1 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
5 * config/nds32/nds32.md (abssi2): New pattern.
7 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
9 * config/i386/i386.md (rex64namesuffix): New mode attribute.
10 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
11 Merge insn pattern from sse_cvtsi2ss<round_name> and
12 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
13 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
14 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
15 using SWI48 mode iterator.
16 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
17 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
18 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
19 pattern from sse_cvttss2si<round_saeonly_name>
20 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
21 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
22 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
23 using SWI48 mode iterator.
24 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
25 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
26 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
27 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
28 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
29 using SWI48 mode iterator.
30 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
31 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
32 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
33 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
34 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
36 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
37 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
38 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
39 pattern from sse_cvttsd2si<round_saeonly_name>
40 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
42 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
44 * config/nds32/nds32-md-auxiliary.c
45 (nds32_valid_smw_lwm_base_p): Refine.
46 (nds32_output_smw_single_word): Refine.
47 (nds32_output_smw_double_word): New.
48 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
50 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
52 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
53 (nds32_output_stack_pop): Refine.
54 (nds32_expand_unaligned_load): Refine.
55 (nds32_expand_unaligned_store): Refine.
57 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
58 Chung-Ju Wu <jasonwucj@gmail.com>
60 * config/nds32/constants.md: Add TP_REGNUM constant.
61 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
62 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
64 * config/nds32/nds32-doubleword.md: Consider flag_pic.
65 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
66 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
67 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
68 and PIC code generation.
69 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
71 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
73 * config/nds32/nds32.md: Support TLS and PIC.
74 * config/nds32/nds32.c: Support TLS and PIC.
75 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
76 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
79 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
81 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
84 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
85 Chung-Ju Wu <jasonwucj@gmail.com>
87 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
88 * config/nds32/nds32-md-auxiliary.c
89 (symbolic_reference_mentioned_p): New.
90 (nds32_legitimize_ict_address): New.
91 (nds32_expand_ict_move): New.
92 (nds32_indirect_call_referenced_p): New.
93 (nds32_symbol_binds_local_p): Delete.
94 (nds32_long_call_p): Modify.
95 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
96 * config/nds32/nds32-protos.h
97 (symbolic_reference_mentioned_p): Declare.
98 (nds32_legitimize_ict_address): Declare.
99 (nds32_expand_ict_move): Declare.
100 (nds32_indirect_call_referenced_p): Declare.
101 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
102 (nds32_relax_group): Use nds32_ict_const_p as condition.
103 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
104 (nds32_asm_file_start): Output ict_model directive in asm code.
105 (nds32_legitimate_address_p): Consider indirect call.
106 (nds32_print_operand): Consider indirect call.
107 (nds32_print_operand_address): Consider indirect call.
108 (nds32_insert_attributes): Handle "indirect_call" attribute.
109 (TARGET_LEGITIMATE_ADDRESS_P): Define.
110 (TARGET_LEGITIMATE_CONSTANT_P): Define.
111 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
112 (TARGET_DELEGITIMIZE_ADDRESS): Define.
113 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
114 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
115 (TARGET_ICT_MODEL_SMALL): Define.
116 (TARGET_ICT_MODEL_LARGE): Define.
117 * config/nds32/nds32.md (movsi): Consider ict model.
118 (call, call_value): Consider ict model.
119 (sibcall, sibcall_value): Consider ict model.
120 * config/nds32/nds32.opt (mict-model): New option.
121 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
124 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
125 Monk Chiang <sh.chiang04@gmail.com>
126 Jim Wilson <jimw@sifive.com>
128 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
129 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
130 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
131 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
132 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
133 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
134 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
135 compute save_libcall_adjustment properly.
136 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
137 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
138 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
139 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
140 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
141 (ABI_SPEC): Handle mabi=ilp32e.
142 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
144 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
145 <-march>: Add rv32e as an example.
147 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
150 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
151 (intra_create_variable_infos): Handle C++ constructors.
153 2018-05-18 Martin Liska <mliska@suse.cz>
155 * passes.def: Remove a redundant pass.
157 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
160 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
162 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
164 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
166 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
168 (armv2, armv2a, armv3, armv3m): Delete architectures.
169 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
170 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
171 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
173 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
174 (*mulsidi3adddi): Likewise.
175 (mulsidi3): Likewise.
176 (*mulsidi3_nov6): Likewise.
177 (umulsidi3): Likewise.
178 (umulsidi3_nov6): Likewise.
179 (umaddsidi4): Likewise.
180 (*umulsidi3adddi): Likewise.
181 (smulsi3_highpart): Likewise.
182 (*smulsi3_highpart_nov6): Likewise.
183 (umulsi3_highpart): Likewise.
184 (*umulsi3_highpart_nov6): Likewise.
185 * config/arm/arm.h (arm_arch3m): Delete.
186 * config/arm/arm.c (arm_arch3m): Delete.
187 (arm_option_override_internal): Update armv3-related comment.
188 (arm_configure_build_target): Delete use of isa_bit_mode32.
189 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
190 (arm_rtx_costs_internal): Delete check of arm_arch3m.
191 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
194 * config/arm/arm-protos.h (arm_arch3m): Delete.
195 * config/arm/arm-tables.opt: Regenerate.
196 * config/arm/arm-tune.md: Likewise.
197 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
198 deleted architectures.
200 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
202 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
203 (armv5t, armv5te): New features.
204 (ARMv5, ARMv5e): Delete fgroups.
205 (ARMv5t, ARMv5te): Adjust for above changes.
207 (armv5, armv5e): Delete arches.
208 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
210 (*call_reg_arm): Likewise.
211 (*call_value_reg_armv5): Likewise.
212 (*call_value_reg_arm): Likewise.
213 (*call_symbol): Likewise.
214 (*call_value_symbol): Likewise.
215 (*sibcall_insn): Likewise.
216 (*sibcall_value_insn): Likewise.
218 (prefetch): Likewise.
219 (define_split and define_peephole2 dependent on arm_arch5):
221 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
223 (TARGET_ARM_QBIT): Likewise.
224 (TARGET_DSP_MULTIPLY): Likewise.
225 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
226 (arm_arch5, arm_arch5e): Delete.
227 (arm_arch5t, arm_arch5te): Declare.
228 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
229 (arm_arch5t): Declare.
230 (arm_option_reconfigure_globals): Update for the above.
231 (arm_options_perform_arch_sanity_checks): Update comment, replace
232 use of arm_arch5 with arm_arch5t.
233 (use_return_insn): Likewise.
234 (arm_emit_call_insn): Likewise.
235 (output_return_instruction): Likewise.
236 (arm_final_prescan_insn): Likewise.
237 (arm_coproc_builtin_available): Likewise.
238 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
239 arm_arch5e with arm_arch5t and arm_arch5te.
240 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
241 (arm_arch5t, arm_arch5te): Declare.
242 * config/arm/arm-tables.opt: Regenerate.
243 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
244 * config/arm/t-multilib: Likewise.
245 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
246 instead of arm_arch5.
247 (*call_reg_thumb1): Likewise.
248 (*call_value_reg_thumb1_v5): Likewise.
249 (*call_value_reg_thumb1): Likewise.
250 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
252 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
254 2018-05-18 Martin Liska <mliska@suse.cz>
256 PR gcov-profile/84846
257 * doc/gcov.texi: Document -t option of gcov tool.
259 2018-05-18 Martin Liska <mliska@suse.cz>
261 PR gcov-profile/84846
262 * gcov.c (print_usage): Add new -t option.
263 (process_args): Handle the option.
264 (generate_results): Use stdout as output when requested by
267 2018-05-18 Martin Liska <mliska@suse.cz>
269 PR gcov-profile/84846
270 * coverage.c (coverage_init): Write PWD to .gcno file.
271 * doc/gcov.texi: Document how working directory is printed.
272 * gcov-dump.c (dump_gcov_file): Print PWD.
273 * gcov.c (output_intermediate_file): Likewise.
274 (read_graph_file): Read PWD string.
275 (output_lines): Print PWD.
277 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
280 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
281 for retval and return false if all args to phi are zero.
283 2018-05-18 Richard Biener <rguenther@suse.de>
285 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
287 (evrp_dom_walker::before_dom_children): Call it.
289 2018-05-18 Richard Biener <rguenther@suse.de>
291 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
292 results when processing array refs with variable index.
294 2018-05-18 Toon Moene <toon@moene.org>
296 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
297 directly after that of -floop-interchange. Indicate that both
298 options are enabled by default when specifying -O3.
300 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
302 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
303 iterator. Delete separate integer-mode vec_set<mode> expander.
304 (aarch64_simd_vec_setv2di): Delete.
305 (vec_setv2di): Delete.
306 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
307 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
308 the "w, r" alternative.
310 2018-05-18 Martin Liska <mliska@suse.cz>
312 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
313 * tree-pass.h (make_pass_lower_switch_O0): New function.
314 * tree-switch-conversion.c (node_has_low_bound): Remove.
315 (node_has_high_bound): Likewise.
316 (node_is_bounded): Likewise.
317 (class pass_lower_switch): Make it a template type and create
319 (pass_lower_switch::execute): Add template argument.
320 (make_pass_lower_switch): New function.
321 (make_pass_lower_switch_O0): New function.
322 (do_jump_if_equal): Remove.
323 (emit_case_nodes): Simplify to just handle all 3 cases and leave
324 all the hard work to tree optimization passes.
326 2018-05-18 Martin Liska <mliska@suse.cz>
328 * dbgcnt.c (limit_low): Renamed from limit.
329 (limit_high): New variable.
330 (dbg_cnt_is_enabled): Check for upper limit.
331 (dbg_cnt): Adjust dumping.
332 (dbg_cnt_set_limit_by_index): Add new argument for high
334 (dbg_cnt_set_limit_by_name): Likewise.
335 (dbg_cnt_process_single_pair): Parse new format.
336 (dbg_cnt_process_opt): Use strtok.
337 (dbg_cnt_list_all_counters): Remove 'value' and add
339 * doc/invoke.texi: Document changes.
341 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
343 * doc/sourcebuild.texi (scalar_all_fma): Document.
344 * tree.def (FMA_EXPR): Delete.
345 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
346 * internal-fn.c (ternary_direct): New macro.
347 (expand_ternary_optab_fn): Likewise.
348 (direct_ternary_optab_supported_p): Likewise.
349 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
350 * builtins.c (fold_builtin_fma): Delete.
351 (fold_builtin_3): Don't call it.
352 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
353 * expr.c (expand_expr_real_2): Likewise.
354 * fold-const.c (operand_equal_p): Likewise.
355 (fold_ternary_loc): Likewise.
356 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
357 * gimple.c (DEFTREECODE): Likewise.
358 * gimplify.c (gimplify_expr): Likewise.
359 * optabs-tree.c (optab_for_tree_code): Likewise.
360 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
361 * tree-eh.c (operation_could_trap_p): Likewise.
362 (stmt_could_throw_1_p): Likewise.
363 * tree-inline.c (estimate_operator_cost): Likewise.
364 * tree-pretty-print.c (dump_generic_node): Likewise.
365 (op_code_prio): Likewise.
366 * tree-ssa-loop-im.c (stmt_cost): Likewise.
367 * tree-ssa-operands.c (get_expr_operands): Likewise.
368 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
369 * fold-const-call.h (fold_fma): Delete.
370 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
371 CFN_FNMA and CFN_FNMS.
373 * genmatch.c (combined_fn): New enum.
374 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
375 (commutative_op): New function.
376 (commutate): Use it. Handle more than 2 operands.
377 (dt_operand::gen_gimple_expr): Use commutative_op.
378 (parser::parse_expr): Allow :c to be used with non-binary
379 operators if the commutative operand is known.
380 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
381 CFN_FMS, CFN_FNMA and CFN_FNMS.
382 (backprop::process_assign_use): Remove FMA_EXPR handling.
383 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
384 (gen_hsa_fma): New function.
385 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
386 IFN_FNMA and IFN_FNMS.
387 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
388 * gimple-fold.h (follow_all_ssa_edges): Declare.
389 * gimple-fold.c (follow_all_ssa_edges): New function.
390 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
391 gimple_build interface and use follow_all_ssa_edges to fold the result.
392 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
393 instead of checking for optabs directly.
394 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
395 rather than FMA_EXPRs.
396 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
397 call to IFN_FMA instead of an FMA_EXPR.
399 2018-05-17 Jim Wilson <jimw@sifive.com>
401 * expr.c (do_tablejump): When converting index to Pmode, if we have a
402 sign extended promoted subreg, and the range does not have the sign bit
403 set, then do a sign extend.
405 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
406 test, check for sign extended subreg and/or constant operands, and
407 do a sign extend in that case.
409 2018-05-17 Steve Ellcey <sellcey@cavium.com>
411 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
412 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
414 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
415 Change logics_shift_reg to logics_shift_imm.
416 (thunderx2t99_fp_loadpair_basic): Delete.
417 (thunderx2t99_fp_storepair_basic): Delete.
418 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
419 (thunderx2t99_asimd_polynomial): Delete.
420 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
421 and neon_fp_mul_d_scalar_q.
422 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
423 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
424 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
425 (thunderx2t99_asimd_lut): Add missing tbl types.
426 (thunderx2t99_asimd_ext): Delete.
427 (thunderx2t99_asimd_load1_1_mult): Delete.
428 (thunderx2t99_asimd_load1_2_mult): Delete.
429 (thunderx2t99_asimd_load1_ldp): New.
430 (thunderx2t99_asimd_load1): New.
431 (thunderx2t99_asimd_load2): Add missing *load2* types.
432 (thunderx2t99_asimd_load3): New.
433 (thunderx2t99_asimd_load4): New.
434 (thunderx2t99_asimd_store1_1_mult): Delete.
435 (thunderx2t99_asimd_store1_2_mult): Delete.
436 (thunderx2t99_asimd_store2_mult): Delete.
437 (thunderx2t99_asimd_store2_onelane): Delete.
438 (thunderx2t99_asimd_store_stp): New.
439 (thunderx2t99_asimd_store1): New.
440 (thunderx2t99_asimd_store2): New.
441 (thunderx2t99_asimd_store3): New.
442 (thunderx2t99_asimd_store4): New.
444 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
446 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
447 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
449 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
450 Segher Boessenkool <segher@kernel.crashing.org>
453 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
456 2018-05-17 Richard Biener <rguenther@suse.de>
458 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
459 for pruning loop and prune defs feeding only already visited PHIs.
461 2018-05-17 Richard Biener <rguenther@suse.de>
463 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
465 2018-05-17 Bin Cheng <bin.cheng@arm.com>
466 Richard Biener <rguenther@suse.de>
468 PR tree-optimization/85793
469 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
470 for VMAT_ELEMENTWISE.
472 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
474 * internal-fn.h (lookup_internal_fn): Declare
475 * internal-fn.c (lookup_internal_fn): New function.
476 * gimple.c (gimple_build_call_from_tree): Handle calls to
478 * gimple-pretty-print.c (dump_gimple_call): Print "." before
479 internal function names.
480 * tree-pretty-print.c (dump_generic_node): Likewise.
481 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
483 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
485 * gimple-fold.h (gimple_build): Make the function forms take
486 combined_fn rather than built_in_function.
487 (gimple_simplify): Likewise.
488 * gimple-match-head.c (gimple_simplify): Likewise.
489 * gimple-fold.c (gimple_build): Likewise.
490 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
491 rather than gimple_build_call_internal.
492 (get_initial_defs_for_reduction): Likewise.
493 (vect_create_epilog_for_reduction): Likewise.
494 (vectorizable_live_operation): Likewise.
496 2018-05-17 Martin Liska <mliska@suse.cz>
498 * gimple-ssa-sprintf.c (format_directive): Do not use
499 space in between 'G_' and '('.
501 2018-05-17 Jakub Jelinek <jakub@redhat.com>
504 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
505 even if the mask is not all ones.
508 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
510 (ix86_gimple_fold_builtin): Likewise.
513 * config/i386/i386.c: Include tree-vector-builder.h.
514 (ix86_vector_shift_count): New function.
515 (ix86_fold_builtin): Fold shift builtins by scalar count.
516 (ix86_gimple_fold_builtin): Likewise.
518 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
519 _mm512_setzero): New intrinsics.
521 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
522 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
524 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
525 code generation for cases where splatting a value is not useful.
526 * simplify-rtx.c (simplify_ternary_operation): Simplify
527 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
528 mode to a vec_concat.
530 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
532 * config.gcc: Support "goldmont-plus".
533 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
535 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
536 PROCESSOR_GOLDMONT_PLUS.
537 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
538 (processor_target_table): Add "goldmont-plus".
539 (PTA_GOLDMONT_PLUS): Define.
540 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
541 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
542 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
543 (fold_builtin_cpu): Add "goldmont-plus".
544 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
545 (ix86_option_override_internal): Add "goldmont-plus".
546 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
547 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
548 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
549 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
551 2018-05-17 Richard Biener <rguenther@suse.de>
553 PR tree-optimization/85757
554 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
555 remove defs that only feed that PHI from further processing.
557 2018-05-16 Jim Wilson <jimw@sifive.com>
559 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
561 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
563 2018-05-16 Mark Wielaard <mark@klomp.org>
565 * dwarf2out.c (count_index_strings): New function.
566 (output_indirect_strings): Call count_index_strings and generate
567 header for dwarf_version >= 5.
569 2018-05-16 Mark Wielaard <mark@klomp.org>
571 * dwarf2out.c (dwarf_FORM): New function.
572 (set_indirect_string): Use dwarf_FORM.
573 (reset_indirect_string): Likewise.
574 (size_of_die): Likewise.
575 (value_format): Likewise.
576 (output_die): Likewise.
577 (add_skeleton_AT_string): Likewise.
578 (output_macinfo_op): Likewise.
579 (index_string): Likewise.
580 (output_index_string_offset): Likewise.
581 (output_index_string): Likewise.
582 (count_index_strings): Likewise.
584 2018-05-16 Carl Love <cel@us.ibm.com>
586 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
587 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
589 2018-05-16 Martin Jambor <mjambor@suse.cz>
591 * ipa-prop.c (ipa_free_all_edge_args): Remove.
592 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
594 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
596 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
597 (fnma<mode>4): Likewise.
598 (fms<mode>4): Likewise.
599 (fnms<mode>4): Likewise.
600 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
601 (aarch64_fnma<mode>4): Likewise.
602 (aarch64_fms<mode>4): Likewise.
603 (aarch64_fnms<mode>4): Likewise.
604 (aarch64_fnmadd<mode>4): Likewise.
606 2018-05-16 Jason Merrill <jason@redhat.com>
608 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
610 2018-05-16 Richard Biener <rguenther@suse.de>
612 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
613 (dump_stmt_cost): Declare.
614 (add_stmt_cost): Dump cost we add.
615 (add_stmt_costs): New function.
616 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
618 (vect_analyze_stmt): Adjust prototype.
619 (vectorizable_condition): Likewise.
620 (vectorizable_live_operation): Likewise.
621 (vectorizable_reduction): Likewise.
622 (vectorizable_induction): Likewise.
623 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
624 cost vector to pass to vectorizable_ and record afterwards.
625 (vect_model_reduction_cost): Take cost vector argument and adjust.
626 (vect_model_induction_cost): Likewise.
627 (vectorizable_reduction): Likewise.
628 (vectorizable_induction): Likewise.
629 (vectorizable_live_operation): Likewise.
630 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
631 SLP_TREE_NUMBER_OF_VEC_STMTS.
632 (vect_analyze_slp_cost_1): Remove.
633 (vect_analyze_slp_cost): Likewise.
634 (vect_slp_analyze_node_operations): Take visited args and
635 a target cost vector. Avoid processing already visited stmt sets.
636 (vect_slp_analyze_operations): Use a local cost vector to gather
637 costs and register those of non-discarded instances.
638 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
639 (vect_schedule_slp_instance): Remove copying of
640 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
642 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
643 adding cost. Record cost entry location.
644 (vect_prologue_cost_for_slp_op): Function to compute cost of
645 a constant or invariant generated for SLP vect in the prologue,
646 split out from vect_analyze_slp_cost_1.
647 (vect_model_simple_cost): Make static. Adjust for SLP costing.
648 (vect_model_promotion_demotion_cost): Likewise.
649 (vect_model_store_cost): Likewise, make static.
650 (vect_model_load_cost): Likewise.
651 (vectorizable_bswap): Add cost vector arg and adjust.
652 (vectorizable_call): Likewise.
653 (vectorizable_simd_clone_call): Likewise.
654 (vectorizable_conversion): Likewise.
655 (vectorizable_assignment): Likewise.
656 (vectorizable_shift): Likewise.
657 (vectorizable_operation): Likewise.
658 (vectorizable_store): Likewise.
659 (vectorizable_load): Likewise.
660 (vectorizable_condition): Likewise.
661 (vectorizable_comparison): Likewise.
662 (can_vectorize_live_stmts): Likewise.
663 (vect_analyze_stmt): Likewise.
664 (vect_transform_stmt): Adjust calls to vectorizable_*.
665 * tree-vectorizer.c: Include gimple-pretty-print.h.
666 (dump_stmt_cost): New function.
668 2018-05-16 Richard Biener <rguenther@suse.de>
670 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
671 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
672 * tree-ssa-dse.c: Include tree-ssa-loop.h.
673 (check_name): New callback.
674 (dse_classify_store): Track cycles via a visited bitmap of PHI
675 defs and simplify handling of in-loop and across loop dead stores
676 and properly fail for loop-variant refs. Handle byte-tracking with
677 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
680 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
682 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
683 (vect_get_mask_type_for_stmt): Likewise.
684 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
686 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
687 to determine the statement's vector type and the vector type that
688 should be used for calculating nunits. Deal with cases in which
689 the type has to be deferred.
690 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
691 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
692 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
693 (vect_determine_vf_for_stmt): New functions, split out from...
694 (vect_determine_vectorization_factor): ...here.
695 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
696 (vect_get_mask_type_for_stmt): New functions, split out from
697 vect_determine_vectorization_factor.
699 2018-05-16 Richard Biener <rguenther@suse.de>
701 * tree-cfg.c (verify_gimple_assign_ternary): Properly
702 verify the [VEC_]COND_EXPR embedded comparison.
704 2018-05-15 Martin Sebor <msebor@redhat.com>
706 PR tree-optimization/85753
707 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
708 RECORD_TYPE in addition to ARRAY_TYPE.
710 2018-05-15 Martin Sebor <msebor@redhat.com>
713 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
715 2018-05-15 Richard Biener <rguenther@suse.de>
717 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
718 add by_clobber_p one. Change algorithm to collect all defs
719 representing uses we need to walk and try reducing them to
720 a single one before failing.
721 (dse_dom_walker::dse_optimize_stmt): Adjust.
723 2018-05-13 Mark Wielaard <mark@klomp.org>
725 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
726 (size_of_loc_descr): Likewise.
727 (output_loc_operands): Likewise.
728 (output_loc_operands_raw): Likewise.
729 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
730 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
731 (hash_loc_operands): Likewise.
