_arg_padding): Likewise.
12208 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12209 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12210 (verify_gimple_assign_ternary): Likewise.
12211 * tree-inline.c (estimate_move_cost): Likewise.
12212 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12213 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12214 (get_address_cost_ainc): Likewise.
12215 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12216 (vect_supportable_dr_alignment): Likewise.
12217 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12218 (vectorizable_reduction): Likewise.
12219 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12220 (vectorizable_operation, vectorizable_load): Likewise.
12221 * tree.c (build_same_sized_truth_vector_type): Likewise.
12222 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12223 * var-tracking.c (emit_note_insn_var_location): Likewise.
12224 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12225 (ADDR_VEC_ALIGN): Likewise.
12227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12228 Alan Hayward <alan.hayward@arm.com>
12229 David Sherwood <david.sherwood@arm.com>
12231 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12233 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12234 or if measurement_type is polynomial.
12235 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12236 * combine.c (make_extraction): Likewise.
12237 * dse.c (find_shift_sequence): Likewise.
12238 * dwarf2out.c (mem_loc_descriptor): Likewise.
12239 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12240 (extract_bit_field, extract_low_bits): Likewise.
12241 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12242 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12243 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12244 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12245 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12246 * reload.c (find_reloads): Likewise.
12247 * reload1.c (alter_reg): Likewise.
12248 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12249 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12250 * tree-if-conv.c (predicate_mem_writes): Likewise.
12251 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12252 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12253 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12254 * valtrack.c (dead_debug_insert_temp): Likewise.
12255 * varasm.c (mergeable_constant_section): Likewise.
12256 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12258 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12259 Alan Hayward <alan.hayward@arm.com>
12260 David Sherwood <david.sherwood@arm.com>
12262 * expr.c (expand_assignment): Cope with polynomial mode sizes
12263 when assigning to a CONCAT.
12265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12266 Alan Hayward <alan.hayward@arm.com>
12267 David Sherwood <david.sherwood@arm.com>
12269 * machmode.h (mode_precision): Change from unsigned short to
12271 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12273 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12274 or if measurement_type is not polynomial.
12275 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12276 in which the mode is already known to be a scalar_int_mode.
12277 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12278 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12280 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12281 for GET_MODE_PRECISION.
12282 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12283 for GET_MODE_PRECISION.
12284 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12286 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12287 (expand_field_assignment, make_extraction): Likewise.
12288 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12289 (get_last_value): Likewise.
12290 * convert.c (convert_to_integer_1): Likewise.
12291 * cse.c (cse_insn): Likewise.
12292 * expr.c (expand_expr_real_1): Likewise.
12293 * lra-constraints.c (simplify_operand_subreg): Likewise.
12294 * optabs-query.c (can_atomic_load_p): Likewise.
12295 * optabs.c (expand_atomic_load): Likewise.
12296 (expand_atomic_store): Likewise.
12297 * ree.c (combine_reaching_defs): Likewise.
12298 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12299 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12300 * tree.h (type_has_mode_precision_p): Likewise.
12301 * ubsan.c (instrument_si_overflow): Likewise.
12303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12304 Alan Hayward <alan.hayward@arm.com>
12305 David Sherwood <david.sherwood@arm.com>
12307 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12308 polynomial numbers of units.
12309 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12310 (valid_vector_subparts_p): New function.
12311 (build_vector_type): Remove temporary shim and take the number
12312 of units as a poly_uint64 rather than an int.
12313 (build_opaque_vector_type): Take the number of units as a
12314 poly_uint64 rather than an int.
12315 * tree.c (build_vector_from_ctor): Handle polynomial
12316 TYPE_VECTOR_SUBPARTS.
12317 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12318 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12319 (build_vector_from_val): If the number of units is variable,
12320 use build_vec_duplicate_cst for constant operands and
12321 VEC_DUPLICATE_EXPR otherwise.
12322 (make_vector_type): Remove temporary is_constant ().
12323 (build_vector_type, build_opaque_vector_type): Take the number of
12324 units as a poly_uint64 rather than an int.
12325 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12327 * cfgexpand.c (expand_debug_expr): Likewise.
12328 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12329 (store_constructor, expand_expr_real_1): Likewise.
12330 (const_scalar_mask_from_tree): Likewise.
12331 * fold-const-call.c (fold_const_reduction): Likewise.
12332 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12333 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12334 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12335 (fold_relational_const): Likewise.
12336 (native_interpret_vector): Likewise. Change the size from an
12337 int to an unsigned int.
12338 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12339 TYPE_VECTOR_SUBPARTS.
12340 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12341 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12342 duplicating a non-constant operand into a variable-length vector.
12343 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12344 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12345 * ipa-icf.c (sem_variable::equals): Likewise.
12346 * match.pd: Likewise.
12347 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12348 * print-tree.c (print_node): Likewise.
12349 * stor-layout.c (layout_type): Likewise.
12350 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12351 * tree-cfg.c (verify_gimple_comparison): Likewise.
12352 (verify_gimple_assign_binary): Likewise.
12353 (verify_gimple_assign_ternary): Likewise.
12354 (verify_gimple_assign_single): Likewise.
12355 * tree-pretty-print.c (dump_generic_node): Likewise.
12356 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12357 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12358 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12359 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12360 (vect_shift_permute_load_chain): Likewise.
12361 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12362 (expand_vector_condition, optimize_vector_constructor): Likewise.
12363 (lower_vec_perm, get_compute_type): Likewise.