732 (compare_loc_operands): Likewise.
734 2018-05-14 Mark Wielaard <mark@klomp.org>
736 * dwarf2out.c (count_index_addrs): New function.
737 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
739 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
741 PR tree-optimization/83648
742 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
743 return value as malloc candidate.
745 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
748 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
749 as true in call to suggest_attribute.
751 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
753 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
756 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
758 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
760 (aarch64_crypto_aesd_fused): Likewise.
762 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
764 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
765 (movsi_aarch64): Likewise.
766 (load_pairsi): Likewise.
767 (load_pairdi): Likewise.
768 (store_pairsi): Likewise.
769 (store_pairdi): Likewise.
770 (load_pairsf): Likewise.
771 (load_pairdf): Likewise.
772 (store_pairsf): Likewise.
773 (store_pairdf): Likewise.
774 (zero_extend): Likewise.
775 (trunc): Swap alternatives.
776 (fcvt_target): Add '?' to prefer w over r.
778 2018-05-14 Jakub Jelinek <jakub@redhat.com>
781 * config/i386/i386.md: Disallow non-commutative arithmetics in
782 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
783 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
784 in the peephole2 before it.
786 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
788 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
789 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
790 (ix86_handle_option): Handle -mcldemote.
791 * config.gcc: New header.
792 * config/i386/cldemoteintrin.h: New file.
793 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
794 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
796 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
797 OPTION_MASK_ISA_CLDEMOTE.
798 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
799 (ix86_valid_target_attribute_inner_p): Ditto.
800 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
801 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
802 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
803 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
804 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
806 * config/i386/i386.opt: Add -mcldemote.
807 * config/i386/x86intrin.h: New header.
808 * doc/invoke.texi: Add -mcldemote.
810 2018-05-14 Richard Biener <rguenther@suse.de>
812 * doc/match-and-simplify.texi: Adjust :s documentation.
814 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
816 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
817 intended memcpy size.
818 (REORDER_45): Likewise.
820 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
823 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
824 * vec.c (qsort_chk): Use gcc_qsort.
825 * Makefile.in (OBJS-libcommon): Add sort.o.
826 (build/sort.o): New target. Use it...
827 (BUILD_RTL): ... here, and...
828 (build/gencfn-macros): ... here, and...
829 (build/genmatch): ... here.
831 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
832 Chung-Ju Wu <jasonwucj@gmail.com>
834 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
835 * config/nds32/nds32-graywolf.md: New file.
836 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
837 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
839 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
840 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
841 * config/nds32/nds32.md (pipeline_model): Add graywolf.
842 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
843 * config/nds32/pipelines.md: Include n15 settings.
845 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
846 Chung-Ju Wu <jasonwucj@gmail.com>
848 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
849 * config/nds32/nds32-n13.md: New file.
850 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
851 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
853 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
854 * config/nds32/nds32.md (pipeline_model): Add n13.
855 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
856 * config/nds32/pipelines.md: Include n13 settings.
858 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
859 Chung-Ju Wu <jasonwucj@gmail.com>
861 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
862 * config/nds32/nds32-n10.md: New file.
863 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
864 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
866 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
867 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
868 * config/nds32/nds32.md (pipeline_model): Add n10.
869 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
870 * config/nds32/pipelines.md: Include n10 settings.
872 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
873 Kito Cheng <kito.cheng@gmail.com>
874 Chung-Ju Wu <jasonwucj@gmail.com>
876 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
877 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
878 Add enum values for DSP extension instructions.
879 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
881 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
882 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
884 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
885 * config/nds32/nds32-dspext.md: New file for DSP implementation.
886 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
887 * config/nds32/nds32-intrinsic.md: Likewise.
888 * config/nds32/nds32_intrinsic.h: Likewise.
889 * config/nds32/nds32-md-auxiliary.c: Likewise.
890 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
891 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
892 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
893 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
894 * config/nds32/nds32-protos.h: New declarations for DSP extension.
895 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
896 TYPE_DMAC in switch statement.
897 * config/nds32/nds32.c: New checking and implementation for DSP
898 extension instructions.
899 * config/nds32/nds32.h: Likewise.
900 * config/nds32/nds32.md: Likewise.
901 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
902 * config/nds32/predicates.md: Implement new predicates for DSP
905 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
907 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
908 Reformat alternatives and attributes so it is easier to identify
909 which constraints/attributes go with which instruction.
910 (mov<mode>_hardfloat32, FMOVE64): Likewise.
911 (mov<mode>_softfloat32, FMOVE64): Likewise.
912 (mov<mode>_hardfloat64, FMOVE64): Likewise.
913 (mov<mode>_softfloat64, FMOVE64): Likewise.
915 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
917 * doc/extend.texi (PowerPC Built-in Functions): Rename this
919 (Basic PowerPC Built-in Functions): The new name of the
920 subsection previously known as "PowerPC Built-in Functions".
921 (Basic PowerPC Built-in Functions Available on all Configurations):
923 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
924 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
925 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
926 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
928 2018-05-11 Martin Jambor <mjambor@suse.cz>
931 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
934 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
937 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
939 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
941 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
942 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
943 (ix86_handle_option): Handle -mwaitpkg.
944 * config.gcc: New header.
945 * config/i386/cpuid.h (bit_WAITPKG): New bit.
946 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
947 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
949 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
950 OPTION_MASK_ISA_WAITPKG.
951 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
952 (ix86_option_override_internal): Add PTA_WAITPKG.
953 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
954 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
956 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
957 __builtin_ia32_umwait and __builtin_ia32_tpause.
958 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
959 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
960 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
961 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
962 UNSPECV_TPAUSE): New.
963 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
964 * config/i386/i386.opt: Add -mwaitpkg.
965 * config/i386/waitpkgintrin.h: New file.
966 * config/i386/x86intrin.h: New header.
967 * doc/invoke.texi: Add -mwaitpkg.
969 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
972 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
974 (cortex-m0): Use armv6s-m isa.
975 (cortex-m0plus): Likewise.
976 (cortex-m1): Likewise.
977 (cortex-m0.small-multiply): Likewise.
978 (cortex-m0plus.small-multiply): Likewise.
979 (cortex-m1.small-multiply): Likewise.
981 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
982 Jakub Jelinek <jakub@redhat.com>
984 PR tree-optimization/85692
985 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
986 source permute as well.
988 2018-05-11 Martin Liska <mliska@suse.cz>
991 * doc/extend.texi: Document LLVM style format for no_sanitize
994 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
996 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
997 mode_supports_vsx_dform_quad to mode_supports_dq_form.
998 (mode_supports_vsx_dform_quad): Likewise.
999 (mode_supports_vmx_dform): Move these functions to be next to the
1000 other mode_supports functions.
1001 (mode_supports_dq_form): Likewise.
1002 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1003 mode_supports_dq_form.
1004 (reg_offset_addressing_ok_p): Likewise.
1005 (offsettable_ok_by_alignment): Likewise.
1006 (rs6000_legitimate_offset_address_p): Likewise.
1007 (legitimate_lo_sum_address_p): Likewise.
1008 (rs6000_legitimize_address): Likewise.
1009 (rs6000_legitimize_reload_address): Likewise.
1010 (rs6000_secondary_reload_inner): Likewise.
1011 (rs6000_preferred_reload_class): Likewise.
1012 (rs6000_output_move_128bit): Likewise.
1014 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1016 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1017 Generate SImode target register for null target.
1018 <case IX86_BUILTIN_XGETBV>: Ditto.
1019 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1020 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1022 2018-05-10 Carl Love <cel@us.ibm.com>
1024 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1025 dcbtt and dcbtstt if operands[2] is 0.
1027 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1030 * config/i386/sse.md (usadv64qi): New expander.
1032 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1034 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1035 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1036 -maltivec=be support.
1037 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1038 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1039 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1040 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1041 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1042 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1043 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1044 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1045 altivec_vsumsws): Adjust.
1046 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1047 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1048 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1050 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1051 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1052 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1053 (altivec_lve<VI_char>x): Delete expand.
1054 (*altivec_lve<VI_char>x_internal): Rename to...
1055 (altivec_lve<VI_char>x): ... this.
1056 (altivec_lvxl_<mode>): Delete expand.
1057 (*altivec_lvxl_<mode>_internal): Rename to ...
1058 (altivec_lvxl_<mode>): ... this.
1059 (altivec_stvxl_<mode>): Delete expand.
1060 (*altivec_stvxl_<mode>_internal): Rename to ...
1061 (altivec_stvxl_<mode>): ... this.
1062 (altivec_stve<VI_char>x): Delete expand.
1063 (*altivec_stve<VI_char>x_internal): Rename to ...
1064 (altivec_stve<VI_char>x): ... this.
1065 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1066 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1067 reduc_plus_scal_<mode>): Adjust.
1068 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1070 (rs6000_cpu_cpp_builtins): Adjust.
1071 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1072 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1073 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1074 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1075 -maltivec=be support.
1076 (rs6000_split_vec_extract_var): Adjust.
1077 (rs6000_split_v4si_init): Adjust.
1078 (swap_selector_for_mode): Delete.
1079 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1080 altivec_expand_stvex_be): Delete.
1081 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1082 -maltivec=be support.
1083 (rs6000_gimple_fold_builtin): Ditto.
1084 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1086 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1087 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1088 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1089 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1090 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1091 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1092 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1093 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1094 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1095 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1096 anonymous split): Adjust.
1097 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1098 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1100 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1102 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1103 when --with-gxx-include-dir is also specified.
1104 * configure: Regenerate.
1106 2018-05-09 Jim Wilson <jimw@sifive.com>
1109 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1110 * config/riscv/t-withmultilib: New.
1111 * config/riscv/withmultilib.h: New.
1112 * doc/install.texi: Document RISC-V --with-multilib-list support.
1114 2018-05-09 Richard Biener <rguenther@suse.de>
1116 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1118 (vect_bb_vectorization_profitable_p): Adjust. Compute
1119 actual scalar cost using the cost vector and the add_stmt_cost
1122 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1124 PR rtl-optimization/85645
1125 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1126 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1128 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1130 PR rtl-optimization/85645
1131 * shrink-wrap.c (spread_components): Return a boolean saying if
1132 anything was changed.
1133 (try_shrink_wrapping_separate): Iterate spread_components until
1134 nothing changes anymore.
1136 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1138 PR rtl-optimization/85645
1139 * regrename.c (build_def_use): Also kill the chains that include the
1140 destination of a REG_CFA_REGISTER note.
1142 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1144 PR rtl-optimization/85645
1145 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1146 insn that has a REG_CFA_REGISTER note.
1148 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1150 * cfgexpand.c (expand_clobber): New function.
1151 (expand_gimple_stmt_1): Use it.
1152 * tree-vect-stmts.c (vect_clobber_variable): New function,
1154 (vectorizable_simd_clone_call): ...here.
1155 (vectorizable_store): Emit a clobber either side of an
1156 IFN_STORE_LANES sequence.
1157 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1159 2018-05-09 Tom de Vries <tom@codesourcery.com>
1162 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1163 (define_insn "trap_if_false"): Add exit after trap.
1165 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1167 PR rtl-optimization/85638
1168 * bb-reorder.c: Include common/common-target.h.
1169 (create_forwarder_block): New function extracted from...
1170 (fix_up_crossing_landing_pad): ...here. Rename into...
1171 (dw2_fix_up_crossing_landing_pad): ...this.
1172 (sjlj_fix_up_crossing_landing_pad): New function.
1173 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1174 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1175 from both partitions and exit the loop after one iteration.
1177 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1180 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1182 (Basic PowerPC Built-in Functions): The new name of the
1183 subsection previously known as "PowerPC Built-in Functions".
1184 (Basic PowerPC Built-in Functions Available on all Configurations):
1186 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1188 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1190 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1192 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1195 2018-05-08 Jim Wilson <jimw@sifive.com>
1197 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1198 (LD_EMUL_SUFFIX): New.
1199 (LINK_SPEC): Use it.
1201 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 Jakub Jelinek <jakub@redhat.com>
1221 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1222 after cmpelim optimization.
1224 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1226 * config.gcc: Support "goldmont".
1227 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1228 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1230 * config/i386/i386.c (m_GOLDMONT): Define.
1231 (processor_target_table): Add "goldmont".
1232 (PTA_GOLDMONT): Define.
1233 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1234 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1235 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1236 (fold_builtin_cpu): Add "goldmont".
1237 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1238 (ix86_option_override_internal): Add "goldmont".
1239 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1240 (processor_type): Add PROCESSOR_GOLDMONT.
1241 * config/i386/i386.md: Add CPU "glm".
1242 * config/i386/glm.md: New file.
1243 * config/i386/x86-tune.def: Add m_GOLDMONT.
1244 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1246 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1249 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1251 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1252 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1253 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1256 * config/i386/i386.c (ix86_fold_builtin): Handle
1257 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1260 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1261 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1263 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1266 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1267 (check_arch): Likewise.
1268 (check_fpu): Return the result rather than printing it.
1269 (end arch): Fix operator precedence.
1270 (end cpu): Likewise.
1271 (END): Print the result from check_fpu.
1273 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1274 Alan Hayward <alan.hayward@arm.com>
1275 David Sherwood <david.sherwood@arm.com>
1277 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1278 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1279 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1280 (*fcmuo<mode>_and): New patterns.
1282 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1284 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1285 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1286 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1287 (cmp_op, sve_imm_con): New code attributes.
1288 (SVE_COND_INT_CMP, imm_con): Delete.
1289 (cmp_op): Remove above unspecs from int attribute.
1290 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1292 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1293 comparison-specific unspecs.
1294 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1295 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1296 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1297 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1298 (*vec_fcm<cmp_op><mode>): Rename to...
1299 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1300 (*vec_fcmuo<mode>): Rename to...
1301 (*fcmuo<mode>): ...this and adjust likewise.
1302 (*pred_fcm<cmp_op><mode>): New pattern.
1303 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1304 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1306 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1308 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1309 (aarch64_emit_sve_predicated_cond): New function.
1310 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1311 (aarch64_emit_unspec_cond_or): Replace with...
1312 (aarch64_emit_sve_or_conds): ...this new function. Use
1313 aarch64_emit_sve_ptrue_op for the individual comparisons and
1314 aarch64_emit_binop to OR them together.
1315 (aarch64_emit_inverted_unspec_cond): Replace with...
1316 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1317 aarch64_emit_sve_ptrue_op for the comparison and
1318 aarch64_emit_unop to invert the result.
1319 (aarch64_expand_sve_vec_cmp_float): Update after the above
1320 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1322 2018-05-07 Nathan Sidwell <nathan@acm.org>
1324 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1325 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1326 (Backwards Compatibility): Likewise.
1328 2018-05-07 Luis Machado <luis.machado@linaro.org>
1332 2018-05-07 Luis Machado <luis.machado@linaro.org>
1334 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1335 <prefetch_dynamic_strides>: New const bool field.
1336 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1337 prefetch_dynamic_strides.
1338 (exynosm1_prefetch_tune): Likewise.
1339 (thunderxt88_prefetch_tune): Likewise.
1340 (thunderx_prefetch_tune): Likewise.
1341 (thunderx2t99_prefetch_tune): Likewise.
1342 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1343 (aarch64_override_options_internal): Update to set
1344 PARAM_PREFETCH_DYNAMIC_STRIDES.
1345 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1346 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1347 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1348 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1349 prefetch-dynamic-strides setting.
1351 2018-05-07 Luis Machado <luis.machado@linaro.org>
1353 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1354 <minimum_stride>: New const int field.
1355 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1356 minimum_stride field.
1357 (exynosm1_prefetch_tune): Likewise.
1358 (thunderxt88_prefetch_tune): Likewise.
1359 (thunderx_prefetch_tune): Likewise.
1360 (thunderx2t99_prefetch_tune): Likewise.
1361 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1362 (aarch64_override_options_internal): Update to set
1363 PARAM_PREFETCH_MINIMUM_STRIDE.
1364 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1365 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1366 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1367 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1368 stride is constant and is below the minimum stride threshold.
1370 2018-05-07 Luis Machado <luis.machado@linaro.org>
1372 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1375 2018-05-07 Luis Machado <luis.machado@linaro.org>
1377 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1378 <prefetch_dynamic_strides>: New const bool field.
1379 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1380 prefetch_dynamic_strides.
1381 (exynosm1_prefetch_tune): Likewise.
1382 (thunderxt88_prefetch_tune): Likewise.
1383 (thunderx_prefetch_tune): Likewise.
1384 (thunderx2t99_prefetch_tune): Likewise.
1385 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1386 (aarch64_override_options_internal): Update to set
1387 PARAM_PREFETCH_DYNAMIC_STRIDES.
1388 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1389 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1390 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1391 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1392 prefetch-dynamic-strides setting.
1394 2018-05-07 Luis Machado <luis.machado@linaro.org>
1396 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1397 <minimum_stride>: New const int field.
1398 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1399 minimum_stride field.
1400 (exynosm1_prefetch_tune): Likewise.
1401 (thunderxt88_prefetch_tune): Likewise.
1402 (thunderx_prefetch_tune): Likewise.
1403 (thunderx2t99_prefetch_tune): Likewise.
1404 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1405 (aarch64_override_options_internal): Update to set
1406 PARAM_PREFETCH_MINIMUM_STRIDE.
1407 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1408 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1409 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1410 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1411 stride is constant and is below the minimum stride threshold.
1413 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1416 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1417 the type is addressable. Don't force op into register if it has
1420 2018-05-05 Roland McGrath <mcgrathr@google.com>
1423 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1424 any section for which we don't know a specific type it should have,
1425 regardless of name. Previously this was done only for the exact
1426 names ".init_array", ".fini_array", and ".preinit_array".
1427 (default_elf_asm_named_section): Add comment about
1428 relationship with default_section_type_flags and SECTION_NOTYPE.
1429 (get_section): Don't consider it a type conflict if one side has
1430 SECTION_NOTYPE and the other doesn't, as long as neither has the
1431 SECTION_BSS et al used in the default_section_type_flags logic.
1433 2018-05-05 Tom de Vries <tom@codesourcery.com>
1436 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1437 (workaround_barsyncs): New function.
1438 (nvptx_reorg): Use workaround_barsyncs.
1439 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1440 (define_expand "nvptx_membar_cta"): New define_expand.
1441 (define_insn "*nvptx_membar_cta"): New insn.
1443 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1445 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1446 To improve optimization opportunities.
1447 * builtin-types.def: The new needed builtin types for the above.
1449 2018-05-04 Richard Biener <rguenther@suse.de>
1451 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1452 * gimple-ssa-store-merging.c
1453 (imm_store_chain_info::output_merged_store): Remove redundant create,
1454 release split_store vector contents on failure.
1455 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1456 scalar stmt vector on cache hit.
1458 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1460 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1462 * config.gcc (powerpc-xilinx-eabi*): Remove.
1463 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1465 (fusion_addis_mem_combo_load): Ditto.
1466 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1468 (rs6000_cpu_cpp_builtins): Ditto.
1469 * config/rs6000/rs6000-linux.c
1470 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1471 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1472 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1474 (rs6000_setup_reg_addr_masks): Ditto.
1475 (rs6000_init_hard_regno_mode_ok): Ditto.
1476 (rs6000_option_override_internal): Ditto.
1477 (legitimate_lo_sum_address_p): Ditto.
1478 (rs6000_legitimize_address): Ditto.
1479 (rs6000_legitimize_reload_address): Ditto.
1480 (rs6000_legitimate_address_p): Ditto.
1481 (abi_v4_pass_in_fpr): Ditto.
1482 (setup_incoming_varargs): Ditto.
1483 (rs6000_gimplify_va_arg): Ditto.
1484 (rs6000_split_multireg_move): Ditto.
1485 (rs6000_savres_strategy): Ditto.
1486 (rs6000_emit_prologue_components): Ditto.
1487 (rs6000_emit_epilogue_components): Ditto.
1488 (rs6000_emit_prologue): Ditto.
1489 (rs6000_emit_epilogue): Ditto.
1490 (rs6000_elf_file_end): Ditto.
1491 (rs6000_function_value): Ditto.
1492 (rs6000_libcall_value): Ditto.
1493 * config/rs6000/rs6000.h: Ditto.
1494 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1495 (TARGET_MINMAX): ... this. New.
1496 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1497 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1498 (*movsi_internal1_single): Delete.
1499 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1500 mfpu=, mxilinx-fpu): Delete.
1501 * config/rs6000/singlefp.h: Delete.
1502 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1503 * config/rs6000/t-rs6000: Ditto.
1504 * config/rs6000/t-xilinx: Delete.
1505 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1506 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1507 (VStype_simple): Delete.
1508 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1509 * config/rs6000/xfpu.h: Delete.
1510 * config/rs6000/xfpu.md: Delete.
1511 * config/rs6000/xilinx.h: Delete.
1512 * config/rs6000/xilinx.opt: Delete.
1513 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1514 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1516 2018-05-04 Tom de Vries <tom@codesourcery.com>
1519 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1522 2018-05-04 Richard Biener <rguenther@suse.de>
1525 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1526 (expand_complex_div_wide): Likewise.
1527 (expand_complex_operations_1): Likewise.
1528 (expand_complex_libcall): Preserve EH info of the original stmt.
1529 (tree_lower_complex): Handle removed blocks.
1530 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1531 on complex multiplication and division libcall builtins.
1533 2018-05-04 Richard Biener <rguenther@suse.de>
1536 * fold-const.c (negate_expr_p): Restrict negation of operand
1537 zero of a division to when we know that can happen without
1539 (fold_negate_expr_1): Likewise.
1541 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1544 * real.h (real_nextafter): Declare.
1545 * real.c (real_nextafter): New function.
1546 * fold-const-call.c (fold_const_nextafter): New function.
1547 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1548 CASE_CFN_NEXTTOWARD.
1549 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1550 even when arg1_mode is different from arg0_mode.
1552 2018-05-03 Nathan Sidwell <nathan@acm.org>
1554 * doc/extend.texi (Deprecated Features): Remove
1556 (Backwards Compatibility): Likewise.
1557 * doc/invoke.texi (C++ Language Options): Likewise.
1558 (C++ Dialect Options): Likewise.
1560 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1563 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1564 _mm512_mask_mullox_epi64): New intrinsics.
1566 2018-05-03 Tom de Vries <tom@codesourcery.com>
1569 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1570 dump files): Add offload-tree.
1572 2018-05-03 Richard Biener <rguenther@suse.de>
1574 PR tree-optimization/85615
1575 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1576 to loops not nested in BBs loop father to avoid creating multi-entry
1579 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1581 PR tree-optimization/70291
1582 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1583 arguments. Change return type to tree. Emit libcall as a new
1584 statement rather than replacing existing one when inplace_p is true.
1585 (expand_complex_multiplication_components): New function.
1586 (expand_complex_multiplication): Expand floating-point complex
1587 multiplication using the above.
1588 (expand_complex_division): Rename inner_type parameter to type.
1589 Update expand_complex_libcall call-site.
1590 (expand_complex_operations_1): Update expand_complex_multiplication
1591 and expand_complex_division call-sites.
1593 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1596 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1597 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1598 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1599 the highest significant bit of the shift count mask is clear. In
1600 check whether and[sq]i3 is needed verify that all significant bits
1601 of the shift count other than the highest are set.
1603 2018-05-02 Tom de Vries <tom@codesourcery.com>
1606 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1607 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1608 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1609 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1610 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1611 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1612 __builtin_goacc_parlevel_size.
1614 2018-05-02 Richard Biener <rguenther@suse.de>
1616 PR tree-optimization/85597
1617 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1618 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1621 2018-05-02 Tom de Vries <tom@codesourcery.com>
1624 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1625 dump files): Add ltrans-tree.
1627 2018-05-02 Tom de Vries <tom@codesourcery.com>
1630 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1631 dump files): Add wpa-ipa.
1633 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1635 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1636 powerpc*-*-linux*paired* target.
1637 * config/rs6000/750cl.h: Delete.
1638 * config/rs6000/paired.h: Delete.
1639 * config/rs6000/paired.md: Delete.
1640 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1642 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1643 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1644 comment. Remove paired float support.
1645 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1646 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1648 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1649 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1651 * config/rs6000/rs6000.c: Remove paired float support.
1652 (paired_expand_vector_init, paired_expand_vector_move,
1653 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1654 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1655 paired_expand_builtin, paired_expand_predicate_builtin,
1656 paired_init_builtins): Delete.
1657 * config/rs6000/rs6000.h: Remove paired float support.
1658 * config/rs6000/rs6000.md: Remove paired float support.
1659 (move_from_CR_ov_bit): Delete.
1660 * config/rs6000/rs6000.opt (mpaired): Delete.
1661 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1662 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1664 2018-05-02 Richard Biener <rguenther@suse.de>
1667 * gimplify.c (gimplify_save_expr): When in SSA form allow
1668 SAVE_EXPRs to compute to SSA vars.
1670 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1673 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1674 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1675 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1676 clobber operands[2], instead use a new pseudo. Formatting fixes.
1678 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1680 PR tree-optimization/85586
1681 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1682 exit early for statements in the same group if the accesses are
1685 2018-05-02 Tom de Vries <tom@codesourcery.com>
1688 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1691 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1693 PR tree-optimization/85143
1694 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1696 2018-05-01 Tom de Vries <tom@codesourcery.com>
1699 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1700 not found" error message.
1702 2018-05-01 Tom de Vries <tom@codesourcery.com>
1705 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1706 * vec.c (test_ordered_remove_if): New function.
1707 (vec_c_tests): Call test_ordered_remove_if.
1708 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1709 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1710 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1711 VEC_ORDERED_REMOVE_IF.
1713 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1715 PR tree-optimization/82665
1716 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1717 pointer subtraction where arguments come from a memchr call.
1719 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1721 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1722 --push-state --as-needed and --pop-state instead of --as-needed and
1723 --no-as-needed if ld supports it.
1724 * configure: Regenerated.
1727 * doc/install.texi2html: Replace _002d with - and _002a with * in
1728 generated html files using sed.
1730 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1733 * gcc-rich-location.c (blank_line_before_p): New function.
1734 (use_new_line): New function.
1735 (gcc_rich_location::add_fixit_insert_formatted): New function.
1736 * gcc-rich-location.h
1737 (gcc_rich_location::add_fixit_insert_formatted): New function.
1739 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1741 * selftest.c (assert_streq): Rename "expected" and "actual" to
1742 "val1" and "val2". Extend NULL-handling to cover both inputs
1743 symmetrically, while still requiring both to be non-NULL for a pass.
1744 * selftest.h (assert_streq): Rename "expected" and "actual" to
1746 (ASSERT_EQ): Likewise.
1747 (ASSERT_EQ_AT): Likewise.
1748 (ASSERT_KNOWN_EQ): Likewise.
1749 (ASSERT_KNOWN_EQ_AT): Likewise.
1750 (ASSERT_NE): Likewise.
1751 (ASSERT_MAYBE_NE): Likewise.
1752 (ASSERT_MAYBE_NE_AT): Likewise.
1753 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1754 the assertion to pass.
1755 (ASSERT_STREQ_AT): Likewise.
1757 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1759 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1760 interaction with -pie.
1762 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1764 * selftest.h: Fix alphabetization of per-source-file selftest
1767 2018-04-30 Jason Merrill <jason@redhat.com>
1769 PR c++/61982 - dead stores to destroyed objects.
1770 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1773 2018-04-30 Jason Merrill <jason@redhat.com>
1775 * tree.c (build_clobber): New.
1776 * tree.h: Declare it.
1777 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1779 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1781 * diagnostic-show-locus.c (layout::layout): Update for
1782 location_get_source_line returning a char_span.
1783 (struct char_span): Move to input.h.
1784 (struct correction): Update for fields in char_span becoming
1786 (struct source_line): Update for location_get_source_line
1787 returning a char_span.
1788 (layout::print_line): Likewise.
1789 * edit-context.c (edited_file::print_content): Likewise.
1790 (edited_file::print_diff_hunk): Likewise.
1791 (edited_file::print_run_of_changed_lines): Likewise.
1792 (edited_file::get_num_lines): Likewise.
1793 (edited_line::edited_line): Likewise.
1794 * final.c (asm_show_source): Likewise.
1795 * input.c (location_get_source_line): Convert return type
1796 from const char * to char_span, losing the final "line_len"
1798 (dump_location_info): Update for the above.
1799 (get_substring_ranges_for_loc): Likewise. Use a char_span
1800 when handling the literal within the line.
1801 (test_reading_source_line): Update for location_get_source_line
1802 returning a char_span.
1803 * input.h (class char_span): Move here from
1804 diagnostic-show-locus.c, converting from a struct to a class.
1805 Make data members private.
1806 (char_span::operator bool): New.
1807 (char_span::length): New.
1808 (char_span::get_buffer): New.
1809 (char_span::operator[]): New.
1810 (char_span::subspan): Make const.
1811 (char_span::xstrdup): New.
1812 (location_get_source_line): Convert return type from const char *
1813 to char_span, losing the final "line_size" param.
1815 2018-04-30 Jan Hubicka <jh@suse.cz>
1817 * lto-wrapper.c (ltrans_priorities): New static var.
1818 (cmp_priority): New.
1819 (run_gcc): Read priorities and if doing parallel build order
1820 the Makefile by them.
1822 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1824 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1826 2018-04-30 Richard Biener <rguenther@suse.de>
1828 * tree-cfg.c (verify_address): Remove base argument, add
1829 flag whether to check TREE_ADDRESSABLE and do that.
1830 (verify_expr): Remove.
1831 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1832 (verify_gimple_assign_single): Likewise.
1833 (verify_gimple_switch): Likewise.
1834 (verify_expr_location_1): Dereference tp once. Add (disabled)
1835 piece from verify_expr.
1836 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1838 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1840 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1842 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1844 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1845 (small_data_pattern): Likewise.
1846 (arc_rewrite_small_data): Likewise.
1847 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1848 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1849 (get_symbol_alignment): New function.
1850 (legitimate_small_data_address_p): Likewise.
1851 (legitimate_scaled_address): Update, call
1852 legitimate_small_data_address_p.
1853 (output_sdata): New static variable.
1854 (arc_print_operand): Update how we handle small data operands.
1855 (arc_print_operand_address): Likewise.
1856 (arc_legitimate_address_p): Update, use
1857 legitimate_small_data_address_p.
1858 (arc_rewrite_small_data_p): Remove.
1859 (arc_rewrite_small_data_1): Likewise.
1860 (arc_rewrite_small_data): Likewise.
1861 (small_data_pattern): Likewise.
1862 (compact_sda_memory_operand): Update to use
1863 legitimate_small_data_address_p and get_symbol_alignment.
1864 (prepare_move_operands): Don't rewite sdata pattern.
1865 (prepare_extend_operands): Remove.
1866 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1868 (zero_extendqisi2): Likewise.
1869 (zero_extendhisi2): Likewise.
1870 (extendqihi2): Likewise.
1871 (extendqisi2): Likewise.
1872 (extendhisi2): Likewise.
1876 * config/arc/constraints.md (Usd): Change it to memory constraint.
1878 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1880 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1881 as source of std instructions.
1882 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1883 allow 6-bit constants as source for store instructions.
1884 (movdi_insn): Update instruction pattern to allow 6-bit constants
1885 as source for store instructions.
1887 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1889 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1891 2018-04-30 Nathan Sidwell <nathan@acm.org>
1892 Sandra Loosemore <sandra@codesourcery.com>
1894 * dumpfile.c (dump_open): Allow '-' for stdout.
1895 * doc/invoke.texi (Developer Options): Document dump filename
1896 determination early. Document stdin/stdout selection.
1898 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1900 Microblaze Target: PIC data text relative
1902 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1903 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1905 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1906 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1907 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1908 New addressing mode for data-text relative position indepenedent code.
1909 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1910 'ADDRESS_SYMBOLIC_TXT_REL'.
1911 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1912 (microblaze_legitimate_pic_operand): Exclude function calls from
1913 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1914 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1916 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1917 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1918 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1919 for 'address + offset'.
1920 (microblaze_expand_prologue): Add new function prologue call for
1922 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1923 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1924 table in case of TARGET_PIC_DATA_TEXT_REL.
1925 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1926 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1927 Add new macros 'UNSPEC_TEXT',
1928 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1929 + exclude function calls from 'UNSPEC_PLT' in case of data text
1931 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1932 new target hook for generating address diff vector tables in case of
1934 * doc/tm.texi : Regenerate.
1935 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1936 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1937 of addr diff vector generation.
1938 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1939 target hook definition.
1940 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1941 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1942 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1943 option for data text relative.
1945 2018-04-30 Richard Biener <rguenther@suse.de>
1947 * tree-chrec.h (evolution_function_is_constant_p): Remove
1949 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1951 2018-04-30 Richard Biener <rguenther@suse.de>
1954 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1956 2018-04-30 Richard Biener <rguenther@suse.de>
1958 PR tree-optimization/28364
1959 PR tree-optimization/85275
1960 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1961 copying first exit test.
1963 2018-04-28 Mark Wielaard <mark@klomp.org>
1965 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1967 (dwarf_AT): Handle DW_AT_addr_base.
1968 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1970 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1973 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1974 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1975 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1976 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1978 2018-04-28 Richard Biener <rguenther@suse.de>
1980 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1981 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1982 to reflect use. Only add interesting stmts.
1984 2018-04-27 Martin Jambor <mjambor@suse.cz>
1987 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1988 the jump function allows for passing through aggregate values.
1990 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1992 * input.h (in_system_header_at): Convert from macro to inline
1994 (from_macro_expansion_at): Likewise.
1995 (from_macro_definition_at): Likewise.
1997 2018-04-27 Jeff Law <law@redhat.com>
1999 * config.gcc: Mark tile* targets as deprecated/obsolete.
2001 2018-04-27 Richard Biener <rguenther@suse.de>
2003 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2006 2018-04-27 Richard Biener <rguenther@suse.de>
2008 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2010 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2012 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2013 with Yd constraint. Set "preferred_for_speed" attribute from
2014 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2016 (*movdi_internal): Ditto.
2017 (movti_interunit splitters): Remove
2018 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2019 (movdi_interunit splitters): Ditto.
2020 * config/i386/constraints.md (Ye): Remove.
2021 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2023 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2026 * config/aarch64/constraints.md (Usg): Limit to 31.
2029 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2031 PR tree-optimization/85529
2032 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2033 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2034 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2035 zero extension or masking of the MSB bit.
2036 (optimize_range_tests): Add FIRST_BB argument, pass it through
2037 to optimize_range_tests_var_bound.
2038 (maybe_optimize_range_tests, reassociate_bb): Adjust
2039 optimize_range_tests callers.
2041 2018-04-26 Richard Biener <rguenther@suse.de>
2042 Jakub Jelinek <jakub@redhat.com>
2044 * cgraph.h (symbol_table): Just declare debug method here.
2045 * symtab.c (symbol_table::debug): Define.
2047 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2049 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2051 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2053 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2054 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2055 (*movdi_internal): Substitute Yi and Yj constraint with x
2056 and Ym and Yn constraint with y constraint. Update "isa"
2057 attribute and set "preferred_for_speed" attribute from
2058 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2059 (*movsi_internal): Ditto.
2060 (*movdf_internal): Ditto.
2061 (*movsf_internal): Ditto.
2062 (*zero_extendsidi2): Ditto.
2063 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2064 (sse2_loadld): Ditto.
2065 (*vec_extract<ssevecmodelower>_0): Ditto.
2066 (*vec_extractv4si_0_zext_sse4): Ditto.
2067 (vec_concatv2di): Ditto.
2068 (*vec_dup<mode>): Ditto.
2069 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2070 * config/i386/constraints.md (Yi): Remove.
2075 2018-04-26 Nathan Sidwell <nathan@acm.org>
2077 * dumpfile.c (dump_open): New.
2078 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2079 (dump_finish): Detect stdio/stderr by value not name.
2081 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2083 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2085 2018-04-26 Tom de Vries <tom@codesourcery.com>
2088 * config/nvptx/nvptx.c (verify_neutering_jumps)
2089 (verify_neutering_labels): New function
2090 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2092 2018-04-26 Tom de Vries <tom@codesourcery.com>
2095 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2096 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2099 2018-04-26 Richard Biener <rguenther@suse.de>
2100 Tom de Vries <tom@codesourcery.com>
2103 * lto-streamer-out.c (output_function): Fixup loops if required to match
2104 discovery done in the reader.
2106 2018-04-26 Richard Biener <rguenther@suse.de>
2108 PR tree-optimization/85116
2109 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2110 have a loop exit from the single latch predecessor. Remove
2111 case of header with just condition.
2112 (ch_base::copy_headers): Exclude infinite loops from any
2114 (pass_ch::execute): Record exits.
2116 2018-04-26 Richard Biener <rguenther@suse.de>
2118 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2119 prologue cost vector and pass it to vect_get_load_cost.
2120 (vect_get_peeling_costs_all_drs): Likewise.
2121 (vect_peeling_hash_get_lowest_cost): Likewise.
2122 (vect_enhance_data_refs_alignment): Likewise.
2124 2018-04-26 Richard Biener <rguenther@suse.de>
2127 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2128 checking of integer<->pointer conversions.
2129 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2130 sign-/zero-extending pointer types.
2131 (expand_omp_for_static_chunk): Likewise.
2133 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2134 Jean Lee <xiaoyur347@gmail.com>
2136 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2137 (TARGET_ASAN_SHADOW_OFFSET): Define.
2138 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2139 true for -fsanitize=address.
2141 2018-04-25 Mark Wielaard <mark@klomp.org>
2143 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2146 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2148 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2149 than "alu", remove explicit "memory" and "imm_disp" attributes.
2150 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2153 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2154 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2155 gen_lowpart_no_emit.
2157 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2160 * config/i386/i386.c (ix86_expand_builtin): Change memory
2161 operand to XI, extend p0 to Pmode.
2162 * config/i386/i386.md: Change unspec volatile and operand
2163 1 mode to XI, change operand 0 mode to P.
2165 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2167 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2168 GET_MODE_MASK before any checking.
2169 (nds32_can_use_bset_p): Likewise.
2170 (nds32_can_use_btgl_p): Likewise.
2172 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2174 * config/nds32/nds32-doubleword.md: New define_split pattern for
2175 illegal register number.
2177 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2179 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2181 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2183 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2185 2018-04-25 Richard Biener <rguenther@suse.de>
2187 * lto-streamer.h (LTO_major_version): Bump to 8.
2189 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2191 * BASE-VER: Set to 9.0.0.
2193 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2195 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2196 in __abskf2 and __powikf2.
2198 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2201 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2202 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2203 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2204 Use the above on operand 2. Reindent.
2205 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2207 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2210 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2212 * config/i386/i386.opt (mcet): Removed.
2213 * doc/install.texi: Remove -mcet documentation.
2214 * doc/invoke.texi: Likewise.
2216 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2219 * doc/install.texi: Remove -mcet from bootstrap-cet.
2221 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2224 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2225 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2229 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2230 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2231 containing a CONST_VECTOR.
2233 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2235 * doc/install.texi: Update newlib dependency for nvptx.
2237 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2240 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2241 instead of INTVAL when shifting x left.
2243 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2245 PR tree-optimization/85478
2246 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2247 vect_grouped_store_supported for single element vectors.
2249 2018-04-24 Richard Biener <rguenther@suse.de>
2252 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2253 load cost increase to the case of non-constant step.
2255 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2258 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2259 destination if any_malformed_asm.
2261 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2264 * expr.c (store_field): In the bitfield case, if the value comes from
2265 a function call and is returned in registers by means of a PARALLEL,
2266 do not change the mode of the temporary unless BLKmode and VOIDmode.
2268 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2270 PR rtl-optimization/85423
2271 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2272 dependencies to debug insns when the previous insn is non-debug.
2274 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2276 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2277 enums into a single definition.
2278 (fls): Fix predicates and printing.
2281 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2283 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2284 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2285 and short u6 immediate.
2286 (check_if_valid_sleep_operand): Remove.
2287 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2289 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2291 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2292 flag_always_save_lp condition.
2293 * config/nds32/nds32.opt (malways-save-lp): New option.
2295 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2297 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2298 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2299 * config/nds32/nds32.h
2300 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2301 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2303 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2305 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2306 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2308 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2309 Chung-Ju Wu <jasonwucj@gmail.com>
2311 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2313 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2314 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2316 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2318 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2320 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2322 * config/nds32/nds32-protos.h (nds32_data_alignment,
2323 nds32_local_alignment): Declare.
2324 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2325 nds32_local_alignment): New functions.
2326 (TARGET_CONSTANT_ALIGNMENT): Define.
2327 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2329 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2331 * config/nds32/nds32.c
2332 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2333 (TARGET_MODES_TIEABLE_P): Likewise.
2335 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2337 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2340 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2341 Chung-Ju Wu <jasonwucj@gmail.com>
2343 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2344 for unaligned access.
2345 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2347 * config/nds32/nds32-intrinsic.md: Likewise.
2348 * config/nds32/nds32_intrinsic.h: Likewise.
2349 * config/nds32/nds32.h (nds32_builtins): Likewise.
2350 * config/nds32/nds32.opt (munaligned-access): New option.
2351 * config/nds32/nds32.c (nds32_asm_file_start): Display
2352 flag_unaligned_access status.
2354 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2356 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2357 -mno-relax is present.
2358 * config/riscv/linux.h (LINK_SPEC): Ditto.
2360 2018-04-20 Martin Sebor <msebor@redhat.com>
2363 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2365 (gimple_fold_builtin_stxcpy_chk): Same.
2366 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2368 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2371 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2372 __powikf2 when long double is IEEE 128-bit.
2374 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2376 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2377 step to make sure stack always aligned.
2379 2018-04-20 Carl Love <cel@us.ibm.com>
2382 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2383 size check for arg0.
2385 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2386 Tom de Vries <tom@codesourcery.com>
2389 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2390 Emit insns for calls too.
2391 (nvptx_find_par): Always look for worker-level predecessor insn.
2392 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2394 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2395 (nvptx_process_pars): Propagate frames for calls.
2397 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2400 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2402 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2403 (ix86_handle_option): Don't handle OPT_mibt.
2404 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2406 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2408 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2409 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2410 (ix86_target_macros): Define __CET__ with flag_cf_protection
2411 for -fcf-protection.
2412 * config/i386/i386.c (isa2_opts): Remove -mibt.
2413 * config/i386/i386.h (TARGET_IBT): Removed.
2414 (TARGET_IBT_P): Likewise.
2415 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2416 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2417 * config/i386/i386.opt (mcet): Update help message.
2420 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2421 -mcet as an alias for -mshstk.
2423 2018-04-20 Richard Biener <rguenther@suse.de>
2426 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2427 complexity by forcing a single use of the multiply operand.
2429 2018-04-20 Martin Jambor <mjambor@suse.cz>
2432 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2433 recursion dependency to only apply to non-clones.
2435 2018-04-20 Martin Jambor <mjambor@suse.cz>
2438 * ipa-cp.c (create_specialized_node): Check that clones of
2439 self-recursive edges exist during IPA-CP.
2441 2018-04-19 Toon Moene <toon@moene.org>
2443 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2446 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2448 PR tree-optimization/85467
2449 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2450 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2451 VECTOR_CST element to type.
2453 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2456 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2457 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2458 (builtin_longjmp): Likewise.
2459 (save_stack_nonlocal): New pattern.
2460 (restore_stack_nonlocal): Likewise.
2462 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2465 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2466 Replace ASM_OUTPUT_LABEL with fprintf.
2468 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2471 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2472 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2473 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2474 define __IBT__ and __SHSTK__ for -fcf-protection.
2475 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2477 (ix86_trampoline_init): Likewise.
2478 (x86_output_mi_thunk): Likewise.
2479 (ix86_notrack_prefixed_insn_p): Likewise.
2480 (ix86_option_override_internal): Don't disallow -fcf-protection.
2481 * config/i386/i386.md (rdssp<mode>): Also enable for
2483 (incssp<mode>): Likewise.
2484 (nop_endbr): Likewise.
2485 * config/i386/i386.opt (mcet): Change help message to built-in
2489 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2490 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2491 enable CET built-in functions.
2493 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2495 * common/config/i386/i386-common.c
2496 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2497 OPTION_MASK_ISA_MOVDIRI_UNSET,
2498 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2499 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2500 * config.gcc (movdirintrin.h): New header.
2501 * config/i386/cpuid.h (bit_MOVDIRI,
2502 bit_MOVDIR64B): New bits.
2503 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2505 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2506 (VOID, PVOID, PCVOID)): New function types.
2507 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2508 __builtin_ia32_directstoreu_u64,
2509 __builtin_ia32_movdir64b): New builtins.
2510 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2511 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2513 (ix86_valid_target_attribute_inner_p): Ditto.
2514 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2515 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2516 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2517 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2518 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2519 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2520 (movdiri<mode>, movdir64b_<mode>): New.
2521 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2522 * config/i386/immintrin.h: Include movdirintrin.h.
2523 * config/i386/movdirintrin.h: New file.
2524 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2526 2018-04-19 Richard Biener <rguenther@suse.de>
2529 * cfg.c (clear_bb_flags): When loop state says we have
2530 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2532 2018-04-19 Richard Biener <rguenther@suse.de>
2534 PR tree-optimization/84737
2535 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2536 copying restrict info.
2537 (vect_setup_realignment): Use it.
2538 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2539 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2540 the first DR to all generated stores.
2541 (vectorizable_load): Likewise for loads.
2543 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2545 PR tree-optimization/85446
2546 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2547 the integral and pointer types to have the same precision.
2549 * doc/install.texi: Document --disable-cet being the default and
2552 2018-04-18 Martin Liska <mliska@suse.cz>
2554 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2557 2018-04-18 Martin Liska <mliska@suse.cz>
2560 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2563 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2564 arguments if they are comparable.