12364 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12365 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12366 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12367 (vect_recog_mask_conversion_pattern): Likewise.
12368 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12369 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12370 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12371 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12372 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12373 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12374 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12375 (supportable_widening_operation): Likewise.
12376 (supportable_narrowing_operation): Likewise.
12377 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12379 * varasm.c (output_constant): Likewise.
12381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12382 Alan Hayward <alan.hayward@arm.com>
12383 David Sherwood <david.sherwood@arm.com>
12385 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12386 so that both the length == 3 and length != 3 cases set up their
12387 own permute vectors. Add comments explaining why we know the
12388 number of elements is constant.
12389 (vect_permute_load_chain): Likewise.
12391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12392 Alan Hayward <alan.hayward@arm.com>
12393 David Sherwood <david.sherwood@arm.com>
12395 * machmode.h (mode_nunits): Change from unsigned char to
12397 (ONLY_FIXED_SIZE_MODES): New macro.
12398 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12399 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12400 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12402 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12403 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12404 or if measurement_type is not polynomial.
12405 * genmodes.c (ZERO_COEFFS): New macro.
12406 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12408 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12409 Use ZERO_COEFFS when emitting initializers.
12410 * data-streamer.h (bp_pack_poly_value): New function.
12411 (bp_unpack_poly_value): Likewise.
12412 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12413 for GET_MODE_NUNITS.
12414 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12415 for GET_MODE_NUNITS.
12416 * tree.c (make_vector_type): Remove temporary shim and make
12417 the real function take the number of units as a poly_uint64
12418 rather than an int.
12419 (build_vector_type_for_mode): Handle polynomial nunits.
12420 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12421 * emit-rtl.c (const_vec_series_p_1): Likewise.
12422 (gen_rtx_CONST_VECTOR): Likewise.
12423 * fold-const.c (test_vec_duplicate_folding): Likewise.
12424 * genrecog.c (validate_pattern): Likewise.
12425 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12426 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12427 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12428 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12429 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12430 * rtlanal.c (subreg_get_info): Likewise.
12431 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12432 (vect_grouped_load_supported): Likewise.
12433 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12434 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12435 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12436 (simplify_const_unary_operation, simplify_binary_operation_1)
12437 (simplify_const_binary_operation, simplify_ternary_operation)
12438 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12439 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12440 instead of CONST_VECTOR_NUNITS.
12441 * varasm.c (output_constant_pool_2): Likewise.
12442 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12443 explicit-encoded elements in the XVEC for variable-length vectors.
12445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12447 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12450 Alan Hayward <alan.hayward@arm.com>
12451 David Sherwood <david.sherwood@arm.com>
12453 * coretypes.h (fixed_size_mode): Declare.
12454 (fixed_size_mode_pod): New typedef.
12455 * builtins.h (target_builtins::x_apply_args_mode)
12456 (target_builtins::x_apply_result_mode): Change type to
12457 fixed_size_mode_pod.
12458 * builtins.c (apply_args_size, apply_result_size, result_vector)
12459 (expand_builtin_apply_args_1, expand_builtin_apply)
12460 (expand_builtin_return): Update accordingly.
12462 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12464 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12465 * cselib.c (cselib_hash_rtx): Likewise.
12466 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12467 CONST_VECTOR encoding.
12469 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12470 Jeff Law <law@redhat.com>
12473 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12474 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12475 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12476 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12479 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12480 explicitly probe *sp in a noreturn function if there were any callee
12481 register saves or frame pointer is needed.
12483 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12486 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12487 BLKmode for ternary, binary or unary expressions.
12490 * var-tracking.c (delete_vta_debug_insn): New inline function.
12491 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12492 insns from get_insns () to NULL instead of each bb separately.
12493 Use delete_vta_debug_insn. No longer static.
12494 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12495 delete_vta_debug_insns callers.
12496 * rtl.h (delete_vta_debug_insns): Declare.
12497 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12498 instead of variable_tracking_main.
12500 2018-01-03 Martin Sebor <msebor@redhat.com>
12502 PR tree-optimization/83603
12503 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12504 arguments past the endof the argument list in functions declared
12505 without a prototype.
12506 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12507 Avoid checking when arguments are null.
12509 2018-01-03 Martin Sebor <msebor@redhat.com>
12512 * doc/extend.texi (attribute const): Fix a typo.
12513 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12514 issuing -Wsuggest-attribute for void functions.
12516 2018-01-03 Martin Sebor <msebor@redhat.com>
12518 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12519 offset_int::from instead of wide_int::to_shwi.
12520 (maybe_diag_overlap): Remove assertion.
12521 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12522 * gimple-ssa-sprintf.c (format_directive): Same.
12523 (parse_directive): Same.
12524 (sprintf_dom_walker::compute_format_length): Same.
12525 (try_substitute_return_value): Same.
12527 2018-01-03 Jeff Law <law@redhat.com>
12529 PR middle-end/83654
12530 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12531 non-constant residual for zero at runtime and avoid probing in
12532 that case. Reorganize code for trailing problem to mirror handling
12535 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12537 PR tree-optimization/83501
12538 * tree-ssa-strlen.c (get_string_cst): New.
12539 (handle_char_store): Call get_string_cst.
12541 2018-01-03 Martin Liska <mliska@suse.cz>
12543 PR tree-optimization/83593
12544 * tree-ssa-strlen.c: Include tree-cfg.h.