2566 2018-04-18 Martin Liska <mliska@suse.cz>
2569 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2572 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2575 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2578 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2579 ENDBR after calling __morestack.
2581 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2584 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2585 by using gcc_base_ver to generate a gcc_driver_version, and use
2586 it when generating GCC_DRIVER_NAME.
2587 * configure: Regenerate.
2589 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2592 * config.gcc: Obsolete powerpc*-*-*spe*.
2594 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2597 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2598 (stabstr_D): Change type of unum from unsigned int to
2599 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2602 2018-04-17 Jim Wilson <jimw@sifive.com>
2605 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2606 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2607 Set arg_pointer_offset after using pretend_args_size.
2609 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2611 PR rtl-optimization/85431
2612 * dse.c (record_store): Ignore zero width stores.
2615 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2616 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2617 __builtin_stack_restore rather than after it.
2618 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2619 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2620 argument instead of virtual_dynamic_stack_rtx.
2622 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2624 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2626 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2627 Add note to error message to explain internal mapping of overloaded
2628 built-in function name to non-overloaded built-in function name.
2629 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2632 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2635 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2636 where the inputs overlap with the output.
2638 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2640 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2641 (=v, v) alternative and explicit "memory" attribute.
2642 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2643 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2645 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2646 "sselog1" type instead of "sselog".
2647 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2648 "sselog". Remove explicit "memory" attribute.
2649 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2650 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2652 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2653 "sselog1" type instead of "sselog". Remove explicit "memory"
2655 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2656 use "sselog1" type instead of "sselog". Remove explicit "memory"
2658 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2659 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2661 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2662 "sselog1" type instead of "sselog". Remove explicit "memory"
2664 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2665 use "sselog1" type instead of "sselog". Remove explicit "memory"
2669 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2672 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2675 2018-04-17 Martin Jambor <mjambor@suse.cz>
2678 * ipa-cp.c (create_specialized_node): Call
2679 expand_all_artificial_thunks if necessary.
2681 2018-04-17 Martin Liska <mliska@suse.cz>
2684 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2685 in message, remote space in between '_G' and '('.
2687 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2690 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2691 avx512f_vmcmp<mode>3<round_saeonly_name>,
2692 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2693 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2694 avx512f_rndscale<mode><round_saeonly_name>,
2695 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2696 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2697 Use %<iptr>2 instead of %2 for -masm=intel.
2698 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2699 avx512f_vcvttss2usi<round_saeonly_name>,
2700 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2702 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2703 avx512f_vcvttsd2usi<round_saeonly_name>,
2704 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2705 Use %q1 instead of %1 for -masm=intel.
2706 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2707 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2708 of %3 for -masm=intel.
2709 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2711 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2713 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2715 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2716 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2718 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2720 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2721 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2722 %g1 and one with %0 and %1.
2723 (avx512er_vmrcp28<mode><round_saeonly_name>,
2724 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2726 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2727 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2728 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2729 of %0 and %{%4%} for -masm=intel.
2730 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2731 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2732 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2733 order of %0 and %{%5%}%{z%} for -masm=intel.
2735 2018-04-17 Jan Hubicka <jh@suse.cz>
2738 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2740 2018-04-17 Martin Liska <mliska@suse.cz>
2743 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2744 for target_clone error message. Make default implementation
2745 clone to be a local declaration.
2746 (separate_attrs): Add new argument and check for an empty
2748 (expand_target_clones): Handle it.
2749 (ipa_target_clone): Make redirection just for target_clones
2752 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2753 Tom de Vries <tom@codesourcery.com>
2756 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2757 tiled basic blocks without omp continue statements.
2759 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2762 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2763 vec_extract expression as having side effects to make sure it gets
2766 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2769 * config/i386/i386.c (get_builtin_code_for_version): Check
2772 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2775 * gcc/config.gcc: Support "skylake".
2776 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2778 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2779 (processor_target_table): Add "skylake".
2780 (ix86_option_override_internal): Add "skylake".
2781 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2782 PROCESSOR_CANNONLAKE.
2783 (get_builtin_code_for_version): Fix priority for
2784 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2785 PROCESSOR_SKYLAKE-AVX512.
2786 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2787 (processor_type): Add PROCESSOR_SKYLAKE.
2789 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2790 Jason Merrill <jason@redhat.com>
2793 * convert.c (convert_to_integer_1): Use direct recursion for
2794 enumeral types and types with a precision less than the number
2795 of bits in their mode.
2797 2018-04-16 Julia Koval <julia.koval@intel.com>
2800 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2801 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2803 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2806 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2807 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2808 and -mno-direct-move.
2810 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2813 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2814 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2815 Ensure negative shifts result in {0}.
2817 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2819 PR rtl-optimization/79916
2820 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2821 regs (if any) to define how to gnerate SD moves when LRA is in
2824 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2826 PR rtl-optimization/85393
2827 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2828 * except.c (expand_dw2_landing_pad_for_region): Make static.
2829 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2830 a label and unconditional jump to old_bb, rather than
2831 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2834 PR rtl-optimization/85376
2835 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2836 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2837 instead of a specific value.
2839 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2840 Bin Cheng <bin.cheng@arm.com>
2842 PR tree-optimization/82965
2843 PR tree-optimization/83991
2844 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2845 by_profile_only parameter.
2846 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2847 information if the loop was predicted to iterate too many times.
2848 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2850 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2853 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2856 2018-04-13 Martin Liska <mliska@suse.cz>
2857 Jakub Jelinek <jakub@redhat.com>
2860 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2861 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2862 * builtins.c (expand_builtin_memory_copy_args): Use
2863 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2864 handle dest_addr == pc_rtx.
2866 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2869 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2870 asked to not generate direct moves.
2871 (fix_trunc<mode>si2_stfiwx): Similar.
2872 (fix_trunc<mode>si2_internal): Similar.
2874 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2877 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2878 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2879 lookup if dest in some wider mode is known to be const0_rtx and
2880 if so, record permanent equivalence for it to be ZERO_EXTEND of
2881 the narrower mode destination.
2883 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2885 * lto-streamer-out.c (output_function): Revert 259346.
2886 * omp-expand.c (expand_oacc_for): Likewise.
2888 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2890 PR rtl-optimization/85354
2891 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2892 * sel-sched.c (sel_global_init): ... here.
2894 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2897 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2898 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2899 mode for PE-COFF targets.
2900 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2901 (i386_pe_asm_lto_end): Likewise.
2902 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2903 (TARGET_ASM_LTO_END): Likewise.
2904 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2905 (i386_pe_asm_lto_start): New function.
2906 (i386_pe_asm_lto_end): Likewise.
2908 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2909 Richard Biener <rguenther@suse.de>
2912 * lto-streamer-out.c (output_function): Fix CFG loop state before
2914 * omp-expand.c (expand_oacc_for): Handle calls to internal
2915 functions like regular functions.
2917 2018-04-12 Richard Biener <rguenther@suse.de>
2920 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2921 for the early LTO debug to properly generate references to it
2922 during DIE emission. Do not re-use that for the skeleton for
2924 (dwarf2out_early_finish): Likewise.
2926 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2929 * config/i386/sse.md
2930 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2931 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2932 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2933 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2934 and output is a reg, avoid creating invalid lowpart subreg, but
2935 instead split into a 512-bit move. Don't split if not AVX512VL,
2936 input is xmm16+ reg and output is a mem.
2937 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2938 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2939 xmm16+ reg and output is a mem.
2941 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2943 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2944 also for flag_dwarf2_cfi_asm.
2946 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2948 PR rtl-optimization/85342
2949 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2950 a bool scalar var inside of the loop instead. Don't try to update
2951 recog_data.operand after failed apply_change_group.
2953 2018-04-12 Tom de Vries <tom@codesourcery.com>
2956 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2957 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2958 array with flexible array member as array without given dimension.
2959 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2960 argument for undefined param to true.
2962 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2965 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2966 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2967 from PowerPC section.
2968 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2969 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2970 help text that is too long.
2971 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2972 help text that is too long.
2973 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2974 help text that is too long.
2976 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2978 * config/alpha/alpha.md (stack_probe_internal): Rename
2979 from "probe_stack". Update all callers.
2981 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2983 PR rtl-optimization/84566
2984 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2985 sched_macro_fuse_insns.
2987 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2990 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2991 (compute_block_dependences): ... from here.
2993 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2995 PR tree-optimization/85331
2996 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2997 from int to HOST_WIDE_INT.
2999 2018-04-11 Martin Jambor <mjambor@suse.cz>
3002 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3003 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3004 not the same as the source val.
3005 (cgraph_edge_brings_value_p): New parameter.
3006 (gather_edges_for_value): Pass destination value to
3007 cgraph_edge_brings_value_p.
3008 (perhaps_add_new_callers): Likewise.
3009 (get_info_about_necessary_edges): Likewise and exclude values brought
3010 only by self-recursive edges.
3011 (create_specialized_node): Redirect only clones of self-calling edges.
3012 (+self_recursive_pass_through_p): New function.
3013 (find_more_scalar_values_for_callers_subset): Use it.
3014 (find_aggregate_values_for_callers_subset): Likewise.
3015 (known_aggs_to_agg_replacement_list): Removed.
3016 (decide_whether_version_node): Re-calculate known constants for all
3017 remaining context clones.
3019 2018-04-11 Richard Biener <rguenther@suse.de>
3022 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3023 from early DWARF output.
3024 (dwarf2out_early_finish): Output line info unconditionally into
3025 early DWARF and add reference to it.
3027 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3030 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3031 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3032 other than V2DFmode using iptr mode attribute.
3033 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3035 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3037 PR rtl-optimization/84659
3038 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3040 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3043 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3045 (output_loc_list): Pass address of a dummy size variable even in the
3046 locview handling loop.
3047 (index_location_lists): Add comment on why skip_loc_list_entry can't
3050 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3053 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3056 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3059 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3060 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3061 and -mstring-compare-inline-limit.
3063 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3066 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3067 for stack clash protection in a register whenever we need it to be in
3070 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3072 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3073 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3075 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3078 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3080 (mlong-double-): Ditto.
3081 * config/rs6000/sysv4.opt (msdata=): Ditto.
3082 (mtls-size=): Ditto.
3084 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3086 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3087 erroneous entries for
3088 "vector int vec_ldl (int, long int *)", and
3089 "vector unsigned int vec_ldl (int, unsigned long int *)".
3090 Add comments and entries for
3091 "vector bool char vec_ldl (int, bool char *)",
3092 "vector bool short vec_ldl (int, bool short *)",
3093 "vector bool int vec_ldl (int, bool int *)",
3094 "vector bool long long vec_ldl (int, bool long long *)",
3095 "vector pixel vec_ldl (int, pixel *)",
3096 "vector long long vec_ldl (int, long long *)",
3097 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3098 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3099 type tree bool_long_long_type_node and correct definition of
3100 bool_V2DI_type_node to make reference to this new type tree.
3101 (rs6000_mangle_type): Replace erroneous reference to
3102 bool_long_type_node with bool_long_long_type_node.
3103 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3104 comments to emphasize sign distinctions for char and int types and
3105 replace RS6000_BTI_bool_long constant with
3106 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3107 use of RS6000_BTI_pixel.
3108 (bool_long_type_node): Remove this macro definition.
3109 (bool_long_long_type_node): New macro definition
3111 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3113 PR rtl-optimization/85300
3114 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3115 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3116 simplify_unary_operation fails.
3118 2018-04-10 Martin Liska <mliska@suse.cz>
3120 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3121 cgraph_edge and ipa_ref.
3123 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3127 * config/i386/sse.md
3128 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3129 computation of the VEC_MERGE selector from mask.
3130 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3131 Fix decoding of the VEC_MERGE selector into mask.
3133 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3135 PR tree-optimization/85286
3136 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3138 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3140 * final.c (final_1): Set insn_last_address as well as
3141 insn_current_address.
3143 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3146 * explow.c (emit_stack_probe): Call validize_mem on memory location
3147 before passing it to gen_probe_stack. Create address operand and
3148 legitimize it for the probe_stack_address case.
3150 2018-04-09 Jan Hubicka <jh@suse.cz>
3153 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3154 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3155 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3157 (free_lang_data): Rebuild type inheritance graph.
3159 2018-04-09 Martin Sebor <msebor@redhat.com>
3161 * invoke.texi (-finline-small-functions): Mention other optimization
3163 (-findirect-inlining, -fpartial-inlining): Same.
3164 (-finline-functions-called-once): Same.
3165 (-freorder-blocks-and-partition): Same.
3167 2018-04-09 Jan Hubicka <jh@suse.cz>
3170 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3171 jumps; choose last target that matches the criteria (i.e.
3172 no partition changes for non-crossing jumps).
3173 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3174 support for redirecting crossing jumps to non-crossing.
3176 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3178 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3179 also for naked functions.
3181 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3183 * config/arc/arc.md (add_shift): New pattern.
3184 (add_shift2): Likewise.
3185 (sub_shift): Likewise.
3186 (sub_shift_cmp0_noout): Likewise.
3187 (compare_si_ashiftsi): Likewise.
3188 (xbfu_cmp0_noout): New combine pattern.
3189 (xbfu_cmp0"): Likewise.
3190 (movsi_set_cc_insn): Place the predicable variant first.
3191 (commutative_binary_cmp0_noout): Remove clobber.
3192 (commutative_binary_cmp0): New pattern.
3193 (noncommutative_binary_cmp0): Likewise.
3194 (noncommutative_binary_cmp0_noout): Likewise.
3195 (noncommutative_binary_comparison_result_used): Removed.
3196 (rsub_cmp0): New pattern.
3197 (rsub_cmp0_noout): Likewise.
3198 (extzvsi): Changed, keep only meaningful variants.
3199 (SQH, SEZ): New iterators.
3200 (SQH_postfix): New mode attribute.
3201 (SEZ_prefix): New code attribute.
3202 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3203 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3204 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3206 (noncommutative_operator): Check the availability of barrel
3209 2018-04-09 Richard Biener <rguenther@suse.de>
3211 PR tree-optimization/85284
3212 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3213 Only use the niter constraining form of simple_iv when the exit
3216 2018-04-09 Tom de Vries <tom@codesourcery.com>
3219 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3220 (define_expand "*memory_barrier"): New define_expand.
3221 (define_insn "memory_barrier"): New insn.
3223 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3225 PR rtl-optimization/80463
3226 PR rtl-optimization/83972
3227 PR rtl-optimization/83480
3229 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3230 correct producer for the insn.
3231 (tidy_control_flow): Fixup seqnos in case of debug insns.
3233 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3235 PR rtl-optimization/83913
3237 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3238 different sched-times when merging exprs.
3240 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3242 PR rtl-optimization/83962
3244 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3245 tidy_fallthru_edge and tidy_control_flow.
3247 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3249 PR rtl-optimization/83530
3251 * sel-sched.c (force_next_insn): New global variable.
3252 (remove_insn_for_debug): When force_next_insn is true, also leave only
3253 next insn in the ready list.
3254 (sel_sched_region): When the region wasn't scheduled, make another pass
3255 over it with force_next_insn set to 1.
3257 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3259 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3261 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3262 for interrupt control.
3263 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3264 functions for interrupt control.
3265 * config/nds32/nds32-intrinsic.md: Likewise.
3266 * config/nds32/nds32_intrinsic.h: Likewise.
3267 * config/nds32/nds32.h (nds32_builtins): Likewise.
3269 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3271 * config/nds32/nds32.c (nds32_init_machine_status,
3272 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3273 strict_aligned_p field.
3274 (nds32_expand_to_rtl_hook): New function.
3275 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3276 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3278 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3279 Chung-Ju Wu <jasonwucj@gmail.com>
3281 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3282 * config/nds32/nds32-n7.md: New file.
3283 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3284 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3286 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3287 * config/nds32/nds32.md (pipeline_model): Add n7.
3288 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3289 * config/nds32/pipelines.md: Include n7 settings.
3291 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3292 Chung-Ju Wu <jasonwucj@gmail.com>
3294 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3295 * config/nds32/nds32-e8.md: New file.
3296 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3297 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3299 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3300 * config/nds32/nds32.md (pipeline_model): Add e8.
3301 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3302 * config/nds32/pipelines.md: Include e8 settings.
3304 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3305 Chung-Ju Wu <jasonwucj@gmail.com>
3307 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3308 * config/nds32/nds32-n8.md: New file.
3309 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3310 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3312 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3313 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3314 * config/nds32/nds32.md (pipeline_model): Add n8.
3315 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3316 * config/nds32/pipelines.md: Include n8 settings.
3318 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3319 Chung-Ju Wu <jasonwucj@gmail.com>
3321 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3322 * config/nds32/nds32-n9-2r1w.md: New file.
3323 * config/nds32/nds32-n9-3r2w.md: New file.
3324 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3325 nds32_register_ports): New or modify for cpu n9.
3326 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3328 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3329 * config/nds32/nds32-utils.c: New file.
3330 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3331 TARGET_MUL_SLOW): Define.
3332 * config/nds32/nds32.md (pipeline_model): New attribute.
3333 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3334 New options that support cpu n9.
3335 * config/nds32/pipelines.md: Include n9 settings.
3336 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3338 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3340 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3341 information if necessary.
3342 (output_cond_branch_compare_zero): Likewise.
3343 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3344 (nds32_target_alignment): Refine for alignment.
3345 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3346 (FUNCTION_BOUNDARY): Modify.
3347 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3349 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3351 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3353 * config/nds32/constants.md (unspec_volatile_element): Add values for
3354 TLB operation and data prefetch.
3355 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3356 functions for TLB operation and data prefetch.
3357 * config/nds32/nds32-intrinsic.md: Likewise.
3358 * config/nds32/nds32_intrinsic.h: Likewise.
3359 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3360 (nds32_print_operand): Likewise.
3361 * config/nds32/nds32.h (nds32_builtins): Likewise.
3363 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3364 Andrew Pinski <pinsika@gcc.gnu.org>
3367 * match.pd: Use constant_boolean_node of correct type instead of
3368 boolean_true_node or boolean_false_node for simplifying
3369 pointer comparisons to zero.
3371 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3373 PR tree-optimization/80021
3374 * tree.c (verify_type_variant): Make error call in verify_variant_match
3375 translatable and remove final full stop.
3377 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3379 * config/nds32/constants.md (unspec_volatile_element): Add
3380 UNSPEC_VOLATILE_EH_RETURN.
3381 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3382 nds32_output_stack_pop): Support dwarf exception handling process.
3383 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3384 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3385 exception handling process.
3386 (nds32_compute_stack_frame): Likewise.
3387 (nds32_return_addr_rtx): Likewise.
3388 (nds32_initial_elimination_offset): Likewise.
3389 (nds32_expand_prologue): Likewise.
3390 (nds32_expand_epilogue): Likewise.
3391 (nds32_dynamic_chain_address): New function.
3392 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3394 (DYNAMIC_CHAIN_ADDRESS): Define.
3395 (EH_RETURN_DATA_REGNO): Define.
3396 (EH_RETURN_STACKADJ_RTX): Define.
3397 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3398 patterns for dwarf exception handling.
3400 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3402 * config/nds32/nds32.h: Clean up obsolete macros.
3404 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3406 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3407 Add enum values for particular instructions.
3408 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3409 particular intrinsic functions.
3410 * config/nds32/nds32-intrinsic.md: Likewise.
3411 * config/nds32/nds32_intrinsic.h: Likewise.
3412 * config/nds32/nds32.h (nds32_builtins): Likewise.
3413 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3414 (btst, ave): New patterns for particular instructions.
3416 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3418 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3419 Add enum values for atomic load/store and memory sync.
3420 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3422 * config/nds32/nds32-intrinsic.md: Likewise.
3423 * config/nds32/nds32_intrinsic.h: Likewise.
3424 * config/nds32/nds32.h (nds32_builtins): Likewise.
3426 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3428 PR tree-optimization/85257
3429 * fold-const.c (native_encode_vector): If not all elts could fit
3430 and off is -1, return 0 rather than offset.
3431 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3432 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3433 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3434 adjust buffer in native_interpret_expr call.
3436 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3438 * config/nds32/constants.md (unspec_volatile_element): Add cache
3439 control enum values.
3440 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3441 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3442 * config/nds32/nds32.c (nds32_cctl_names): New.
3443 (nds32_print_operand): Handle cache control register names.
3444 * config/nds32/nds32.h (nds32_builtins): New enum values.
3445 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3447 * config/nds32/nds32.md (type): Add mmu.
3448 * config/nds32/pipelines.md (simple_insn): Add mmu.
3450 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3452 * config/nds32/nds32.md (type): Remove call.
3453 * config/nds32/pipelines.md (simple_insn): Likewise.
3455 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3457 * config/nds32/constants.md (unspec_volatile_element): Add
3458 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3459 UNSPEC_VOLATILE_FMFCFG.
3460 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3461 description for fmfcfg and fmfcsr.
3462 (bdesc_1arg): Add fmtcsr.
3463 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3464 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3465 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3466 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3467 unspec_fmfcfg): New patterns.
3468 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3469 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3470 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3471 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3472 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3473 __nds32__fmfcfg): Define.
3475 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3477 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3478 intrinsic register names.
3479 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3480 intrinsic register enum values and macros.
3482 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3484 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3485 for load/store addressing form.
3486 (nds32_print_operand_address): Likewise.
3488 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3491 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3492 based on LABEL_REF. Remove useless assertion.
3493 (pic_address_needs_scratch): Fix formatting.
3494 (sparc_legitimize_pic_address): Minor tweaks.
3495 (sparc_delegitimize_address): Adjust assertion accordingly.
3496 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3497 into symbolic_operand.
3498 (movsi_high_pic_label_ref): Likewise.
3499 (movsi_lo_sum_pic_label_ref): Likewise.
3500 (movdi_pic_label_ref): Likewise.
3501 (movdi_high_pic_label_ref): Likewise.
3502 (movdi_lo_sum_pic_label_ref): Likewise.
3504 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3506 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3507 custom LIB_SPEC setup.
3509 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3510 Kito Cheng <kito.cheng@gmail.com>
3512 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3513 * config/riscv/freebsd.h: New.
3515 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3517 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3518 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3521 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3522 Kito Cheng <kito.cheng@gmail.com>
3524 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3525 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3526 * config/nds32/nds32-protos.h (nds32_output_call,
3527 nds32_output_return): Declare.
3528 * config/nds32/nds32.md: Refine all the call and return patterns.
3530 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3533 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3534 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3536 PR rtl-optimization/84872
3537 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3538 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3541 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3543 * expr.c (copy_blkmode_to_reg): Revert 254862.
3544 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3546 2018-04-06 Richard Biener <rguenther@suse.de>
3549 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3550 after seeing a component reference with an adjacent field. Treat
3551 refs to arrays at struct end of external decls similar to
3552 refs to unconstrained commons.
3554 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3557 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3558 look through SAVE_EXPRs with non-side-effects argument. Adjust
3560 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3563 2018-04-06 Richard Biener <rguenther@suse.de>
3566 * alias.c (find_base_term): New wrapper around find_base_term
3567 unwinding CSELIB_VAL_PTR changes.
3568 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3571 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3573 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3575 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3576 constant definitions.
3577 ("nop"): lr 0,0 -> nopr r0
3578 ("nop_lr0", "nop_lr1"): New insn definitions.
3580 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3582 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3583 NDS32_V3PUSH_AVAILABLE_P macro.
3585 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3586 Chung-Ju Wu <jasonwucj@gmail.com>
3588 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3589 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3590 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3591 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3592 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3593 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3594 * config/nds32/constraints.md: New constraints and checking for hard
3595 float configuration.
3596 * config/nds32/iterators.md: New mode iterator and attribute for hard
3597 float configuration.
3598 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3600 * config/nds32/nds32-fpu.md: New file.
3601 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3602 deal with hard float code generation.
3603 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3605 (abi_type, float_reg_number): New enum type.
3606 * config/nds32/nds32-predicates.c: New predicates for hard float.
3607 * config/nds32/nds32-protos.h: Declare functions for hard float.
3608 * config/nds32/nds32.c: Implementation for hard float configuration.
3609 * config/nds32/nds32.h: Definitions for hard float configuration.
3610 * config/nds32/nds32.md: Include hard float machine description and
3611 modify patterns for hard float configuration.
3612 * config/nds32/nds32.opt: New options for hard float configuration.
3613 * config/nds32/predicates.md: New predicates for hard float
3616 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3618 * common/config/nds32/nds32-common.c
3619 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3621 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3624 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3625 CONSTRUCTOR_ELT (ctor, ...)->value.
3627 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3630 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3632 2018-04-05 Tom de Vries <tom@codesourcery.com>
3635 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3638 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3639 Kito Cheng <kito.cheng@gmail.com>
3641 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3642 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3643 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3644 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3646 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3647 Kito Cheng <kito.cheng@gmail.com>
3649 * config/nds32/constraints.md (Ufe): New memory constraint.
3650 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3651 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3652 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3654 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3655 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3657 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3659 * config/nds32/nds32.md: Use optimize_size in the condition for
3660 alu-shift instructions.
3662 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3664 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3666 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3668 * config/nds32/nds32.md (negsi2): Refine pattern.
3670 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3671 Chung-Ju Wu <jasonwucj@gmail.com>
3673 * config/nds32/iterators.md (shift_rotate): New code iterator.
3674 (shift): New code attribute.
3675 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3676 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3677 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3678 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3679 bit-wise operations.
3680 (andsi3, *andsi3): Ditto.
3681 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3682 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3683 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3684 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3685 nds32_ior_operand, nds32_xor_operand): New predicates.
3687 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3689 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3690 (addsi3, subsi3): ... this.
3692 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3694 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3696 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3698 * config/nds32/nds32.md: Adjust indention.
3700 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3702 * config/nds32/nds32.md (feature): New attribute.
3704 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3706 * config/nds32/nds32.md (subtype): New attribute.
3708 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3711 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3712 expansion to perform a bitwise AND of the argument followed by a
3713 boolean negation of the result.
3715 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3717 PR rtl-optimization/84878
3718 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3719 the basic block. Assert the use reference is not artificial and that
3720 it has an associated insn.
3722 2018-04-04 Michael Matz <matz@suse.de>
3724 * builtins.c (compute_objsize): Pass correct operand
3725 to array_at_struct_end_p.
3727 2018-04-04 Richard Biener <rguenther@suse.de>
3730 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3731 from contexts for DINFO_LEVEL_TERSE and below.
3733 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3735 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3736 resiter_operand condition.
3737 * config/nds32/nds32.md (*move<mode>): Ditto.
3739 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3740 Monk Chiang <sh.chiang04@gmail.com>
3742 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3744 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3746 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3748 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3749 Kito Cheng <kito.cheng@gmail.com>
3751 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3752 nds32_cond_code_str, output_cond_branch,
3753 output_cond_branch_compare_zero, nds32_expand_cbranch,
3754 nds32_expand_cstore, nds32_expand_movcc,
3755 nds32_output_cbranchsi4_equality_zero,
3756 nds32_output_cbranchsi4_equality_reg,
3757 nds32_output_cbranchsi4_equality_reg_or_const_int,
3758 nds32_output_cbranchsi4_greater_less_zero: New functions.
3759 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3760 nds32_expand_cstore, nds32_expand_movcc,
3761 nds32_output_cbranchsi4_equality_zero,
3762 nds32_output_cbranchsi4_equality_reg,
3763 nds32_output_cbranchsi4_equality_reg_or_const_int,
3764 nds32_output_cbranchsi4_greater_less_zero): Declare.
3765 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3766 nds32_rimm11s_operand): New predicates.
3767 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3768 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3771 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3773 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3774 * config/nds32/nds32.md: Ditto.
3775 * config/nds32/pipelines.md: Ditto.
3777 2018-04-04 Richard Biener <rguenther@suse.de>
3779 PR tree-optimization/85168
3780 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3781 propagating abnormals.
3783 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3785 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3787 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3788 Kito Cheng <kito.cheng@gmail.com>
3790 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3791 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3792 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3793 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3794 * config/nds32/nds32.md (sibcall_internal): New.
3795 (sibcall_register): Remove.
3796 (sibcall_immediate): Remove.
3797 (sibcall_value_internal): New.
3798 (sibcall_value_register): Remove.
3799 (sibcall_value_immediate): Remove.
3800 * config/nds32/predicates.md (nds32_general_register_operand): New.
3801 (nds32_call_address_operand): New.
3803 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3805 PR rtl-optimization/85167
3806 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3807 bb_defs if *split_p, instead preinitialize it to NULL.
3809 PR tree-optimization/85156
3810 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3811 evaluating the argument multiple times.
3813 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3815 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3817 (_mm_cvtpd_ps): Likewise.
3818 (_mm_cvttpd_epi32): Likewise.
3819 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3820 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3821 vector, pixel, and bool following altivec.h include.
3823 2018-04-03 Martin Sebor <msebor@redhat.com>
3825 * doc/extend.texi (Common Function Attributes): Clarify.
3826 (const attribute): Likewise.
3827 (pure attribute): Likewise.
3829 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3832 * config/i386/i386.c (ix86_expand_vector_set): Use
3833 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3835 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3837 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3838 instructions when changing rounding bits to preserve precision bits
3839 in the x87 control word.
3841 2018-04-03 Martin Liska <mliska@suse.cz>
3843 PR tree-optimization/82491
3844 * rtl.h (strip_offset_and_add): Replace += suboffset with
3845 poly_uint64 () + suboffset.
3847 2018-03-29 Martin Liska <mliska@suse.cz>
3848 Martin Jambor <mjambor@suse.cz>
3851 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3852 param_type is not an integral or pointer type.
3854 2018-04-03 Richard Biener <rguenther@suse.de>
3856 * sese.h (recompute_all_dominators): Remove.
3858 2018-04-02 Martin Sebor <msebor@redhat.com>
3860 * doc/invoke.texi (-Wrestrict): Fix typos.
3862 2018-04-02 Jim Wilson <jimw@sifive.com>
3864 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3865 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3866 (<optab>di3, <optab>si3_extend): Likewise.
3867 (<optab>si3_mask, <optab>si3_mask_1): New.
3868 (<optab>di3_mask, <optab>di3_mask_1): New.
3869 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3870 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3871 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3873 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3875 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3878 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3880 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3881 (nds32_canonicalize_comparison): New function.
3883 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3884 Kito Cheng <kito.cheng@gmail.com>
3885 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3887 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3888 * config/nds32/constants.md (unspec_volatile_element): Add
3889 UNSPEC_VOLATILE_RELAX_GROUP.
3890 * config/nds32/nds32-relax-opt.c: New file.
3891 * config/nds32/nds32-predicates.c
3892 (nds32_symbol_load_store_p): New function.
3893 * config/nds32/nds32-protos.h
3894 (nds32_symbol_load_store_p): Declare function.
3895 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3896 * config/nds32/nds32.c
3897 (nds32_register_pass): New function to register pass.
3898 (nds32_register_passes): New function to register passes.
3899 * config/nds32/nds32.md (relax_group): New pattern.
3900 * config/nds32/nds32.opt (mrelax-hint): New option.
3901 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3903 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3905 * config/nds32/t-nds32: Modify files dependency.
3907 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3909 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3910 (PROFILE_HOOK): Define its implementation.
3912 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3914 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3915 type and 32-bit size.
3917 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3920 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3921 (V_128_256): New mode iterator.
3922 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3923 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3924 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3926 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3927 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3929 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3932 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3933 NaN inputs correctly.
3935 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3938 * config/rs6000/vsx.md (??r): New mode attribute.
3939 (*vsx_mov<mode>_64bit): Use it.
3940 (*vsx_mov<mode>_32bit): Likewise.
3942 2018-03-30 Martin Sebor <msebor@redhat.com>
3944 PR tree-optimization/84818
3945 * builtins.c (check_access): Use warning_n.
3947 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3950 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3952 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3955 2018-03-30 Julia Koval <julia.koval@intel.com>
3958 * x86-tune.def (movx, partial_reg_dependency): Enable for
3961 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3964 * lra-constraints.c (process_alt_operands): Move setting
3965 this_alternative_matches below.
3967 2018-03-29 Martin Liska <mliska@suse.cz>
3970 * doc/invoke.texi: Document how LTO works with debug info.
3971 Describe auto-load support of binutils. Mention 'x86-64'
3972 as valid option value of -march option.
3974 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3976 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3979 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3980 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3981 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3984 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3987 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3988 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3989 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3990 for RS6000_BTM_POWERPC64.
3991 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3992 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3993 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3998 2018-03-28 Carl Love <cel@us.ibm.com>
4001 2017-09-27 Carl Love <cel@us.ibm.com>
4003 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4004 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4005 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4008 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4010 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4011 instead of __vector bool.
4012 (_mm_max_pu8): Likewise.
4013 (_mm_min_pi16): Likewise.
4015 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4018 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4019 (DIVWEUO): Likewise.
4021 (DIVDEUO): Likewise.
4022 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4023 DIVWEUO and DIVDEUO.
4024 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4025 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4026 (div_extend): Likewise.
4027 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4029 (__builtin_divweuo): Likewise.
4030 (__builtin_divdeo): Likewise.
4031 (__builtin_divdeuo): Likewise.
4033 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4036 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4037 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4039 PR tree-optimization/82004
4040 * gimple-match-head.c (optimize_pow_to_exp): New function.
4041 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4042 Don't fold to exp if optimize_pow_to_exp is false.
4044 2018-03-28 Martin Liska <mliska@suse.cz>
4047 * calls.c (initialize_argument_information): Fix trailing space.
4048 * common.opt: Fix typo and provide better explanation for
4049 -fsanitize-coverage option.
4050 * config/i386/i386.opt: Fix typo.
4052 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4053 Martin Liska <mliska@suse.cz>
4056 * gimplify.c (asan_poison_variable): Don't do the check for
4057 gimplify_omp_ctxp here.
4058 (gimplify_decl_expr): Do it here.
4059 (gimplify_target_expr): Likewise.
4061 2018-03-28 Martin Liska <mliska@suse.cz>
4064 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4065 chkp_type_bounds_count if MPX is not enabled.
4067 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4069 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4071 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4074 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4075 function to create the function decl for complex long double
4076 multiply and divide for -mabi=ieeelongdouble.
4077 (init_float128_ieee): Call it.
4079 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4082 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4083 -fcf-protection=branch -mibt.
4084 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4086 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4089 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4091 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4094 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4095 extern array declarations.
4097 2018-03-27 Richard Biener <rguenther@suse.de>
4100 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4101 explicit single_use checks.
4103 2018-03-27 Richard Biener <rguenther@suse.de>
4105 PR tree-optimization/85082
4106 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4109 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4111 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4112 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4113 Turn on fasynchronous-unwind-tables and funwind-tables.
4115 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4118 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4119 (*bmi_blsr_<mode>_ccz): Ditto.
4121 2018-03-26 Tom de Vries <tom@codesourcery.com>
4123 PR tree-optimization/85063
4124 * omp-general.c (offloading_function_p): New function. Factor out
4126 * omp-offload.c (pass_omp_target_link::gate): ... here.
4127 * omp-general.h (offloading_function_p): Declare.
4128 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4129 with attribute omp declare target for offloading functions.
4131 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4133 PR tree-optimization/84005
4134 * tree-data-ref.h (get_base_for_alignment): Declare.
4135 * tree-data-ref.c (get_base_for_alignment_1): New function.
4136 (get_base_for_alignment): Likewise.
4137 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4138 get_base_for_alignment to find a suitable base object, instead
4139 of always using drb->base_address.
4141 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4144 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4145 known size by default.
4147 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4150 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4151 and non BLKmode operands.
4153 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4156 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4157 Clean up attributes.
4159 2018-03-23 Richard Biener <rguenther@suse.de>
4162 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4163 we are going to emit early debug for LTO.
4165 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4168 * function.c (match_asm_constraints_1): Don't optimize if input
4169 doesn't satisfy general_operand predicate for output's mode.
4172 * alias.c (write_dependence_p): Don't require for x_canonicalized
4173 non-VOIDmode if x has VOIDmode.
4176 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4177 just don't try to optimize it rather than assert it never happens.
4179 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4181 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4182 macro expansions for definition of ST_INTERNAL_<mode> and
4183 LD_INTERNAL_<mode> builtins.
4184 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4186 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4188 (altivec_expand_st_builtin): Likewise.
4189 (altivec_expand_builtin): Remove calls to deleted functions.
4190 (rs6000_address_for_altivec): Delete this function.
4191 * config/rs6000/vector.md: Remove expands for
4192 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4194 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4197 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4198 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4199 re-computing once computed.
4200 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4201 (arm_init_machine_status): Initialize
4202 machine->static_chain_stack_bytes.
4204 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4207 * doc/extend.texi: Add four new prototypes for vec_ld.
4208 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4209 definitions for more logical presentation.
4210 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4211 entries for V1TI variants of __builtin_altivec_ld builtin.
4212 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4213 handling of V1TI variant of LVX icode pattern.
4214 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4215 (rs6000_gimple_fold_builtin): Likewise.
4216 (altivec_init_builtins): Add code to define
4217 __builtin_altivec_lvx_v1ti function.
4219 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4222 * function.c (match_asm_constraints_1): Don't do the optimization
4223 if input isn't a REG, SUBREG, MEM or constant.
4225 2018-03-22 Tom de Vries <tom@codesourcery.com>
4227 PR tree-optimization/84956
4228 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4229 bb_has_abnormal_pred.
4231 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4234 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4235 DECL_INITIAL (decl) to decl at the end.
4236 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4239 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4241 * doc/extend.texi (__builtin_tgmath): Document when complex
4242 integer types are treated as _Complex _Float64.
4244 2018-03-21 Tom de Vries <tom@codesourcery.com>
4246 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4248 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4250 PR tree-optimization/84960
4251 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4252 if it is ENTRY block, move them into single succ of ENTRY in that case.
4254 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4256 PR tree-optimization/84811
4257 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4258 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4259 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4260 (known_subrange_p): Update accordingly. Cast each value involved
4261 in the size comparison, rather than casting the result of the
4264 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4266 PR tree-optimization/84982
4267 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4268 by flipping the least significant bit rather than all bits from
4269 bitpos to bitpos + bitsize - 1.
4271 2018-03-21 Nathan Sidwell <nathan@acm.org>
4273 * doc/extend.texi (Deprecated Features): Remove mention of
4274 long-deleted deprecations.
4276 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4279 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4280 * configure: Regenerate.
4282 2018-03-21 Tom de Vries <tom@codesourcery.com>
4284 PR tree-optimization/83126
4285 * tree-parloops.c (num_phis): New function.
4286 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4288 2018-03-21 Nathan Sidwell <nathan@acm.org>
4290 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4291 mention anon-struct/union members and trailing attributes.
4293 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4295 PR tree-optimization/84969
4296 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4297 builtin memset partitions if they set different rhs values.
4299 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4301 PR rtl-optimization/84989
4302 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4303 VEC_DUPLICATE with scalar result mode.
4305 2018-03-21 Martin Liska <mliska@suse.cz>
4308 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4309 not intended return statement.
4311 2018-03-21 Martin Liska <mliska@suse.cz>
4314 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4315 (chkp_find_bound_slots_1): Limit number of iterations.
4317 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4320 * Minor grammar fixes for x86 options.
4322 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4325 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4326 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4328 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4331 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4332 (altivec_lvx_<mode>_1op): Likewise.
4333 (altivec_stvx_<mode>_2op): Likewise.
4334 (altivec_stvx_<mode>_1op): Likewise.
4335 (altivec_lvx_<VM2:mode>): New define_expand.
4336 (altivec_stvx_<VM2:mode>): Likewise.
4337 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4338 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4339 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4340 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4341 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4342 (rs6000_gen_lvx): Likewise.
4343 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4344 (altivec_expand_stv_builtin): Likewise.
4345 (altivec_expand_builtin): Likewise.
4346 * config/rs6000/vector.md: Likewise.
4348 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4351 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4354 2018-03-20 Richard Biener <rguenther@suse.de>
4357 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4358 sign-conversions as zero, fall back to standard scalar_stmt
4361 2018-03-20 Martin Liska <mliska@suse.cz>
4364 * predict.c (rebuild_frequencies): Handle case when we have
4365 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4367 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4370 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4371 flag_section_anchors.
4372 * varasm.c (use_blocks_for_decl_p): Remove hack for
4373 dw2_force_const_mem.
4376 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4378 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4379 be created, use lowpart_subreg of operands[0] rather than operands[0]
4381 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4382 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4383 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4384 and n constraint instead of aarch64_shift_imm_di and Usd.
4385 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4386 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4388 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4391 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4392 to favor GPR over NEON registers.
4393 (<shift>di3_neon): Likewise.
4395 2018-03-20 Tom de Vries <tom@codesourcery.com>
4398 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4399 (nvptx_process_pars): Emit bar.sync asap and alap.
4401 2018-03-20 Tom de Vries <tom@codesourcery.com>
4404 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4405 seen_label if seen_label is already set.
4407 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4410 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4411 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4412 Use 1U instead of 1. Formatting fixes.
4415 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4416 instead of TREE_TYPE (s1) for the return value.
4418 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4420 PR tree-optimization/84946
4421 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4422 bitsize + bitsize in poly_uint64 rather than poly_int64.
4425 * dwarf2asm.c: Include fold-const.c.
4426 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4427 of decl rather than decl itself.
4429 PR rtl-optimization/84643
4430 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4432 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4435 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4436 calling assemble_variable.
4438 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4441 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4442 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4444 2018-03-19 Jim Wilson <jimw@sifive.com>
4447 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4448 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4449 (riscv_first_stack_step): Likewise.
4450 (riscv_option_override): Use STACK_BOUNDARY instead of
4452 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4454 (BIGGEST_ALIGNMENT): Set to 128.
4455 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4456 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4459 2018-03-19 Richard Biener <rguenther@suse.de>
4461 PR tree-optimization/84933
4462 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4463 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4465 2018-03-19 Richard Biener <rguenther@suse.de>
4467 PR tree-optimization/84859
4468 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4469 (cond_if_else_store_replacement): Perform sinking operation on
4470 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4471 Generalize what a BB with a single eligible store is.
4473 2018-03-19 Richard Biener <rguenther@suse.de>
4475 PR tree-optimization/84929
4476 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4477 chrec_is_positive against non-chrec arg.
4479 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4482 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4484 2018-03-18 Martin Liska <mliska@suse.cz>
4486 PR rtl-optimization/84635
4487 * regrename.c (build_def_use): Use matches_mode only when
4490 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4492 PR tree-optimization/84913
4493 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4494 vectorize chains of COND_EXPRs.
4496 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4498 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4500 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4502 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4504 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4506 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4508 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4509 Kito Cheng <kito.cheng@gmail.com>
4511 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4512 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4513 (nds32_adjust_reg_alloc_order): New function.
4514 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4516 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4518 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4519 nds32_print_operand, nds32_print_operand_address): Use
4520 HOST_WIDE_INT_PRINT_DEC instead.
4522 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4524 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4526 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4529 * config/i386/i386.c (initial_ix86_tune_features,
4530 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4532 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4533 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4534 rather than 1u << ix86_tune. Formatting fix.
4535 (ix86_option_override_internal): Change ix86_arch_mask from
4536 unsigned int to unsigned HOST_WIDE_INT, initialize to
4537 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4538 (ix86_function_specific_restore): Likewise.
4540 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4543 * postreload.c (reload_combine_recognize_pattern): Perform
4544 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4545 truncate_int_for_mode the result for the destination's mode.
4548 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4549 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4551 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4554 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4557 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4558 regno_allocno_class_array and sorted_pseudos.
4559 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4560 insns where regno is used.
4562 2018-03-16 Martin Liska <mliska@suse.cz>
4565 * multiple_target.c (create_dispatcher_calls): Redirect
4566 reference in the symbol table.
4568 2018-03-16 Martin Liska <mliska@suse.cz>
4571 * multiple_target.c (create_dispatcher_calls): Redirect also
4574 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4578 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4579 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4580 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4582 2018-03-16 Julia Koval <julia.koval@intel.com>
4584 * doc/invoke.texi (Skylake Server): Add CLWB.
4585 Cannonlake): Remove CLWB.
4587 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4589 PR tree-optimization/84841
4590 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4592 (FLOAT_ONE_CONST_TYPE): Define.
4593 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4594 (sort_by_operand_rank): Put entries with higher constant_type last
4595 rather than first to match comments.
4597 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4599 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4602 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4605 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4606 check and use address of target always.
4608 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4611 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4612 (indirect_thunk_bnd_needed): Likewise.
4613 (indirect_thunks_used): Likewise.
4614 (indirect_thunks_bnd_used): Likewise.
4615 (indirect_return_needed): New.
4616 (indirect_return_bnd_needed): Likewise.
4617 (output_indirect_thunk_function): Add a bool argument for
4619 (output_indirect_thunk_function): Don't generate alias for
4620 function return thunk.
4621 (ix86_code_end): Call output_indirect_thunk_function to generate
4622 function return thunks.
4623 (ix86_output_function_return): Set indirect_return_bnd_needed
4624 and indirect_return_needed instead of indirect_thunk_bnd_needed
4625 and indirect_thunk_needed.
4627 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4629 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4630 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4631 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4633 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4634 Paul Hua <paul.hua.gm@gmail.com>
4637 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4639 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4641 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4642 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4645 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4648 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4649 instead of GET_MODE_SIZE when comparing Units.
4651 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4654 * varasm.c (hash_section): Return an unchangeble hash value
4655 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4656 Return !aarch64_can_use_per_function_literal_pools_p ().
4658 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4661 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4662 rather than address of cmode as last argument to prepare_cmp_insn.
4664 2018-03-15 Julia Koval <julia.koval@intel.com>
4666 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4667 F_AVX512VNNI, F_AVX512BITALG): New.