12545 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12546 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12547 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12549 (strlen_dom_walker::before_dom_children): Call
12550 gimple_purge_dead_eh_edges. Dump tranformation with details
12552 (strlen_dom_walker::before_dom_children): Update call by adding
12553 new argument cleanup_eh.
12554 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12556 2018-01-03 Martin Liska <mliska@suse.cz>
12559 * cif-code.def (VARIADIC_THUNK): New enum value.
12560 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12563 2018-01-03 Jan Beulich <jbeulich@suse.com>
12565 * sse.md (mov<mode>_internal): Tighten condition for when to use
12566 vmovdqu<ssescalarsize> for TI and OI modes.
12568 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12570 Update copyright years.
12572 2018-01-03 Martin Liska <mliska@suse.cz>
12575 * ipa-visibility.c (function_and_variable_visibility): Skip
12576 functions with noipa attribure.
12578 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12580 * gcc.c (process_command): Update copyright notice dates.
12581 * gcov-dump.c (print_version): Ditto.
12582 * gcov.c (print_version): Ditto.
12583 * gcov-tool.c (print_version): Ditto.
12584 * gengtype.c (create_file): Ditto.
12585 * doc/cpp.texi: Bump @copying's copyright year.
12586 * doc/cppinternals.texi: Ditto.
12587 * doc/gcc.texi: Ditto.
12588 * doc/gccint.texi: Ditto.
12589 * doc/gcov.texi: Ditto.
12590 * doc/install.texi: Ditto.
12591 * doc/invoke.texi: Ditto.
12593 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12595 * vector-builder.h (vector_builder::m_full_nelts): Change from
12596 unsigned int to poly_uint64.
12597 (vector_builder::full_nelts): Update prototype accordingly.
12598 (vector_builder::new_vector): Likewise.
12599 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12600 (vector_builder::operator ==): Likewise.
12601 (vector_builder::finalize): Likewise.
12602 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12603 Take the number of elements as a poly_uint64 rather than an
12605 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12606 from unsigned int to poly_uint64.
12607 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12608 (vec_perm_indices::new_vector): Likewise.
12609 (vec_perm_indices::length): Likewise.
12610 (vec_perm_indices::nelts_per_input): Likewise.
12611 (vec_perm_indices::input_nelts): Likewise.
12612 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12613 number of elements per input as a poly_uint64 rather than an
12614 unsigned int. Use the original encoding for variable-length
12615 vectors, rather than clamping each individual element.
12616 For the second and subsequent elements in each pattern,
12617 clamp the step and base before clamping their sum.
12618 (vec_perm_indices::series_p): Handle polynomial element counts.
12619 (vec_perm_indices::all_in_range_p): Likewise.
12620 (vec_perm_indices_to_tree): Likewise.
12621 (vec_perm_indices_to_rtx): Likewise.
12622 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12623 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12624 (tree_vector_builder::new_binary_operation): Handle polynomial
12625 element counts. Return false if we need to know the number
12626 of elements at compile time.
12627 * fold-const.c (fold_vec_perm): Punt if the number of elements
12628 isn't known at compile time.
12630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12632 * vec-perm-indices.h (vec_perm_builder): Change element type
12633 from HOST_WIDE_INT to poly_int64.
12634 (vec_perm_indices::element_type): Update accordingly.
12635 (vec_perm_indices::clamp): Handle polynomial element_types.
12636 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12637 (vec_perm_indices::all_in_range_p): Likewise.
12638 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12640 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12641 polynomial vec_perm_indices element types.
12642 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12643 * fold-const.c (fold_vec_perm): Likewise.
12644 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12645 * tree-vect-generic.c (lower_vec_perm): Likewise.
12646 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12647 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12648 element type to 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 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12655 rather than an int. Use plus_constant.
12656 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12657 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12660 Alan Hayward <alan.hayward@arm.com>
12661 David Sherwood <david.sherwood@arm.com>
12663 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12664 a HOST_WIDE_INT to a poly_int64.
12666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12667 Alan Hayward <alan.hayward@arm.com>
12668 David Sherwood <david.sherwood@arm.com>
12670 * calls.c (load_register_parameters): Cope with polynomial
12671 mode sizes. Require a constant size for BLKmode parameters
12672 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12673 forces a parameter to be padded at the lsb end in order to
12674 fill a complete number of words, require the parameter size
12675 to be ordered wrt UNITS_PER_WORD.
12677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12678 Alan Hayward <alan.hayward@arm.com>
12679 David Sherwood <david.sherwood@arm.com>
12681 * reload1.c (spill_stack_slot_width): Change element type
12682 from unsigned int to poly_uint64_pod.
12683 (alter_reg): Treat mode sizes as polynomial.
12685 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12686 Alan Hayward <alan.hayward@arm.com>
12687 David Sherwood <david.sherwood@arm.com>
12689 * reload.c (complex_word_subreg_p): New function.
12690 (reload_inner_reg_of_subreg, push_reload): Use it.
12692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12693 Alan Hayward <alan.hayward@arm.com>
12694 David Sherwood <david.sherwood@arm.com>
12696 * lra-constraints.c (process_alt_operands): Reject matched
12697 operands whose sizes aren't ordered.
12698 (match_reload): Refer to this check here.
12700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12701 Alan Hayward <alan.hayward@arm.com>
12702 David Sherwood <david.sherwood@arm.com>
12704 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12705 that the mode size is in the set {1, 2, 4, 8, 16}.