4669 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4672 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4673 insn for floating-point loads and stores.
4675 2018-03-14 Carl Love <cel@us.ibm.com>
4677 * config/rs6000/rs6000-c.c: Add macro definitions for
4678 ALTIVEC_BUILTIN_VEC_PERMXOR.
4679 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4680 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4681 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4682 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4684 * config/doc/extend.texi: Add prototypes for vec_permxor.
4686 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4689 * diagnostic-show-locus.c (class layout_point): Convert m_line
4690 from int to linenum_type.
4691 (line_span::comparator): Use linenum "compare" function when
4692 comparing line numbers.
4693 (test_line_span): New function.
4694 (layout_range::contains_point): Convert param "row" from int to
4696 (layout_range::intersects_line_p): Likewise.
4697 (layout::will_show_line_p): Likewise.
4698 (layout::print_source_line): Likewise.
4699 (layout::should_print_annotation_line_p): Likewise.
4700 (layout::print_annotation_line): Likewise.
4701 (layout::print_leading_fixits): Likewise.
4702 (layout::annotation_line_showed_range_p): Likewise.
4703 (struct line_corrections): Likewise for field m_row.
4704 (line_corrections::line_corrections): Likewise for param "row".
4705 (layout::print_trailing_fixits): Likewise.
4706 (layout::get_state_at_point): Likewise.
4707 (layout::get_x_bound_for_row): Likewise.
4708 (layout::print_line): Likewise.
4709 (diagnostic_show_locus): Likewise for locals "last_line" and
4711 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4712 * input.c (selftest::test_linenum_comparisons): New function.
4713 (selftest::input_c_tests): Call it.
4714 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4715 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4716 * selftest.h (ASSERT_GT): New macro.
4717 (ASSERT_GT_AT): New macro.
4718 (ASSERT_LT): New macro.
4719 (ASSERT_LT_AT): New macro.
4721 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4723 PR rtl-optimization/84780
4724 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4726 2018-03-14 Martin Liska <mliska@suse.cz>
4728 * tree.c (record_node_allocation_statistics): Use
4729 get_stats_node_kind.
4730 (get_stats_node_kind): New function extracted from
4731 record_node_allocation_statistics.
4732 (free_node): Use get_stats_node_kind.
4734 2018-03-14 Richard Biener <rguenther@suse.de>
4736 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4737 that the value-set of ANTIC_IN doesn't grow.
4740 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4742 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4743 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4745 2018-03-14 Julia Koval <julia.koval@intel.com>
4747 * config.gcc (icelake-client, icelake-server): New.
4749 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4750 (initial_ix86_arch_features): Ditto.
4751 (PTA_SKYLAKE): Add SGX.
4752 (PTA_ICELAKE): Remove.
4753 (PTA_ICELAKE_CLIENT): New.
4754 (PTA_ICELAKE_SERVER): New.
4755 (ix86_option_override_internal): Split up icelake on icelake client and
4757 (get_builtin_code_for_version): Ditto.
4758 (fold_builtin_cpu): Ditto.
4759 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4760 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4761 * config/i386/i386.h (processor_type): Ditto.
4762 * doc/invoke.texi: Ditto.
4764 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4767 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4768 INTEGER_CST offset, add it together with bitpos / 8 and
4769 sign extend based on POINTER_SIZE.
4773 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4776 * config/i386/constraints.md (Yc): New register constraint.
4777 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4778 Use Yc constraint for alternative 2 of operand 0. Remove
4779 preferred_for_speed attribute.
4781 2018-03-14 Richard Biener <rguenther@suse.de>
4783 PR tree-optimization/84830
4784 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4785 with the old one to avoid oscillations.
4787 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4790 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4792 (assign_by_spills): Return a flag of reload assignment failure.
4793 Do not process the reload assignment failures. Do not spill other
4794 reload pseudos if they has the same reg class. Update n if
4796 (lra_assign): Add a return arg. Set up from the result of
4797 assign_by_spills call.
4798 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4799 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4800 usage_insns if it is not NULL.
4801 (spill_hard_reg_in_range): New function.
4802 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4803 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4804 function prototypes.
4805 (lra_assign): Change prototype.
4806 * lra.c (lra): Add code to deal with fails by splitting hard reg
4809 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4811 * config/riscv/riscv.opt (mrelax): New option.
4812 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4813 "norelax" when riscv_mrelax is disabled.
4814 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4816 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4819 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4820 reassociation for int modes.
4822 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4824 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4825 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4827 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4828 * config/aarch64/aarch64-sve.md
4829 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4830 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4831 (*extend<mode><Vwide>2): Rename to...
4832 (aarch64_sve_extend<mode><Vwide>2): ...this.
4833 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4834 renaming the old pattern to...
4835 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4837 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4838 define_expand, renaming the old pattern to...
4839 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4840 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4841 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4842 account when deciding which SVE instruction the optab should use.
4843 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4845 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4847 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4848 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4849 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4850 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4851 (tlsdesc_small_<mode>): Turn a define_expand and use
4852 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4853 (tlsdesc_small_advsimd_<mode>): ...this.
4854 (tlsdesc_small_sve_<mode>): New pattern.
4856 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4858 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4859 (UNSPEC_UMUL_HIGHPART): New constants.
4860 (MUL_HIGHPART): New int iteraor.
4861 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4862 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4864 (*<su>mul<mode>3_highpart): New define_insn.
4866 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4869 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4872 2018-03-13 Martin Liska <mliska@suse.cz>
4875 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4877 (sem_item_optimizer::~sem_item_optimizer): Release it.
4878 (sem_item_optimizer::merge_classes): Register variable aliases.
4879 (sem_item_optimizer::fixup_pt_set): New function.
4880 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4881 * ipa-icf.h: Declare new variables and functions.
4883 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4886 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4887 integer_pow2p@2 and test integer_pow2p in condition.
4888 (A < 0 ? C : 0): Similarly for @1.
4891 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4892 characters starting at p contain '\0' character, don't look beyond
4896 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4897 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4900 * reg-stack.c (change_stack): Change update_end var from int to
4901 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4902 also call set_block_for_insn on the newly added insns and rescan.
4905 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4906 on the last operand.
4909 * tree.c (stabilize_reference_1): Return save_expr (e) for
4910 STATEMENT_LIST even if it doesn't have side-effects.
4912 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4914 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4916 2018-03-12 Renlin Li <renlin.li@arm.com>
4918 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4919 aarch64_output_scalar_simd_mov_immediate.
4921 2018-03-12 Martin Sebor <msebor@redhat.com>
4923 PR tree-optimization/83456
4924 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4925 for perfectly overlapping calls to memcpy.
4926 (gimple_fold_builtin_memory_chk): Same.
4927 (gimple_fold_builtin_strcpy): Handle no-warning.
4928 (gimple_fold_builtin_stxcpy_chk): Same.
4929 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4931 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4933 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4934 parameter. Use it for SFmode.
4935 (rs6000_function_arg_advance_1): Adjust.
4936 (rs6000_function_arg): Adjust.
4937 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4939 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4941 PR rtl-optimization/84169
4942 PR rtl-optimization/84780
4943 * combine.c (can_combine_p): Check for a 2-insn combination whether
4944 the destination register is used between the two insns, too.
4946 2018-03-12 Richard Biener <rguenther@suse.de>
4948 PR tree-optimization/84803
4949 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4950 for refs DR analysis didn't process.
4952 2018-03-12 Richard Biener <rguenther@suse.de>
4954 PR tree-optimization/84777
4955 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4956 force-vectorize loops ignore whether we are optimizing for size.
4958 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4960 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4961 (TARGET_MD_ASM_ADJUST): Define.
4963 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4964 Kito Cheng <kito.cheng@gmail.com>
4965 Chung-Ju Wu <jasonwucj@gmail.com>
4967 * config/nds32/nds32.c (nds32_compute_stack_frame,
4968 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4969 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4970 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4971 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4972 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4973 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4974 * config/nds32/nds32.md (prologue, epilogue): Use macro
4975 NDS32_V3PUSH_AVAILABLE_P to do checking.
4977 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4980 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4981 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4982 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4983 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4984 addition of most attributes on !orig_type_die or the attribute not
4985 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4987 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4988 Chung-Ju Wu <jasonwucj@gmail.com>
4990 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4992 * config/nds32/nds32.opt (mvh): New option.
4994 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4995 Chung-Ju Wu <jasonwucj@gmail.com>
4997 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4999 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5000 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5003 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5004 Chung-Ju Wu <jasonwucj@gmail.com>
5006 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5008 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5009 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5011 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5012 Kito Cheng <kito.cheng@gmail.com>
5013 Chung-Ju Wu <jasonwucj@gmail.com>
5015 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5016 UNSPEC_FFMISM and UNSPEC_FLMISM.
5017 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5018 for ffb, ffmism and flmism.
5019 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5020 (unspec_ffmism): Ditto.
5021 (unspec_flmism): Ditto.
5022 (nds32_expand_builtin_impl): Check if string extension is available.
5023 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5024 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5026 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5029 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5032 * lra-assigns.c (assign_by_spills): Return a flag of reload
5033 assignment failure. Do not process the reload assignment
5034 failures. Do not spill other reload pseudos if they has the same
5036 (lra_assign): Add a return arg. Set up from the result of
5037 assign_by_spills call.
5038 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5039 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5040 usage_insns if it is not NULL.
5041 (spill_hard_reg_in_range): New function.
5042 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5043 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5044 function prototypes.
5045 (lra_assign): Change prototype.
5046 * lra.c (lra): Add code to deal with fails by splitting hard reg
5049 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5052 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5054 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5057 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5058 (dw_val_node): Add val_symbolic_view.
5059 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5060 (symview_upper_bound): New.
5061 (new_line_info_table): Initialize symviews_since_reset.
5062 (dwarf2out_source_line): Count symviews_since_reset and set
5063 symview_upper_bound.
5064 (dw_val_equal_p): Handle symview.
5065 (add_AT_symview): New.
5066 (print_dw_val): Handle symview.
5067 (attr_checksum, attr_checksum_ordered): Likewise.
5068 (same_dw_val_p, size_of_die): Likewise.
5069 (value_format, output_die): Likewise.
5070 (add_high_low_attributes): Use add_AT_symview for entry_view.
5071 (dwarf2out_finish): Reset symview_upper_bound, clear
5074 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5077 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5078 Add strict argument and use it.
5079 (rs6000_split_multireg_move): Update for new strict argument.
5080 (mem_operand_gpr): Disallow all non-offsettable addresses.
5081 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5083 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5086 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5087 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5088 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5091 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5092 decl, use remap_type if we want to use the type.
5094 2018-03-09 Martin Sebor <msebor@redhat.com>
5096 PR tree-optimization/84526
5097 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5099 (builtin_access::generic_overlap): Be prepared to handle non-array
5102 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5104 PR rtl-optimization/84682
5105 * lra-constraints.c (process_address_1): Check is_address flag
5106 for address constraints.
5107 (process_alt_operands): Likewise.
5108 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5109 preprocess_constraints.
5110 * recog.h (preprocess_constraints): Add oploc parameter.
5112 * recog.c (preprocess_constraints): Test address_operand for
5113 CT_ADDRESS constraints.
5115 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5118 * lra-assigns.c (assign_by_spills): Return a flag of reload
5119 assignment failure. Do not process the reload assignment
5120 failures. Do not spill other reload pseudos if they has the same
5122 (lra_assign): Add a return arg. Set up from the result of
5123 assign_by_spills call.
5124 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5125 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5126 usage_insns if it is not NULL.
5127 (spill_hard_reg_in_range): New function.
5128 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5129 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5130 function prototypes.
5131 (lra_assign): Change prototype.
5132 * lra.c (lra): Add code to deal with fails by splitting hard reg
5135 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5138 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5139 Accept complain bool parameter. Only emit errors if it is true.
5140 (arm_parse_cpu_option_name): Likewise.
5141 (arm_target_thumb_only): Adjust callers of the above.
5142 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5143 prototype to take a default true bool parameter.
5144 (arm_parse_arch_option_name): Likewise.
5146 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5147 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5151 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5152 * configure: Regenerate.
5153 * configure.ac ("linker --version-script option"): New.
5154 ("linker soname option"): New.
5156 2018-03-09 Richard Biener <rguenther@suse.de>
5158 PR tree-optimization/84775
5159 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5160 immediate uses of predicate stmts and mark them modified.
5163 PR tree-optimization/84178
5164 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5166 (version_loop_for_if_conversion): Delay update_ssa call.
5167 (tree_if_conversion): Delay update_ssa until after predicate
5170 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5173 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5174 when the function accesses prior frames.
5176 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5179 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5180 gen_llsym, otherwise call maybe_gen_llsym.
5183 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5184 has ',' character inside of it.
5186 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5189 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5190 as clobbering CC_REGNUM.
5192 2018-03-08 Richard Biener <rguenther@suse.de>
5195 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5196 (follow_copies_to_constant): Do not follow SSA names registered
5199 2018-03-08 Richard Biener <rguenther@suse.de>
5201 PR tree-optimization/84178
5202 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5204 (version_loop_for_if_conversion): Delay update_ssa call.
5205 (tree_if_conversion): Delay update_ssa until after predicate
5208 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5210 PR tree-optimization/84178
5211 * tree-if-conv.c (release_bb_predicate): Remove the
5212 the assertion that the stmts have NULL use_ops.
5213 Discard the statements, asserting that they haven't
5214 yet been added to a BB.
5216 2018-03-08 Richard Biener <rguenther@suse.de>
5218 PR tree-optimization/84746
5219 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5220 (phi_translate): Pass in destination ANTIC_OUT set.
5221 (phi_translate_1): Likewise. For a simplified result lookup
5222 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5223 (phi_translate_set): Adjust.
5224 (do_pre_regular_insertion): Likewise.
5225 (do_pre_partial_partial_insertion): Likewise.
5227 2018-03-08 Martin Liska <mliska@suse.cz>
5229 PR gcov-profile/84735
5230 * doc/gcov.texi: Document usage of profile files.
5231 * gcov-io.h: Document changes in the format.
5233 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5237 * dwarf2out.c (struct dw_line_info_table): Update comments for
5239 (FORCE_RESET_NEXT_VIEW): New.
5240 (FORCE_RESETTING_VIEW_P): New.
5241 (RESETTING_VIEW_P): Check for -1 too.
5242 (ZERO_VIEW_P): Likewise.
5243 (new_line_info_table): Force-reset next view.
5244 (dwarf2out_begin_function): Likewise.
5245 (dwarf2out_source_line): Simplify zero_view_p initialization.
5246 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5247 view directly. Omit view when omitting .loc at line 0.
5249 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5251 PR tree-optimization/84740
5252 * tree-switch-conversion.c (process_switch): Call build_constructors
5253 only if info.phi_count is non-zero.
5255 PR tree-optimization/84739
5256 * tree-tailcall.c (find_tail_calls): Check call arguments against
5257 DECL_ARGUMENTS (current_function_decl) rather than
5258 DECL_ARGUMENTS (func) when checking for tail recursion.
5260 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5262 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5263 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5264 Volker Reichelt's entry and add entries for people that perform
5265 GCC fuzzy testing and report numerous bugs.
5267 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5270 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5271 readonly data in sdata, if that is disabled.
5272 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5273 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5274 -mreadonly-in-sdata option.
5276 2018-03-07 Martin Sebor <msebor@redhat.com>
5278 PR tree-optimization/84468
5279 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5280 basic block when looking for nul assignment.
5282 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5285 * except.h (output_function_exception_table): Adjust prototype.
5286 * except.c (output_function_exception_table): Remove FNNAME parameter
5287 and add SECTION parameter. Ouput one part of the table at a time.
5288 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5289 the first part of the exception table and emit unwind directives.
5290 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5291 (i386_pe_seh_cold_init): Likewise.
5292 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5293 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5294 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5295 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5296 * config/i386/winnt.c: Include except.h.
5297 (struct seh_frame_state): Add reg_offset, after_prologue and
5298 in_cold_section fields.
5299 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5300 (i386_pe_seh_cold_init): New function.
5301 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5302 to seh->in_cold_section.
5303 (seh_emit_push): Record the offset of the push.
5304 (seh_emit_save): Record the offet of the save.
5305 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5306 Test seh->after_prologue to disregard the epilogue.
5307 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5308 (i386_pe_end_cold_function): New function.
5310 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5313 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5314 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5317 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5319 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5320 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5323 * multiple_target.c: Include tree-inline.h and intl.h.
5324 (expand_target_clones): Diagnose and fail if node->definition and
5325 !tree_versionable_function_p (node->decl).
5327 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5329 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5331 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5332 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5333 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5335 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5338 * combine.c (try_combine): Use reg_or_subregno instead of handling
5339 just paradoxical SUBREGs and REGs.
5341 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5343 * config/arc/arc.c (arc_finalize_pic): Remove function.
5344 (arc_must_save_register): We use single base PIC register, remove
5345 checks to save/restore the PIC register.
5346 (arc_expand_prologue): Likewise.
5347 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5349 (arc_verify_short): Likewise.
5350 (arc_attr_type): Likewise.
5351 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5352 (walk_stores): Likewise.
5353 (arc_address_cost): Make it static.
5354 (arc_verify_short): Likewise.
5355 (branch_dest): Likewise.
5356 (arc_attr_type): Likewise.
5357 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5358 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5359 (arc_final_prescan_insn): Remove inserting the nops due to
5360 hardware hazards. It is done in reorg step.
5361 (insn_length_variant_t): Remove.
5362 (insn_length_parameters_t): Likewise.
5363 (arc_insn_length_parameters): Likewise.
5364 (arc_get_insn_variants): Likewise.
5365 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5367 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5370 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5373 PR tree-optimization/84687
5374 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5376 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5378 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5380 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5381 Rename to ppc_speculation_barrier.
5382 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5383 __builtin_ppc_speculation_barrier.
5385 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5388 * combine.c (combine_simplify_rtx): Don't try to simplify if
5389 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5392 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5394 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5395 to 32 bytes when compiling for POWER9.
5397 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5400 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5401 regparm >= 3 with no arg reg available also for calls with
5402 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5405 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5407 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5409 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5412 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5414 2018-03-05 Richard Biener <rguenther@suse.de>
5416 PR tree-optimization/84486
5417 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5418 When inserting a __builtin_assume_aligned call set the LHS
5419 SSA name alignment info accordingly.
5421 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5423 PR tree-optimization/84114
5424 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5425 Avoid reassociation of FLOAT_MODE addition.
5427 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5429 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5430 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5431 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5432 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5433 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5434 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5435 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5437 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5438 __builtin_ia32_wbinvd): New builtins.
5439 (SPECIAL_ARGS2): New.
5440 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5441 (SPECIAL_ARGS2): New.
5442 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5443 (ix86_valid_target_attribute_inner_p): Ditto.
5444 (ix86_init_mmx_sse_builtins): Add special_args2.
5445 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5446 TARGET_WBNOINVD_P): New.
5447 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5448 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5449 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5450 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5451 * config/i386/pconfigintrin.h: New file.
5452 * config/i386/wbnoinvdintrin.h: Ditto.
5453 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5454 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5456 2018-03-05 Richard Biener <rguenther@suse.de>
5458 PR tree-optimization/84670
5459 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5461 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5462 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5463 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5464 doesn't grow if all successors (recursively) were visited at least
5467 2018-03-05 Richard Biener <rguenther@suse.de>
5469 PR tree-optimization/84650
5470 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5471 if executed in the loop pipeline.
5473 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5475 * doc/configfiles.texi (Configuration Files): Move info about
5476 conditionalizing $target-protos.h to...
5477 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5478 differs from $target-protos.h.
5480 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5481 Chung-Ju Wu <jasonwucj@gmail.com>
5483 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5484 * config/nds32/nds32-multiple.md (setmemsi): Define.
5485 * config/nds32/nds32-memory-manipulation.c
5486 (nds32_gen_dup_4_byte_to_word_value): New.
5487 (emit_setmem_word_loop): New.
5488 (emit_setmem_byte_loop): New.
5489 (nds32_expand_setmem_loop): New.
5490 (nds32_expand_setmem_loop_v3m): New.
5491 (nds32_expand_setmem_unroll): New.
5492 (nds32_expand_setmem): New.
5494 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5495 Chung-Ju Wu <jasonwucj@gmail.com>
5497 * config/nds32/nds32-memory-manipulation.c
5498 (nds32_emit_load_store): New.
5499 (nds32_emit_post_inc_load_store): New.
5500 (nds32_emit_mem_move): New.
5501 (nds32_emit_mem_move_block): New.
5502 (nds32_expand_movmemsi_loop_unknown_size): New.
5503 (nds32_expand_movmemsi_loop_known_size): New.
5504 (nds32_expand_movmemsi_loop): New.
5505 (nds32_expand_movmemsi_unroll): New.
5506 (nds32_expand_movmemqi): Rename ...
5507 (nds32_expand_movmemsi): ... to this.
5508 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5509 (movmemsi): ... to this.
5510 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5511 (nds32_expand_movmemsi): ... to this.
5513 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5514 Monk Chiang <sh.chiang04@gmail.com>
5515 Chung-Ju Wu <jasonwucj@gmail.com>
5517 * config/nds32/nds32-protos.h
5518 (nds32_expand_load_multiple): New arguments.
5519 (nds32_expand_store_multiple): Ditto.
5520 (nds32_valid_multiple_load_store): Rename ...
5521 (nds32_valid_multiple_load_store_p): ... to this.
5522 * config/nds32/nds32-memory-manipulation.c
5523 (nds32_expand_load_multiple): Refine implementation.
5524 (nds32_expand_store_multiple): Ditto.
5525 * config/nds32/nds32-multiple.md
5526 (load_multiple): Update nds32_expand_load_multiple interface.
5527 (store_multiple): Update nds32_expand_store_multiple interface.
5528 * config/nds32/nds32-predicates.c
5529 (nds32_valid_multiple_load_store): Rename ...
5530 (nds32_valid_multiple_load_store_p): ... to this and refine
5532 * config/nds32/predicates.md
5533 (nds32_load_multiple_and_update_address_operation): New predicate.
5534 (nds32_store_multiple_and_update_address_operation): New predicate.
5536 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5537 Chung-Ju Wu <jasonwucj@gmail.com>
5539 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5540 (combo): New attribute.
5541 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5543 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5545 * config/nds32/nds32.opt: Change -mcmodel= default value.
5547 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5548 Monk Chiang <sh.chiang04@gmail.com>
5549 Chung-Ju Wu <jasonwucj@gmail.com>
5551 * config/nds32/constants.md (unspec_element): New enum.
5552 * config/nds32/constraints.md (Umw): New constraint.
5553 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5554 * config/nds32/nds32-intrinsic.md: Likewise.
5555 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5556 (nds32_valid_smw_lwm_base_p): New.
5557 (nds32_output_smw_single_word): New.
5558 (nds32_output_lmw_single_word): New.
5559 (nds32_expand_unaligned_load): New.
5560 (nds32_expand_unaligned_store): New.