12707 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12708 Alan Hayward <alan.hayward@arm.com>
12709 David Sherwood <david.sherwood@arm.com>
12711 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12712 Use plus_constant instead of gen_rtx_PLUS.
12714 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12715 Alan Hayward <alan.hayward@arm.com>
12716 David Sherwood <david.sherwood@arm.com>
12718 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12719 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12720 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12721 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12722 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12723 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12724 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12725 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12726 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12727 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12729 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12730 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12731 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12732 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12733 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12734 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12735 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12736 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12737 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12738 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12740 * expr.c (emit_move_resolve_push): Treat the input and result
12741 of PUSH_ROUNDING as a poly_int64.
12742 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12743 (emit_push_insn): Likewise.
12744 * lra-eliminations.c (mark_not_eliminable): Likewise.
12745 * recog.c (push_operand): Likewise.
12746 * reload1.c (elimination_effects): Likewise.
12747 * rtlanal.c (nonzero_bits1): Likewise.
12748 * calls.c (store_one_arg): Likewise. Require the padding to be
12749 known at compile time.
12751 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12752 Alan Hayward <alan.hayward@arm.com>
12753 David Sherwood <david.sherwood@arm.com>
12755 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12756 Use plus_constant instead of gen_rtx_PLUS.
12758 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12759 Alan Hayward <alan.hayward@arm.com>
12760 David Sherwood <david.sherwood@arm.com>
12762 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12763 rather than an int.
12765 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12766 Alan Hayward <alan.hayward@arm.com>
12767 David Sherwood <david.sherwood@arm.com>
12769 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12770 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12771 via stack temporaries. Treat the mode size as polynomial 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 * expr.c (expand_expr_real_2): When handling conversions involving
12778 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12779 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12780 as a poly_uint64 too.
12782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12783 Alan Hayward <alan.hayward@arm.com>
12784 David Sherwood <david.sherwood@arm.com>
12786 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12789 Alan Hayward <alan.hayward@arm.com>
12790 David Sherwood <david.sherwood@arm.com>
12792 * combine.c (can_change_dest_mode): Handle polynomial
12793 REGMODE_NATURAL_SIZE.
12794 * expmed.c (store_bit_field_1): Likewise.
12795 * expr.c (store_constructor): Likewise.
12796 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12797 and polynomial REGMODE_NATURAL_SIZE.
12798 (gen_lowpart_common): Likewise.
12799 * reginfo.c (record_subregs_of_mode): Likewise.
12800 * rtlanal.c (read_modify_subreg_p): Likewise.
12802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12803 Alan Hayward <alan.hayward@arm.com>
12804 David Sherwood <david.sherwood@arm.com>
12806 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12807 numbers of elements.
12809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12810 Alan Hayward <alan.hayward@arm.com>
12811 David Sherwood <david.sherwood@arm.com>
12813 * match.pd: Cope with polynomial numbers of vector elements.
12815 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12816 Alan Hayward <alan.hayward@arm.com>
12817 David Sherwood <david.sherwood@arm.com>
12819 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12820 in a POINTER_PLUS_EXPR.
12822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12823 Alan Hayward <alan.hayward@arm.com>
12824 David Sherwood <david.sherwood@arm.com>
12826 * omp-simd-clone.c (simd_clone_subparts): New function.
12827 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12828 (ipa_simd_modify_function_body): Likewise.
12830 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12831 Alan Hayward <alan.hayward@arm.com>
12832 David Sherwood <david.sherwood@arm.com>
12834 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12835 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12836 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12837 (expand_vector_condition, vector_element): Likewise.
12838 (subparts_gt): New function.
12839 (get_compute_type): Use subparts_gt.
12840 (count_type_subparts): Delete.
12841 (expand_vector_operations_1): Use subparts_gt instead of
12842 count_type_subparts.
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-data-refs.c (vect_no_alias_p): Replace with...
12849 (vect_compile_time_alias): ...this new function. Do the calculation
12850 on poly_ints rather than trees.
12851 (vect_prune_runtime_alias_test_list): Update call accordingly.
12853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12854 Alan Hayward <alan.hayward@arm.com>
12855 David Sherwood <david.sherwood@arm.com>
12857 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12859 (vect_schedule_slp_instance): Likewise.
12861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12862 Alan Hayward <alan.hayward@arm.com>
12863 David Sherwood <david.sherwood@arm.com>
12865 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12866 constant and extern definitions for variable-length vectors.
12867 (vect_get_constant_vectors): Note that the number of units
12868 is known to be constant.
12870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12871 Alan Hayward <alan.hayward@arm.com>
12872 David Sherwood <david.sherwood@arm.com>
12874 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12875 of units as polynomial. Choose between WIDE and NARROW based
12878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12879 Alan Hayward <alan.hayward@arm.com>
12880 David Sherwood <david.sherwood@arm.com>
12882 * tree-vect-stmts.c (simd_clone_subparts): New function.
12883 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12886 Alan Hayward <alan.hayward@arm.com>
12887 David Sherwood <david.sherwood@arm.com>
12889 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12890 vectors as polynomial. Use build_index_vector for
12891 IFN_GOMP_SIMD_LANE.
12893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12894 Alan Hayward <alan.hayward@arm.com>
12895 David Sherwood <david.sherwood@arm.com>
12897 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12898 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12899 for variable-length vectors.
12900 (vectorizable_mask_load_store): Treat the number of units as
12901 polynomial, asserting that it is constant if the condition has
12902 already been enforced.
12903 (vectorizable_store, vectorizable_load): Likewise.
12905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12906 Alan Hayward <alan.hayward@arm.com>
12907 David Sherwood <david.sherwood@arm.com>
12909 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12910 of units as polynomial. Punt if we can't tell at compile time
12911 which vector contains the final result.
12913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12914 Alan Hayward <alan.hayward@arm.com>
12915 David Sherwood <david.sherwood@arm.com>
12917 * tree-vect-loop.c (vectorizable_induction): Treat the number
12918 of units as polynomial. Punt on SLP inductions. Use an integer
12919 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12920 cast of such a series for variable-length floating-point
12923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12924 Alan Hayward <alan.hayward@arm.com>
12925 David Sherwood <david.sherwood@arm.com>
12927 * tree.h (build_index_vector): Declare.
12928 * tree.c (build_index_vector): New function.
12929 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12930 of units as polynomial, forcibly converting it to a constant if
12931 vectorizable_reduction has already enforced the condition.
12932 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12933 to create a {1,2,3,...} vector.
12934 (vectorizable_reduction): Treat the number of units as polynomial.
12935 Choose vectype_in based on the largest scalar element size rather
12936 than the smallest number of units. Enforce the restrictions
12939 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12940 Alan Hayward <alan.hayward@arm.com>
12941 David Sherwood <david.sherwood@arm.com>
12943 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12944 number of units as polynomial.
12946 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12947 Alan Hayward <alan.hayward@arm.com>
12948 David Sherwood <david.sherwood@arm.com>
12950 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12951 * target.def (autovectorize_vector_sizes): Return the vector sizes
12952 by pointer, using vector_sizes rather than a bitmask.
12953 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12954 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12955 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12957 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12958 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12959 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12960 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12961 * omp-general.c (omp_max_vf): Likewise.
12962 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12963 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12964 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12965 * tree-vect-slp.c (vect_slp_bb): Likewise.
12966 * doc/tm.texi: Regenerate.
12967 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12969 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12970 the vector size as a poly_uint64 rather than an unsigned int.
12971 (current_vector_size): Change from an unsigned int to a poly_uint64.
12972 (get_vectype_for_scalar_type): Update accordingly.
12973 * tree.h (build_truth_vector_type): Take the size and number of
12974 units as a poly_uint64 rather than an unsigned int.
12975 (build_vector_type): Add a temporary overload that takes
12976 the number of units as a poly_uint64 rather than an unsigned int.
12977 * tree.c (make_vector_type): Likewise.
12978 (build_truth_vector_type): Take the number of units as a poly_uint64
12979 rather than an unsigned int.
12981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12982 Alan Hayward <alan.hayward@arm.com>
12983 David Sherwood <david.sherwood@arm.com>
12985 * target.def (get_mask_mode): Take the number of units and length
12986 as poly_uint64s rather than unsigned ints.
12987 * targhooks.h (default_get_mask_mode): Update accordingly.
12988 * targhooks.c (default_get_mask_mode): Likewise.
12989 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12990 * doc/tm.texi: Regenerate.
12992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12993 Alan Hayward <alan.hayward@arm.com>
12994 David Sherwood <david.sherwood@arm.com>
12996 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12997 * omp-general.c (omp_max_vf): Likewise.
12998 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12999 (expand_omp_simd): Handle polynomial safelen.
13000 * omp-low.c (omplow_simd_context): Add a default constructor.
13001 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13002 (lower_rec_simd_input_clauses): Update accordingly.
13003 (lower_rec_input_clauses): Likewise.
13005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13006 Alan Hayward <alan.hayward@arm.com>
13007 David Sherwood <david.sherwood@arm.com>
13009 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13010 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13011 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13012 (vect_analyze_slp_cost): Likewise.
13013 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13014 (vect_model_load_cost): Likewise.
13016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13017 Alan Hayward <alan.hayward@arm.com>
13018 David Sherwood <david.sherwood@arm.com>
13020 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13021 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13022 from an unsigned int * to a poly_uint64_pod *.
13023 (calculate_unrolling_factor): New function.
13024 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13027 Alan Hayward <alan.hayward@arm.com>
13028 David Sherwood <david.sherwood@arm.com>
13030 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13031 from an unsigned int to a poly_uint64.
13032 (_loop_vec_info::slp_unrolling_factor): Likewise.
13033 (_loop_vec_info::vectorization_factor): Change from an int
13035 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13036 (vect_get_num_vectors): New function.
13037 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13038 (vect_get_num_copies): Use vect_get_num_vectors.
13039 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13040 to an unsigned int *.
13041 (vect_analyze_data_refs): Change min_vf from an int * to a
13043 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13044 than an unsigned HOST_WIDE_INT.
13045 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13046 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13047 to an unsigned int *.
13048 (vect_analyze_data_ref_dependences): Likewise.
13049 (vect_compute_data_ref_alignment): Handle polynomial vf.
13050 (vect_enhance_data_refs_alignment): Likewise.
13051 (vect_prune_runtime_alias_test_list): Likewise.
13052 (vect_shift_permute_load_chain): Likewise.
13053 (vect_supportable_dr_alignment): Likewise.
13054 (dependence_distance_ge_vf): Take the vectorization factor as a
13055 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13056 (vect_analyze_data_refs): Change min_vf from an int * to a
13058 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13059 vfm1 as a poly_uint64 rather than an int. Make the same change
13060 for the returned bound_scalar.
13061 (vect_gen_vector_loop_niters): Handle polynomial vf.