5561 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5562 (nds32_output_smw_single_word): Declare.
5563 (nds32_output_lmw_single_word): Declare.
5564 (nds32_expand_unaligned_load): Declare.
5565 (nds32_expand_unaligned_store): Declare.
5566 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5567 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5568 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5569 NDS32_BUILTIN_UASTORE_DW.
5570 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5573 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5574 Kito Cheng <kito.cheng@gmail.com>
5575 Chung-Ju Wu <jasonwucj@gmail.com>
5577 * config/nds32/nds32-intrinsic.c
5578 (nds32_expand_builtin_null_ftype_reg): Delete.
5579 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5580 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5581 (nds32_read_argument): New.
5582 (nds32_legitimize_target): Ditto.
5583 (nds32_legitimize_argument): Ditto.
5584 (nds32_check_constant_argument): Ditto.
5585 (nds32_expand_unop_builtin): Ditto.
5586 (nds32_expand_unopimm_builtin): Ditto.
5587 (nds32_expand_binop_builtin): Ditto.
5588 (nds32_builtin_decl_impl): Ditto.
5589 (builtin_description): Ditto.
5590 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5591 (nds32_init_builtins_impl): Ditto.
5592 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5593 (nds32_builtin_decl): New.
5594 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5595 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5597 2018-03-02 Jeff Law <law@redhat.com>
5599 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5600 (redundant_insn, fill_simple_delay_slots): Likewise.
5601 (fill_slots_from_thread): Likewise.
5602 * resource.c (mark_referenced_resources): Likewise.
5603 (mark_set_resources, find_dead_or_set_registers): Likewise.
5605 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5607 * substring-locations.h (format_warning_va): Formatting fix for
5609 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5611 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5612 * substring-locations.c: Include intl.h.
5613 (format_warning_va): Turned into small wrapper around
5614 format_warning_n_va, renamed to ...
5615 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5616 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5618 (format_warning_at_substring_n): New function.
5619 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5620 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5621 format_warning_at_substring with just a shorter name instead of
5622 const function pointer.
5623 (fmtwarn_n): New function.
5624 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5625 appropriate, get rid of all the fmtstr temporaries, move conditionals
5626 with G_() wrapped string literals directly into fmtwarn arguments,
5627 cast dir.len to (int), formatting fixes.
5629 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5631 * doc/invoke.texi: Remove "Cilk Plus" references.
5633 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5634 Richard Biener <rguenther@suse.de>
5637 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5638 for error or warning attributes if CALL_FROM_THUNK_P is set.
5641 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5644 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5645 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5648 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5649 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5651 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5652 instead of -1U in last predictors element's probability member.
5654 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5657 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5658 arguments if they are comparable.
5660 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5662 PR tree-optimization/84634
5663 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5664 masks and masked_loop_p with a single loop_masks, making sure it's
5665 null for bb vectorization.
5667 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5669 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5670 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5671 loop->force_vectorize to check whether there is no alias.
5673 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5676 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5678 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5680 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5681 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5682 instead of a loop around prev_real_insn.
5683 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5687 * config/i386/i386.c (ix86_print_operand): Use conditional
5688 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5691 2018-03-02 Richard Biener <rguenther@suse.de>
5693 PR tree-optimization/84427
5694 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5695 (bitmap_set_subtract_values): Rewrite to handle multiple
5698 (prune_clobbered_mems): Likewise.
5699 (phi_translate): Take edge instead of pred/phiblock.
5700 (phi_translate_1): Likewise.
5701 (phi_translate_set): Likewise. Insert all translated
5702 exprs for a value into the set, keeping possibly multiple
5703 expressions per value.
5704 (compute_antic_aux): Adjust for phi_translate changes.
5705 When intersecting union the expressions and prune those
5706 not in the final value set, keeping possibly multiple
5707 expressions per value. Do not use value-insertion
5708 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5709 all expressions. Add verification that the value-sets
5710 only shrink during iteration.
5711 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5712 (do_pre_regular_insertion): Likewise.
5713 (do_pre_partial_partial_insertion): Likewise.
5715 2018-03-02 Richard Biener <rguenther@suse.de>
5718 * config/darwin.c (saved_debug_info_level): New static global.
5719 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5720 (darwin_asm_lto_end): Restore debug info generation settings.
5722 2018-03-01 Martin Liska <mliska@suse.cz>
5725 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5728 2018-03-01 Richard Biener <rguenther@suse.de>
5731 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5732 type annotation with LTO when debug was disabled at compile-time.
5734 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5736 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5738 (mips_final_postscan_insn): Likewise.
5740 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5742 PR rtl-optimization/84528
5743 * alias.c (init_alias_target): Add commentary.
5744 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5745 a unique base value if the frame pointer is not eliminated
5746 to the stack pointer.
5748 2018-03-01 Tom de Vries <tom@codesourcery.com>
5750 PR rtl-optimization/83327
5751 * lra-int.h (hard_regs_spilled_into): Declare.
5752 * lra.c (hard_regs_spilled_into): Define.
5753 (init_reg_info): Init hard_regs_spilled_into.
5754 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5755 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5756 (process_bb_lives): Handle hard_regs_spilled_into.
5757 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5758 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5760 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5762 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5763 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5764 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5765 * config/rs6000/aix72.h: New file.
5767 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5769 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5770 instead of warning_at with conditional singular and plural messages
5774 * stor-layout.c (update_alignment_for_field): For
5775 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5776 && !DECL_PACKED (field), do the alignment update, just use
5777 only desired_align instead of MAX (type_align, desired_align)
5779 (place_field): Don't do known_align < desired_align handling
5780 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5781 is non-NULL, instead do it after rli->prev_field handling and
5782 only if not within a bitfield word. For DECL_PACKED (field)
5783 use type_align of BITS_PER_UNIT.
5785 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5787 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5788 superfluous parentheses and trailing spaces.
5790 2018-02-28 Richard Biener <rguenther@suse.de>
5792 PR tree-optimization/84584
5793 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5794 SCoPs with fake exit edge.
5796 2018-02-28 Martin Liska <mliska@suse.cz>
5799 * timevar.c (timer::print): Fix format to properly print 100%
5802 2018-02-28 Richard Biener <rguenther@suse.de>
5805 * genmatch.c (capture_info::walk_match): Do not mark
5806 captured expressions without operands as expr_p given
5807 they act more like predicates and should be subject to
5808 "lost tail" side-effect preserving.
5810 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5812 PR rtl-optimization/81611
5813 * auto-inc-dec.c (attempt_change): Move dead note from
5814 mem_insn if it's the next use of regno
5815 (find_address): Take address use of reg holding
5816 non-incremented value. Add parm to limit search to the named
5818 (merge_in_block): Attempt to use a mem insn that is the next
5819 use of the original regno.
5821 2018-02-27 Martin Sebor <msebor@redhat.com>
5824 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5825 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5827 2018-02-27 Martin Sebor <msebor@redhat.com>
5829 PR translation/84207
5830 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5831 n argument to unsigned HOST_WIDE_INT.
5832 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5833 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5834 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5835 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5837 2018-02-27 Richard Biener <rguenther@suse.de>
5839 PR tree-optimization/84512
5840 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5841 Do not use the estimate returned from record_stmt_cost for
5842 the scalar iteration cost but sum properly using add_stmt_cost.
5844 2018-02-27 Richard Biener <rguenther@suse.de>
5846 PR tree-optimization/84466
5847 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5848 Adjust last change to less strictly validate use operands.
5850 2018-02-27 Martin Liska <mliska@suse.cz>
5852 PR gcov-profile/84548
5853 * gcov.c (process_file): Allow partial overlap and consider it
5854 also as group functions.
5855 (output_lines): Properly calculate range of lines for a group.
5857 2018-02-27 Martin Liska <mliska@suse.cz>
5859 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5860 'ggc' suffixes. Change first column width.
5861 (timer::print): Fix formatting of the column.
5863 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5865 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5866 preserve inline entry blocks for the sake of debug inline
5867 entry point markers alone.
5868 (remove_unused_locals): Suggest in comments a better place to
5869 force the preservation of inline entry blocks that are
5870 otherwise unused, but do not preserve them.
5872 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5874 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5876 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5879 * config/i386/constraints.md (Bs): Replace
5880 ix86_indirect_branch_register with
5881 TARGET_INDIRECT_BRANCH_REGISTER.
5883 * config/i386/i386.md (indirect_jump): Likewise.
5884 (tablejump): Likewise.
5885 (*sibcall_memory): Likewise.
5886 (*sibcall_value_memory): Likewise.
5887 Peepholes of indirect call and jump via memory: Likewise.
5888 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5889 (*sibcall_value_GOT_32): Likewise.
5890 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5891 (GOT_memory_operand): Likewise.
5892 (call_insn_operand): Likewise.
5893 (sibcall_insn_operand): Likewise.
5894 (GOT32_symbol_operand): Likewise.
5895 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5897 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5899 PR rtl-optimization/83496
5900 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5901 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5902 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5903 redundant insn, if any.
5904 (relax_delay_slots): Likewise.
5905 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5907 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5909 PR tree-optimization/83965
5910 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5911 that grouped statements are part of a reduction chain. Return
5912 true if the statement is not marked as a reduction itself but
5914 (vect_recog_dot_prod_pattern): Don't check whether the statement
5915 is part of a group here.
5916 (vect_recog_sad_pattern): Likewise.
5917 (vect_recog_widen_sum_pattern): Likewise.
5919 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5922 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5924 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5927 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5929 (ix86_output_indirect_function_return): New prototype.
5930 (ix86_split_simple_return_pop_internal): Likewise.
5931 * config/i386/i386.c (indirect_return_via_cx): New.
5932 (indirect_return_via_cx_bnd): Likewise.
5933 (indirect_thunk_name): Handle return va CX_REG.
5934 (output_indirect_thunk_function): Create alias for
5935 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5936 (ix86_output_indirect_jmp): Remove the bool argument.
5937 (ix86_output_indirect_function_return): New function.
5938 (ix86_split_simple_return_pop_internal): Likewise.
5939 * config/i386/i386.md (*indirect_jump): Don't pass false
5940 to ix86_output_indirect_jmp.
5941 (*tablejump_1): Likewise.
5942 (simple_return_pop_internal): Change it to define_insn_and_split.
5943 Call ix86_split_simple_return_pop_internal to split it for
5945 (simple_return_indirect_internal): Call
5946 ix86_output_indirect_function_return instead of
5947 ix86_output_indirect_jmp.
5949 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5952 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5953 memset and value initialization afterwards.
5955 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5957 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5959 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5962 * common/config/aarch64/aarch64-common.c
5963 (aarch_option_optimization_table[]): Switch
5964 off fomit-frame-pointer
5966 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5967 Chung-Ju Wu <jasonwucj@gmail.com>
5969 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5971 (store_multiple): Ditto.
5973 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5975 * config.gcc: Add --with-cpu support for nds32 target.
5976 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5977 * config/nds32/nds32.opt: Add -mcpu= option.
5979 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5981 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5982 isel=yes): Warn for these deprecated options.
5984 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5986 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5987 ISA_2_5_MASKS_EMBEDDED.
5989 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5991 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5992 p->max as pointers rather than using iterative_hash_expr.
5994 2018-02-23 Carl Love <cel@us.ibm.com>
5996 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5997 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5999 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6000 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6001 P8V_BUILTIN_VEC_VUNSIGNED2.
6003 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6006 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6007 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6009 * lra-constraints.c (curr_insn_transform): Set up
6010 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6012 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6013 * lra-eliminations.c (spill_pseudos): Ditto.
6014 (process_insn_for_elimination): Ditto.
6015 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6016 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6017 LRA_NON_CLOBBERED_ALT.
6019 2018-02-22 Martin Sebor <msebor@redhat.com>
6021 PR tree-optimization/84480
6022 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6023 to maybe_diag_stxncpy_trunc. Call it.
6024 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6025 from gimple_fold_builtin_strcpy. Print inlining stack.
6026 (handle_builtin_stxncpy): Print inlining stack.
6027 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6029 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6032 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6033 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6034 and -fcheck-pointer-bounds are used together.
6035 (indirect_thunk_prefix): New enum.
6036 (indirect_thunk_need_prefix): New function.
6037 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6038 "_nt" instead of "_bnd" for NOTRACK prefix.
6039 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6040 (output_indirect_thunk_function): Likewise.
6042 (ix86_code_end): Update output_indirect_thunk_function calls.
6043 (ix86_output_indirect_branch_via_reg): Replace
6044 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6045 (ix86_output_indirect_branch_via_push): Likewise.
6046 (ix86_output_function_return): Likewise.
6047 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6048 incompatible with -fcf-protection=branch and
6049 -fcheck-pointer-bounds.
6051 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6054 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6055 Change gcc_assert call to output_operand_lossage.
6057 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6059 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6061 2018-02-22 DJ Delorie <dj@redhat.com>
6062 Sebastian Perta <sebastian.perta@renesas.com>
6063 Oleg Endo <olegendo@gcc.gnu.org>
6065 * config/rx/rx.c (rx_rtx_costs): New function.
6066 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6068 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6070 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6072 2018-02-22 Martin Liska <mliska@suse.cz>
6075 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6076 Add "native" as a possible value.
6078 2018-02-22 Martin Liska <mliska@suse.cz>
6081 * config/i386/i386.c (ix86_option_override_internal):
6082 Add "native" as a possible value for -march and -mtune.
6084 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6087 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6088 to all type variants.
6090 PR tree-optimization/84503
6091 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6092 width as info->bitpos + info->bitsize - start.
6093 (merged_store_group::merge_overlapping): Simplify width computation.
6094 (check_no_overlap): New function.
6095 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6096 start + width and last_order of the group, fail if check_no_overlap
6098 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6099 to group if check_no_overlap fails.
6101 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6103 * config/rs6000/altivec.md: Delete contraint arguments to
6104 define_expand, define_split, and define_peephole2, and in
6105 define_insn_and_split if always unused.
6106 * config/rs6000/darwin.md: Ditto.
6107 * config/rs6000/dfp.md: Ditto.
6108 * config/rs6000/rs6000.md: Ditto.
6109 * config/rs6000/sync.md: Ditto.
6110 * config/rs6000/vector.md: Ditto.
6111 * config/rs6000/vsx.md: Ditto.
6113 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6115 * config/rs6000/altivec.md: Write output control strings as braced
6116 blocks instead of double-quoted strings.
6117 * config/rs6000/darwin.md: Ditto.
6118 * config/rs6000/rs6000.md: Ditto.
6119 * config/rs6000/vector.md: Ditto.
6120 * config/rs6000/vsx.md: Ditto.
6122 2018-02-21 Jason Merrill <jason@redhat.com>
6124 PR c++/84314 - ICE with templates and fastcall attribute.
6125 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6127 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6129 * ipa-cp.c (determine_versionability): Fix comment typos.
6131 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6134 * ipa-cp.c (determine_versionability): Do not version functions caling
6137 2018-02-21 Martin Liska <mliska@suse.cz>
6140 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6141 Add "native" as a possible value.
6142 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6143 the macro when native cpu detection is available.
6145 2018-02-21 Martin Liska <mliska@suse.cz>
6148 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6149 Add "native" as a possible value.
6150 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6151 when native cpu detection is available.
6153 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6154 Martin Sebor <msebor@redhat.com>
6156 PR tree-optimization/84478
6157 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6159 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6160 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6161 support which is conservatively correct, for 2 only stay conservative
6162 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6163 argument to the 2 argument get_range_strlen, adjust 6 arg
6164 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6166 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6167 (gimple_fold_builtin_strlen): Pass true as last argument to
6170 2018-02-20 Martin Sebor <msebor@redhat.com>
6173 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6174 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6175 (builtin_memref::builtin_memref): Factor out parts into
6176 set_base_and_offset and call it.
6178 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6181 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6182 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6183 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6184 search at the associated MODE_INT.
6186 2018-02-20 Jeff Law <law@redhat.com>
6189 PR tree-optimization/81592
6191 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6192 for range data rather than using global data.
6193 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6194 range data rather than using global data.
6195 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6196 pass it to children as needed.
6197 (struct directive::fmtresult): Similarly.
6198 (struct directive::set_width): Similarly.
6199 (struct directive::set_precision): Similarly.
6200 (format_integer, format_directive, parse_directive): Similarly.
6201 (format_none): Accept unnamed vr_values parameter.
6202 (format_percent, format_floating, format_character): Similarly.
6203 (format_string, format_plain): Similarly.
6204 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6205 the EVRP range analyzer for range data rather than using global data.
6206 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6207 gimple-ssa-evrp-analyze.h
6208 (class sprintf_dom_walker): Add after_dom_children member function.
6209 Add evrp_range_analyzer member.
6210 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6211 range analyzer as needed.
6212 (sprintf_dom_walker::after_dom_children): New member function.
6213 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6215 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6216 (evrp_range_analyzer::pop_to_marker): Likewise.
6218 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6220 PR tree-optimization/84419
6221 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6222 with the required type if its current type is compatible but
6225 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6228 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6229 after vectorization.
6231 2018-02-20 Martin Liska <mliska@suse.cz>
6234 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6235 possible values if we don't have a hint.
6237 2018-02-20 Martin Liska <mliska@suse.cz>
6241 * final.c (shorten_branches): Build align_tab array with one
6243 * opts.c (finish_options): Add alignment option limit check.
6244 (MAX_CODE_ALIGN): Likewise.
6245 (MAX_CODE_ALIGN_VALUE): Likewise.
6246 * doc/invoke.texi: Document maximum allowed option value for
6247 all -falign-* options.
6249 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6252 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6253 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6254 * var-tracking.c (emit_note_insn_var_location): Remove all references
6255 to NOTE_INSN_CALL_ARG_LOCATION.
6256 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6257 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6258 Use copy_rtx_if_shared.
6259 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6260 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6261 (dwarf2out_var_location): Remove handling of
6262 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6264 * final.c (final_scan_insn): Remove all references to
6265 NOTE_INSN_CALL_ARG_LOCATION.
6266 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6267 before dumping final insns.
6268 * except.c (emit_note_eh_region_end): Remove all references to
6269 NOTE_INSN_CALL_ARG_LOCATION.
6270 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6271 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6272 * config/arc/arc.c (hwloop_optimize): Likewise.
6273 * config/arm/arm.c (create_fix_barrier): Likewise.
6274 * config/s390/s390.c (s390_chunkify_start): Likewise.
6275 * config/sh/sh.c (find_barrier): Likewise.
6276 * config/i386/i386.c (rest_of_insert_endbranch,
6277 ix86_seh_fixup_eh_fallthru): Likewise.
6278 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6279 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6280 * config/frv/frv.c (frv_function_prologue): Likewise.
6281 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6283 (note_outside_basic_block_p): Remove all references to
6284 NOTE_INSN_CALL_ARG_LOCATION.
6285 * gengtype.c (adjust_field_rtx_def): Likewise.
6286 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6288 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6289 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6292 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6295 PR tree-optimization/84452
6296 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6297 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6300 2018-02-19 Martin Liska <mliska@suse.cz>
6303 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6305 2018-02-19 Martin Liska <mliska@suse.cz>
6306 Richard Sandiford <richard.sandiford@linaro.org>
6308 PR tree-optimization/82491
6309 * gimple-fold.c (get_base_constructor): Make earlier bail out
6312 2018-02-19 Carl Love <cel@us.ibm.com>
6314 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6315 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6317 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6318 P8V_BUILTIN_VEC_NEG.
6320 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6322 * config/rl78/rl78.md (movdf): New define expand.
6324 2018-02-19 Martin Liska <mliska@suse.cz>
6327 * doc/invoke.texi: Fix typo.
6328 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6330 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6332 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6333 handle rs6000_single_float and rs6000_double_float specially for
6336 2018-02-16 Jeff Law <law@redhat.com>
6338 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6341 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6344 * ipa-inline.c (inline_small_functions): Fix a typo.
6346 2018-02-16 Nathan Sidwell <nathan@acm.org>
6348 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6350 2018-02-16 Carl Love <cel@us.ibm.com>
6352 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6353 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6354 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6355 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6356 expansion to P8V_BUILTIN_VEC_FLOAT2.
6358 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6360 PR rtl-optimization/70023
6361 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6362 src_regno into account.
6364 2018-02-16 Carl Love <cel@us.ibm.com>
6366 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6367 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6368 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6369 * config/rs6000/rs6000.c: Remove case statements for
6370 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6371 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6372 and P9V_BUILTIN_VEC_VINSERT4B.
6373 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6374 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6375 * config/rs6000/vsx.md:
6376 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6379 2018-02-16 Carl Love <cel@us.ibm.com>
6381 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6383 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6385 * config/rs6000/rs6000-c.c: Add the definitions for
6386 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6387 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6388 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6389 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6390 definition for insert4b and define insn *insert3b_internal.
6391 * doc/extend.texi: Add documentation for vec_extract4b.
6393 2018-02-16 Nathan Sidwell <nathan@acm.org>
6395 * doc/extend.texi (Backwards Compatibility): Mention friend
6396 injection. Note for-scope is deprecated.
6397 * doc/invoke.texi (-ffriend-injection): Deprecate.
6399 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6401 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6402 that moved to I2, also allow destinations that are a paradoxical
6403 subreg (instead of a normal reg).
6405 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6408 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6411 2018-02-16 Richard Biener <rguenther@suse.de>
6413 PR tree-optimization/84037
6414 PR tree-optimization/84016
6416 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6417 Adjust vec_construct for the fact we need additional higher latency
6418 128bit inserts for AVX256 and AVX512 vector builds.
6419 (ix86_add_stmt_cost): Scale vector construction cost for
6422 2018-02-16 Richard Biener <rguenther@suse.de>
6424 PR tree-optimization/84417
6425 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6426 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6427 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6429 2018-02-16 Martin Liska <mliska@suse.cz>
6432 * internal-fn.def (ASAN_CHECK): Set proper flags.
6433 (ASAN_MARK): Likewise.
6435 2018-02-16 Julia Koval <julia.koval@intel.com>
6437 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6438 from PTA_CANNONLAKE.
6440 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6443 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6444 Use ++iter rather than iter++ for std::list iterators.
6445 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6446 defer deleting them until all nodes in the forest are processed. Do
6447 free even leaf nodes. Change to_process into auto_vec.
6450 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6451 * vec.h (vec_default_construct): Use memset instead of placement new
6452 if BROKEN_VALUE_INITIALIZATION is defined.
6453 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6454 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6457 PR rtl-optimization/83723
6458 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6459 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6460 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6462 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6464 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6466 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6468 PR rtl-optimization/81443
6469 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6470 from inner REGs to paradoxical SUBREGs.
6472 2018-02-16 Richard Biener <rguenther@suse.de>
6474 PR tree-optimization/84399
6475 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6476 For operands we can analyze at their definition make sure we can
6477 analyze them at each use as well.
6479 2018-02-16 Richard Biener <rguenther@suse.de>
6481 PR tree-optimization/84190
6482 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6483 volatile accesses if the decl isn't volatile.