13062 (vect_do_peeling): Likewise. Update call to
13063 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13064 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13066 * tree-vect-loop.c (vect_determine_vectorization_factor)
13067 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13068 (vect_get_known_peeling_cost): Likewise.
13069 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13070 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13071 (vect_transform_loop): Likewise. Use the lowest possible VF when
13072 updating the upper bounds of the loop.
13073 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13074 rather than an int.
13075 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13076 polynomial unroll factors.
13077 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13078 (vect_make_slp_decision): Likewise.
13079 (vect_supported_load_permutation_p): Likewise, and polynomial
13081 (vect_analyze_slp_cost): Handle polynomial vf.
13082 (vect_slp_analyze_node_operations): Likewise.
13083 (vect_slp_analyze_bb_1): Likewise.
13084 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13085 than an unsigned HOST_WIDE_INT.
13086 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13087 (vectorizable_load): Handle polynomial vf.
13088 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13090 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13093 Alan Hayward <alan.hayward@arm.com>
13094 David Sherwood <david.sherwood@arm.com>
13096 * match.pd: Handle bit operations involving three constants
13097 and try to fold one pair.
13099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13101 * tree-vect-loop-manip.c: Include gimple-fold.h.
13102 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13103 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13104 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13105 Add a path that uses a step of VF instead of 1, but disable it
13107 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13108 and niters_no_overflow parameters. Update calls to
13109 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13110 Create a new SSA name if the latter choses to use a ste other
13111 than zero, and return it via niters_vector_mult_vf_var.
13112 * tree-vect-loop.c (vect_transform_loop): Update calls to
13113 vect_do_peeling, vect_gen_vector_loop_niters and
13114 slpeel_make_loop_iterate_ntimes.
13115 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13116 (vect_gen_vector_loop_niters): Update declarations after above changes.
13118 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13120 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13121 128-bit round to integer instructions.
13122 (ceil<mode>2): Likewise.
13123 (btrunc<mode>2): Likewise.
13124 (round<mode>2): Likewise.
13126 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13128 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13129 unaligned VSX load/store on P8/P9.
13130 (expand_block_clear): Allow the use of unaligned VSX
13131 load/store on P8/P9.
13133 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13135 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13137 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13138 swap associated with both a load and a store.
13140 2018-01-02 Andrew Waterman <andrew@sifive.com>
13142 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13143 * config/riscv/riscv.md (clear_cache): Use it.
13145 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13147 * web.c: Remove out-of-date comment.
13149 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13151 * expr.c (fixup_args_size_notes): Check that any existing
13152 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13153 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13154 (emit_single_push_insn): ...here.
13156 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13158 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13159 (const_vector_encoded_nelts): New function.
13160 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13161 (const_vector_int_elt, const_vector_elt): Declare.
13162 * emit-rtl.c (const_vector_int_elt_1): New function.
13163 (const_vector_elt): Likewise.
13164 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13165 of CONST_VECTOR_ELT.
13167 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13169 * expr.c: Include rtx-vector-builder.h.
13170 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13171 directly on the tree encoding.
13172 (const_vector_from_tree): Likewise.
13173 * optabs.c: Include rtx-vector-builder.h.
13174 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13175 sequence of "u" values.
13176 * vec-perm-indices.c: Include rtx-vector-builder.h.
13177 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13178 directly on the vec_perm_indices encoding.
13180 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13182 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13183 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13184 * rtx-vector-builder.h: New file.
13185 * rtx-vector-builder.c: Likewise.
13186 * rtl.h (rtx_def::u2): Add a const_vector field.
13187 (CONST_VECTOR_NPATTERNS): New macro.
13188 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13189 (CONST_VECTOR_DUPLICATE_P): Likewise.
13190 (CONST_VECTOR_STEPPED_P): Likewise.
13191 (CONST_VECTOR_ENCODED_ELT): Likewise.
13192 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13193 (unwrap_const_vec_duplicate): Likewise.
13194 (const_vec_series_p): Check for a non-duplicated vector encoding.
13195 Say that the function only returns true for integer vectors.
13196 * emit-rtl.c: Include rtx-vector-builder.h.
13197 (gen_const_vec_duplicate_1): Delete.
13198 (gen_const_vector): Call gen_const_vec_duplicate instead of
13199 gen_const_vec_duplicate_1.
13200 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13201 (gen_const_vec_duplicate): Use rtx_vector_builder.
13202 (gen_const_vec_series): Likewise.
13203 (gen_rtx_CONST_VECTOR): Likewise.
13204 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13205 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13206 Build a new vector rather than modifying a CONST_VECTOR in-place.
13207 (handle_special_swappables): Update call accordingly.
13208 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13209 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13210 Build a new vector rather than modifying a CONST_VECTOR in-place.
13211 (handle_special_swappables): Update call accordingly.
13213 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13215 * simplify-rtx.c (simplify_const_binary_operation): Use
13216 CONST_VECTOR_ELT instead of XVECEXP.
13218 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13220 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13221 the selector elements to be different from the data elements
13222 if the selector is a VECTOR_CST.
13223 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13224 ssizetype for the selector.
13226 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13228 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13229 before testing each element individually.
13230 * tree-vect-generic.c (lower_vec_perm): Likewise.
13232 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13234 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13235 * selftest-run-tests.c (selftest::run_tests): Call it.
13236 * vector-builder.h (vector_builder::operator ==): New function.
13237 (vector_builder::operator !=): Likewise.
13238 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13239 (vec_perm_indices::all_from_input_p): New function.
13240 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13241 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13242 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13243 instead of reading the VECTOR_CST directly. Detect whether both
13244 vector inputs are the same before constructing the vec_perm_indices,
13245 and update the number of inputs argument accordingly. Use the
13246 utility functions added above. Only construct sel2 if we need to.
13248 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13250 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13251 the broadcast of the low byte.
13252 (expand_mult_highpart): Use an explicit encoding for the permutes.
13253 * optabs-query.c (can_mult_highpart_p): Likewise.
13254 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13255 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13256 (vectorizable_bswap): Likewise.
13257 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13258 explicit encoding for the power-of-2 permutes.
13259 (vect_permute_store_chain): Likewise.
13260 (vect_grouped_load_supported): Likewise.
13261 (vect_permute_load_chain): Likewise.
13263 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13265 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13266 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13267 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13268 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13269 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13270 (vect_gen_perm_mask_any): Likewise.
13272 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13274 * int-vector-builder.h: New file.
13275 * vec-perm-indices.h: Include int-vector-builder.h.
13276 (vec_perm_indices): Redefine as an int_vector_builder.
13277 (auto_vec_perm_indices): Delete.
13278 (vec_perm_builder): Redefine as a stand-alone class.
13279 (vec_perm_indices::vec_perm_indices): New function.
13280 (vec_perm_indices::clamp): Likewise.
13281 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13282 (vec_perm_indices::new_vector): New function.
13283 (vec_perm_indices::new_expanded_vector): Update for new
13284 vec_perm_indices class.
13285 (vec_perm_indices::rotate_inputs): New function.
13286 (vec_perm_indices::all_in_range_p): Operate directly on the
13287 encoded form, without computing elided elements.
13288 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13289 encoding. Update for new vec_perm_indices class.
13290 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13291 the given vec_perm_builder.
13292 (expand_vec_perm_var): Update vec_perm_builder constructor.
13293 (expand_mult_highpart): Use vec_perm_builder instead of
13294 auto_vec_perm_indices.
13295 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13296 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13297 or double series encoding as appropriate.
13298 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13299 vec_perm_indices instead of auto_vec_perm_indices.
13300 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13301 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13302 (vect_permute_store_chain): Likewise.
13303 (vect_grouped_load_supported): Likewise.
13304 (vect_permute_load_chain): Likewise.
13305 (vect_shift_permute_load_chain): Likewise.
13306 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13307 (vect_transform_slp_perm_load): Likewise.
13308 (vect_schedule_slp_instance): Likewise.
13309 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13310 (vectorizable_mask_load_store): Likewise.
13311 (vectorizable_bswap): Likewise.
13312 (vectorizable_store): Likewise.
13313 (vectorizable_load): Likewise.
13314 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13315 vec_perm_indices instead of auto_vec_perm_indices. Use
13316 tree_to_vec_perm_builder to read the vector from a tree.
13317 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13318 vec_perm_builder instead of a vec_perm_indices.
13319 (have_whole_vector_shift): Use vec_perm_builder and
13320 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13321 truncation to calc_vec_perm_mask_for_shift.
13322 (vect_create_epilog_for_reduction): Likewise.
13323 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13324 from auto_vec_perm_indices to vec_perm_indices.
13325 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13326 instead of changing individual elements.
13327 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13328 the vector in d.perm.
13329 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13330 from auto_vec_perm_indices to vec_perm_indices.
13331 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13332 instead of changing individual elements.
13333 (arm_vectorize_vec_perm_const): Use new_vector to install
13334 the vector in d.perm.
13335 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13336 Update vec_perm_builder constructor.
13337 (rs6000_expand_interleave): Likewise.
13338 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13339 (rs6000_expand_interleave): Likewise.
13341 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13343 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13344 to qimode could truncate the indices.
13345 * optabs.c (expand_vec_perm_var): Likewise.
13347 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13349 * Makefile.in (OBJS): Add vec-perm-indices.o.
13350 * vec-perm-indices.h: New file.
13351 * vec-perm-indices.c: Likewise.
13352 * target.h (vec_perm_indices): Replace with a forward class
13354 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13355 * optabs.h: Include vec-perm-indices.h.
13356 (expand_vec_perm): Delete.
13357 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13358 (expand_vec_perm_const): Declare.
13359 * target.def (vec_perm_const_ok): Replace with...
13360 (vec_perm_const): ...this new hook.
13361 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13362 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13363 * doc/tm.texi: Regenerate.
13364 * optabs.def (vec_perm_const): Delete.
13365 * doc/md.texi (vec_perm_const): Likewise.
13366 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13367 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13368 expand_vec_perm for constant permutation vectors. Assert that
13369 the mode of variable permutation vectors is the integer equivalent
13370 of the mode that is being permuted.
13371 * optabs-query.h (selector_fits_mode_p): Declare.
13372 * optabs-query.c: Include vec-perm-indices.h.
13373 (selector_fits_mode_p): New function.
13374 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13375 is defined, instead of checking whether the vec_perm_const_optab
13376 exists. Use targetm.vectorize.vec_perm_const instead of
13377 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13378 fit in the vector mode before using a variable permute.
13379 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13380 vec_perm_indices instead of an rtx.
13381 (expand_vec_perm): Replace with...
13382 (expand_vec_perm_const): ...this new function. Take the selector
13383 as a vec_perm_indices rather than an rtx. Also take the mode of
13384 the selector. Update call to shift_amt_for_vec_perm_mask.
13385 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13386 Use vec_perm_indices::new_expanded_vector to expand the original
13387 selector into bytes. Check whether the indices fit in the vector
13388 mode before using a variable permute.
13389 (expand_vec_perm_var): Make global.
13390 (expand_mult_highpart): Use expand_vec_perm_const.
13391 * fold-const.c: Includes vec-perm-indices.h.
13392 * tree-ssa-forwprop.c: Likewise.
13393 * tree-vect-data-refs.c: Likewise.
13394 * tree-vect-generic.c: Likewise.
13395 * tree-vect-loop.c: Likewise.
13396 * tree-vect-slp.c: Likewise.
13397 * tree-vect-stmts.c: Likewise.
13398 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13400 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13401 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13402 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13403 (aarch64_vectorize_vec_perm_const): ...this new function.
13404 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13405 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13406 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13407 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13408 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13409 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13410 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13412 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13413 check for NEON modes.
13414 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13415 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13416 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13417 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13419 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13420 the old VEC_PERM_CONST conditions.
13421 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13422 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13423 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13424 (ia64_vectorize_vec_perm_const_ok): Merge into...
13425 (ia64_vectorize_vec_perm_const): ...this new function.
13426 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13427 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13428 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13429 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13430 * config/mips/mips.c (mips_expand_vec_perm_const)
13431 (mips_vectorize_vec_perm_const_ok): Merge into...
13432 (mips_vectorize_vec_perm_const): ...this new function.
13433 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13434 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13435 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13436 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13437 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13438 (rs6000_expand_vec_perm_const): Delete.
13439 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13441 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13442 (altivec_expand_vec_perm_const_le): Take each operand individually.
13443 Operate on constant selectors rather than rtxes.
13444 (altivec_expand_vec_perm_const): Likewise. Update call to
13445 altivec_expand_vec_perm_const_le.
13446 (rs6000_expand_vec_perm_const): Delete.
13447 (rs6000_vectorize_vec_perm_const_ok): Delete.
13448 (rs6000_vectorize_vec_perm_const): New function.
13449 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13450 an element count and rtx array.
13451 (rs6000_expand_extract_even): Update call accordingly.
13452 (rs6000_expand_interleave): Likewise.
13453 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13454 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13455 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13456 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13457 (rs6000_expand_vec_perm_const): Delete.
13458 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13459 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13460 (altivec_expand_vec_perm_const_le): Take each operand individually.
13461 Operate on constant selectors rather than rtxes.
13462 (altivec_expand_vec_perm_const): Likewise. Update call to
13463 altivec_expand_vec_perm_const_le.
13464 (rs6000_expand_vec_perm_const): Delete.
13465 (rs6000_vectorize_vec_perm_const_ok): Delete.
13466 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13467 reference to the SPE evmerge intructions.
13468 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13469 an element count and rtx array.
13470 (rs6000_expand_extract_even): Update call accordingly.
13471 (rs6000_expand_interleave): Likewise.
13472 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13473 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13475 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13477 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13479 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13480 vector mode and that that mode matches the mode of the data
13482 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13483 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13484 directly using expand_vec_perm_1 when forcing selectors into
13486 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13488 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13490 * optabs-query.h (can_vec_perm_p): Delete.
13491 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13492 * optabs-query.c (can_vec_perm_p): Split into...
13493 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13494 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13495 particular selector is valid.
13496 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13497 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13498 (vect_grouped_load_supported): Likewise.
13499 (vect_shift_permute_load_chain): Likewise.
13500 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13501 (vect_transform_slp_perm_load): Likewise.
13502 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13503 (vectorizable_bswap): Likewise.
13504 (vect_gen_perm_mask_checked): Likewise.
13505 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13506 implementations of variable permutation vectors into account
13507 when deciding which selector to use.
13508 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13509 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13510 with a false third argument.
13511 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13512 to test whether the constant selector is valid and can_vec_perm_var_p
13513 to test whether a variable selector is valid.
13515 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13517 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13518 * optabs-query.c (can_vec_perm_p): Likewise.
13519 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13520 instead of vec_perm_indices.
13521 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13522 (vect_gen_perm_mask_checked): Likewise,
13523 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13524 (vect_gen_perm_mask_checked): Likewise,
13526 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13528 * optabs-query.h (qimode_for_vec_perm): Declare.
13529 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13530 (qimode_for_vec_perm): ...this new function.
13531 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13533 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13535 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13536 does not have a conditional at the top.
13538 2018-01-02 Richard Biener <rguenther@suse.de>
13540 * ipa-inline.c (big_speedup_p): Fix expression.
13542 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13545 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13548 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13552 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13553 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13554 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13555 cond_taken_branch_cost 3->4.
13557 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13559 PR tree-optimization/83581
13560 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13561 TODO_cleanup_cfg if any changes have been made.
13563 PR middle-end/83608
13564 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13565 convert_modes if target mode has the right side, but different mode
13568 PR middle-end/83609
13569 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13570 last argument when extracting from CONCAT. If either from_real or
13571 from_imag is NULL, use expansion through memory. If result is not
13572 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13573 the parts directly to inner mode, if even that fails, use expansion
13576 PR middle-end/83623
13577 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13578 check for bswap in mode rather than HImode and use that in expand_unop
13581 Copyright (C) 2018 Free Software Foundation, Inc.
13583 Copying and distribution of this file, with or without modification,
13584 are permitted in any medium without royalty provided the copyright
13585 notice and this notice are preserved.