6485 2018-02-15 Jason Merrill <jason@redhat.com>
6487 PR c++/84314 - ICE with templates and fastcall attribute.
6488 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6489 TYPE_CANONICAL on an existing type.
6491 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6493 PR tree-optimization/84383
6494 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6495 dstoff nor call operand_equal_p if dstbase is NULL.
6497 PR tree-optimization/84334
6498 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6499 also a CONSTANT_CLASS_P, punt.
6501 2018-02-14 Jim Wilson <jimw@sifive.com>
6503 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6504 first SMALL_OPERAND check. New local min_second_step. Move assert
6505 to where locals are set. Add TARGET_RVC support.
6506 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6508 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6510 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6512 2018-02-14 Martin Sebor <msebor@redhat.com>
6514 PR tree-optimization/83698
6515 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6516 arrays constrain the offset range to their bounds.
6517 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6518 (builtin_access::overlap): Avoid setting the size of overlap if it's
6520 (maybe_diag_overlap): Also consider arrays when deciding what values
6521 of offsets to include in diagnostics.
6523 2018-02-14 Martin Sebor <msebor@redhat.com>
6526 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6527 that correspond to the kind of a declaration.
6529 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6532 * config/pa/pa.md: Load address of PIC label using the linkage table
6533 if the label is nonlocal.
6535 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6537 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6538 warning message if user requests -maltivec=be.
6539 * doc/invoke.texi: Document deprecation of -maltivec=be.
6541 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6544 * config/rs6000/rs6000-c.c: Update definitions for
6545 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6546 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6548 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6551 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6552 add _get_ssp intrinsics. Remove argument from
6553 __builtin_ia32_rdssp[d|q].
6554 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6555 * config/i386/i386-builtin.def: Remove argument from
6556 __builtin_ia32_rdssp[d|q].
6557 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6558 ix86_expand_special_args_builtin for _rdssp[d|q].
6559 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6560 Clear register before usage.
6561 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6562 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6564 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6566 PR tree-optimization/84357
6567 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6568 operand 1 of an ARRAY_REF too.
6570 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6573 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6574 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6576 (set_of_reg): New struct.
6577 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6578 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6579 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6581 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6582 Split into bitclr, bitset, bitinvert patterns if appropriate.
6583 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6584 use rx_fuse_in_memory_bitop.
6585 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6586 to named insn, correct maximum insn length.
6588 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6591 * machmode.def: Define a complex mode for PARTIAL_INT.
6592 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6594 * doc/rtl.texi: Document CSPImode.
6595 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6597 (msp430_hard_regno_nregs_with_padding): Likewise.
6599 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6602 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6604 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6606 PR rtl-optimization/84169
6607 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6608 we generated a parallel as new i3 and we split that to new i2 and i3
6609 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6610 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6611 those to i2, not i1. Partially rewrite this scan code.
6613 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6616 * stor-layout.c (place_field): For variable length fields, adjust
6617 offset_align afterwards not just based on the field's alignment,
6618 but also on the size.
6621 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6622 of exps and logs in the use_exp2 case.
6624 2018-02-13 Jeff Law <law@redhat.com>
6626 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6629 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6632 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6636 * common.opt (gas-loc-support, gas-locview-support): New.
6637 (ginline-points, ginternal-reset-location-views): New.
6638 * doc/invoke.texi: Document them. Use @itemx where intended.
6639 (gvariable-location-views): Adjust.
6640 * target.def (reset_location_view): New.
6641 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6642 (TARGET_RESET_LOCATION_VIEW): New.
6643 * doc/tm.texi: Rebuilt.
6644 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6645 (dwarf2out_default_as_locview_support): New.
6646 (output_asm_line_debug_info): Use option variables.
6647 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6648 (output_loc_list): Likewise.
6649 (add_high_low_attributes): Check option variables.
6650 Don't output entry view attribute in strict mode.
6651 (gen_inlined_subroutine_die): Check option variables.
6652 (dwarf2out_inline_entry): Likewise.
6653 (init_sections_and_labels): Likewise.
6654 (dwarf2out_early_finish): Likewise.
6655 (maybe_reset_location_view): New, from...
6656 (dwarf2out_var_location): ... here. Call it.
6657 * debug.h (dwarf2out_default_as_loc_support): Declare.
6658 (dwarf2out_default_as_locview_support): Declare.
6659 * hooks.c (hook_int_rtx_insn_0): New.
6660 * hooks.h (hook_int_rtx_insn_0): Declare.
6661 * toplev.c (process_options): Take -gas-loc-support and
6662 -gas-locview-support from dwarf2out. Enable
6663 -gvariable-location-views by default only with locview
6664 assembler support. Enable -ginternal-reset-location-views by
6665 default only if the target defines the corresponding hook.
6666 Enable -ginline-points by default if location views are
6667 enabled; force it disabled if statement frontiers are
6669 * tree-inline.c (expand_call_inline): Check option variables.
6670 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6672 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6674 PR tree-optimization/84321
6675 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6676 handling. Also check whether the anti-range contains any values
6677 that satisfy the mask; switch to a VR_RANGE if not.
6679 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6682 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6684 2018-02-13 Martin Jambor <mjambor@suse.cz>
6687 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6688 of call statements, also set location of a load to a temporary.
6690 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6692 * config/rl78/rl78.c (add_vector_labels): New function.
6693 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6694 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6695 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6696 which checks that no arguments are passed.
6697 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6698 * doc/extend.texi: Documentation for the new attribute.
6700 2018-02-13 Andreas Schwab <schwab@suse.de>
6702 * config/riscv/linux.h (CPP_SPEC): Define.
6704 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6707 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6708 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6709 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6710 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6711 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6712 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6713 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6714 temporarily for AES and PCLMUL builtins.
6716 PR tree-optimization/84339
6717 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6718 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6722 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6723 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6724 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6726 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6728 * omp-simd-clone.h: New file.
6729 * omp-simd-clone.c: Include omp-simd-clone.h.
6730 (expand_simd_clones): No longer static.
6731 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6732 cgraph.h and omp-simd-clone.h.
6733 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6734 (vect_recog_widen_shift_pattern): Formatting fix.
6735 (vect_pattern_recog_1): Don't check optab for calls.
6738 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6739 operands[2] into a REG before using gen_lowpart on it.
6741 2018-02-12 Jeff Law <law@redhat.com>
6744 * config/sh/sh.c (find_barrier): Consider a sibling call
6747 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6748 successfully back substituting a reg.
6750 2018-02-12 Richard Biener <rguenther@suse.de>
6752 PR tree-optimization/84037
6753 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6754 parameter, move visited init to caller.
6755 (vect_slp_analyze_operations): Separate cost from validity
6756 check, initialize visited once for all instances.
6757 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6759 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6761 (vect_model_promotion_demotion_cost): Likewise.
6762 (vectorizable_bswap): Guard cost modeling with !slp_node
6763 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6765 (vectorizable_call): Likewise.
6766 (vectorizable_conversion): Likewise.
6767 (vectorizable_assignment): Likewise.
6768 (vectorizable_shift): Likewise.
6769 (vectorizable_operation): Likewise.
6770 (vectorizable_store): Likewise.
6771 (vectorizable_load): Likewise.
6772 (vectorizable_condition): Likewise.
6773 (vectorizable_comparison): Likewise.
6775 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6778 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6779 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6780 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6782 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6785 * params.def (inline-min-speedup): Increase from 8 to 15.
6786 (max-inline-insns-auto): Decrease from 40 to 30.
6787 * ipa-split.c (consider_split): Add some buffer for function to
6788 be considered inlining candidate.
6789 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6792 2018-02-12 Richard Biener <rguenther@suse.de>
6794 PR tree-optimization/84037
6795 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6796 matched stmts if we cannot swap the non-matched ones.
6798 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6800 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6801 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6802 _mm_maskz_scalef_round_ss): New intrinsics.
6803 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6804 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6805 __builtin_ia32_scalefss_round): Remove.
6806 (__builtin_ia32_scalefsd_mask_round,
6807 __builtin_ia32_scalefss_mask_round): New intrinsics.
6808 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6809 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6810 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6811 "<round_constraint>")): Changed to ...
6812 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6813 "<round_scalar_constraint>")): ... this.
6814 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6815 %0, %1, %2<round_op3>}"): Changed to ...
6816 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6817 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6818 %2<round_scalar_mask_op3>}"): ... this.
6819 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6821 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6823 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6824 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6825 (_mm_maskz_sqrt_round_ss): New intrinsics.
6826 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6827 (__builtin_ia32_sqrtsd_mask_round)
6828 (__builtin_ia32_sqrtss_mask_round): New builtins.
6829 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6830 (__builtin_ia32_sqrtss_round): Remove.
6831 (__builtin_ia32_sqrtsd_mask_round)
6832 (__builtin_ia32_sqrtss_mask_round): New builtins.
6833 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6834 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6835 ((match_operand:VF_128 1 "vector_operand"
6836 "xBm,<round_constraint>")): Changed to ...
6837 ((match_operand:VF_128 1 "vector_operand"
6838 "xBm,<round_scalar_constraint>")): ... this.
6839 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6840 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6841 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6842 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6843 %<iptr>1<round_scalar_mask_op3>}): ... this.
6844 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6845 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6847 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6850 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6851 Cast vec_cmpeq result to correct type.
6852 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6853 Cast vec_cmpgt result to correct type.
6855 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6857 * final.c (final_scan_insn_1): Renamed from...
6858 (final_scan_insn): ... this. New wrapper, to recover
6859 seen from the outermost call in recursive ones.
6860 * config/sparc/sparc.c (output_return): Drop seen from call.
6861 (output_sibcall): Likewise.
6862 * config/visium/visium.c (output_branch): Likewise.
6864 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6866 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6869 2018-02-10 Alan Modra <amodra@gmail.com>
6872 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6873 Specify LR as an input.
6875 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6878 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6879 remove_member_access_dummy_vars): New functions.
6880 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6881 lower_omp_1, execute_lower_omp): Use them.
6883 PR rtl-optimization/84308
6884 * shrink-wrap.c (spread_components): Release todo vector.
6886 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6888 PR rtl-optimization/57193
6889 * ira-color.c (struct allocno_color_data): Add member
6890 conflict_allocno_hard_prefs.
6891 (update_conflict_allocno_hard_prefs): New.
6892 (bucket_allocno_compare_func): Add a preference based on
6893 conflict_allocno_hard_prefs.
6894 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6895 (color_allocnos): Remove a dead code. Initiate
6896 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6898 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6901 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6902 constraint from =wa to wa. Avoid a subreg on the output operand,
6903 instead use a pseudo and subreg it in a move.
6904 (p9_xxbrd_<mode>): Changed to ...
6905 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6906 (p9_xxbrd_v2df): New expander.
6907 (p9_xxbrw_<mode>): Changed to ...
6908 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6909 (p9_xxbrw_v4sf): New expander.
6911 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6913 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6915 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6918 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6919 multiply in 32-bit mode.
6920 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6921 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6924 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6926 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6927 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6928 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6929 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6931 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6934 * dwarf2out.c (is_trivial_indirect_ref): New function.
6935 (dwarf2out_late_global_decl): Do not generate a location
6936 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6937 and that are not defined in the current unit.
6939 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6941 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6942 instead of a libcall for UNORDERED.
6944 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6947 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6948 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6950 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6953 * config/s390/s390.c (s390_set_current_function): Invoke
6954 s390_indirect_branch_settings also if fndecl didn't change.
6956 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6958 * config/rs6000/rs6000.md (blockage): Set length to zero.
6960 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6962 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6964 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6967 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6968 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6972 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6973 PARALLEL incoming that failed vt_get_decl_and_offset check.
6976 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6977 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6979 (get_variable_section): For decls in named .bss* sections pass true as
6980 second argument to bss_initializer_p.
6982 2018-02-09 Marek Polacek <polacek@redhat.com>
6983 Jakub Jelinek <jakub@redhat.com>
6986 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6987 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6988 Sync some changes from cxx_fold_indirect_ref.
6990 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6992 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6994 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6995 (BLOCK_INLINE_ENTRY_LABEL): New.
6996 (dwarf2out_var_location): Disregard inline entry markers.
6997 (inline_entry_data): New struct.
6998 (inline_entry_data_hasher): New hashtable type.
6999 (inline_entry_data_hasher::hash): New.
7000 (inline_entry_data_hasher::equal): New.
7001 (inline_entry_data_table): New variable.
7002 (add_high_low_attributes): Add DW_AT_entry_pc and
7003 DW_AT_GNU_entry_view attributes if a pending entry is found
7004 in inline_entry_data_table. Add old entry_pc attribute only
7005 if debug nonbinding markers are disabled.
7006 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7007 markers are enabled.
7008 (block_within_block_p, dwarf2out_inline_entry): New.
7009 (dwarf2out_finish): Check that no entries remained in
7010 inline_entry_data_table.
7011 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7012 (final_scan_insn, notice_source_line): Likewise.
7013 (rest_of_clean_state): Skip inline entry markers.
7014 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7016 * gimple.c (gimple_build_debug_inline_entry): New.
7017 * gimple.h (enum gimple_debug_subcode): Add
7018 GIMPLE_DEBUG_INLINE_ENTRY.
7019 (gimple_build_debug_inline_entry): Declare.
7020 (gimple_debug_inline_entry_p): New.
7021 (gimple_debug_nonbind_marker_p): Adjust.
7022 * insn-notes.def (INLINE_ENTRY): New.
7023 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7024 inline entry marker notes.
7025 (print_insn): Likewise.
7026 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7027 (INSN_DEBUG_MARKER_KIND): Likewise.
7028 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7029 * tree-inline.c (expand_call_inline): Build and insert
7030 debug_inline_entry stmt.
7031 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7032 inline entry blocks early, if nonbind markers are enabled.
7033 (dump_scope_block): Dump fragment info.
7034 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7035 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7036 (gimple_build_debug_inline_entry): New.
7037 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7038 Enable/disable inline entry points too.
7039 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7040 (DEBUG_INSN): Describe inline entry markers.
7042 * common.opt (gvariable-location-views): New.
7043 (gvariable-location-views=incompat5): New.
7044 * config.in: Rebuilt.
7045 * configure: Rebuilt.
7046 * configure.ac: Test assembler for view support.
7047 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7048 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7049 * dwarf2out.c (var_loc_view): New typedef.
7050 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7051 (dwarf2out_locviews_in_attribute): New.
7052 (dwarf2out_locviews_in_loclist): New.
7053 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7054 (enum dw_line_info_opcode): Add LI_adv_address.
7055 (struct dw_line_info_table): Add view.
7056 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7057 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7058 (zero_view_p): New variable.
7059 (ZERO_VIEW_P): New macro.
7060 (output_asm_line_debug_info): New.
7061 (struct var_loc_node): Add view.
7062 (add_AT_view_list, AT_loc_list): New.
7063 (add_var_loc_to_decl): Add view param. Test it against last.
7064 (new_loc_list): Add view params. Record them.
7065 (AT_loc_list_ptr): Handle loc and view lists.
7066 (view_list_to_loc_list_val_node): New.
7067 (print_dw_val): Handle dw_val_class_view_list.
7068 (size_of_die): Likewise.
7069 (value_format): Likewise.
7070 (loc_list_has_views): New.
7071 (gen_llsym): Set vl_symbol too.
7072 (maybe_gen_llsym, skip_loc_list_entry): New.
7073 (dwarf2out_maybe_output_loclist_view_pair): New.
7074 (output_loc_list): Output view list or entries too.
7075 (output_view_list_offset): New.
7076 (output_die): Handle dw_val_class_view_list.
7077 (output_dwarf_version): New.
7078 (output_compilation_unit_header): Use it.
7079 (output_skeleton_debug_sections): Likewise.
7080 (output_rnglists, output_line_info): Likewise.
7081 (output_pubnames, output_aranges): Update version comments.
7082 (output_one_line_info_table): Output view numbers in asm comments.
7083 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7084 Call maybe_gen_llsym.
7085 (loc_list_from_tree_1): Adjust.
7086 (add_AT_location_description): Create view list attribute if
7087 needed, check it's absent otherwise.
7088 (convert_cfa_to_fb_loc_list): Adjust.
7089 (maybe_emit_file): Call output_asm_line_debug_info for test.
7090 (dwarf2out_var_location): Reset views as needed. Precompute
7091 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7092 attribute. Set view.
7093 (new_line_info_table): Reset next view.
7094 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7095 (dwarf2out_source_line): Likewise. Output view resets and labels to
7096 the assembler, or select appropriate line info opcodes.
7097 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7098 (optimize_string_length): Catch it. Adjust.
7099 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7100 dw_val_class_view_list, and remove it if no longer needed.
7101 (hash_loc_list): Hash view numbers.
7102 (loc_list_hasher::equal): Compare them.
7103 (optimize_location_lists): Check whether a view list symbol is
7104 needed, and whether the locview attribute is present, and
7105 whether they match. Remove the locview attribute if no longer
7107 (index_location_lists): Call skip_loc_list_entry for test.
7108 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7109 Use output_dwarf_version.
7110 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7111 (struct dw_val_node): Add val_view_list.
7112 * final.c (SEEN_NEXT_VIEW): New.
7113 (set_next_view_needed): New.
7114 (clear_next_view_needed): New.
7115 (maybe_output_next_view): New.
7116 (final_start_function): Rename to...
7117 (final_start_function_1): ... this. Take pointer to FIRST,
7118 add SEEN parameter. Emit param bindings in the initial view.
7119 (final_start_function): Reintroduce SEEN-less interface.
7120 (final): Rename to...
7121 (final_1): ... this. Take SEEN parameter. Output final pending
7122 next view at the end.
7123 (final): Reintroduce seen-less interface.
7124 (final_scan_insn): Output pending next view before switching
7125 sections or ending a block. Mark the next view as needed when
7126 outputting variable locations. Notify debug backend of section
7127 changes, and of location view changes.
7128 (rest_of_handle_final): Adjust.
7129 * toplev.c (process_options): Autodetect value for debug variable
7130 location views option. Warn on incompat5 without -gdwarf-5.
7131 * doc/invoke.texi (gvariable-location-views): New.
7132 (gvariable-location-views=incompat5): New.
7133 (gno-variable-location-views): New.
7135 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7137 PR tree-optimization/84136
7138 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7139 that the result of find_edge is non-NULL.
7141 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7144 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7145 storing integer register in SImode. Fix cost of 256 and 512
7146 byte aligned SSE register store.
7148 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7150 * config/i386/i386.c (ix86_multiplication_cost): Fix
7151 multiplication cost for TARGET_AVX512DQ.
7153 2018-02-08 Marek Polacek <polacek@redhat.com>
7155 PR tree-optimization/84238
7156 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7159 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7161 PR tree-optimization/84265
7162 * tree-vect-stmts.c (vectorizable_store): Don't treat
7163 VMAT_CONTIGUOUS accesses as grouped.
7164 (vectorizable_load): Likewise.
7166 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7168 PR tree-optimization/81635
7169 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7170 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7171 (test_round_for_mask): New functions.
7172 (wide_int_cc_tests): Call test_round_for_mask.
7173 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7174 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7175 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7176 range returned by get_range_info.
7178 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7181 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7182 * symtab.c: Include builtins.h
7183 (symtab_node::output_to_lto_symbol_table_p): Move here
7184 from lto-streamer-out.c:output_symbol_p.
7185 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7186 (output_symbol_p): Move all logic to symtab.c
7187 (produce_symtab): Update.
7189 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7191 * config/s390/s390-opts.h (enum indirect_branch): Define.
7192 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7193 (s390_indirect_branch_via_thunk)
7194 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7195 (enum s390_indirect_branch_type): Define.
7196 * config/s390/s390.c (struct s390_frame_layout, struct
7197 machine_function): Remove.
7198 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7199 (indirect_branch_table_label_no, indirect_branch_table_name):
7201 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7202 (enum s390_indirect_branch_option): Define.
7203 (s390_return_addr_from_memory): New function.
7204 (s390_handle_string_attribute): New function.
7205 (s390_attribute_table): Add new attribute handler.
7206 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7207 (s390_indirect_branch_via_thunk): New function.
7208 (s390_indirect_branch_via_inline_thunk): New function.
7209 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7210 sibling call optimization for non z10 compiles.
7211 (s390_emit_call): Force indirect branch target to be a single
7212 register. Add r1 clobber for non-z10 compiles.
7213 (s390_emit_epilogue): Emit return jump via return_use expander.
7214 (s390_reorg): Handle JUMP_INSNs as execute targets.
7215 (s390_option_override_internal): Perform validity checks for the
7216 new command line options.
7217 (s390_indirect_branch_attrvalue): New function.
7218 (s390_indirect_branch_settings): New function.
7219 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7220 (s390_output_indirect_thunk_function): New function.
7221 (s390_code_end): Implement target hook.
7222 (s390_case_values_threshold): Implement target hook.
7223 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7225 * config/s390/s390.h (struct s390_frame_layout)
7226 (struct machine_function): Move here from s390.c.
7227 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7228 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7229 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7230 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7231 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7232 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7233 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7234 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7235 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7236 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7237 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7238 (mnemonic attribute): Add values which aren't recognized
7240 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7241 pattern for branch conversion. Fix mnemonic attribute.
7242 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7243 indirect branch via thunk if requested.
7244 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7245 ("*indirect_jump"): Disable for branch conversion using out of
7247 ("indirect_jump_via_thunk<mode>_z10")
7248 ("indirect_jump_via_thunk<mode>")
7249 ("indirect_jump_via_inlinethunk<mode>_z10")
7250 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7251 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7252 ("casesi_jump_via_inlinethunk<mode>_z10")
7253 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7254 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7255 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7256 ("*indirect2_jump"): Disable for branch conversion.
7257 ("casesi_jump"): Turn into expander and expand patterns for branch
7259 ("return_use"): New expander.
7260 ("*return"): Emit return via thunk and rename it to ...
7261 ("*return<mode>"): ... this one.
7262 * config/s390/s390.opt: Add new options and and enum for the
7265 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7267 * lra-constraints.c (match_reload): Unconditionally use
7268 gen_lowpart_SUBREG, rather than selecting between that
7269 and equivalent gen_rtx_SUBREG code.
7271 2018-02-08 Richard Biener <rguenther@suse.de>
7273 PR tree-optimization/84233
7274 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7275 changed flag instead of boguously re-using phi_inserted.
7277 2018-02-08 Martin Jambor <mjambor@suse.cz>
7279 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7280 static local variables.
7282 2018-02-08 Richard Biener <rguenther@suse.de>
7284 PR tree-optimization/84278
7285 * tree-vect-stmts.c (vectorizable_store): When looking for
7286 smaller vector types to perform grouped strided loads/stores
7287 make sure the mode is supported by the target.
7288 (vectorizable_load): Likewise.
7290 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7292 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7293 Increase LDP/STP opportunities by adding adjacent callee-saves.
7295 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7297 PR rtl-optimization/84068
7298 PR rtl-optimization/83459
7299 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7301 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7303 PR tree-optimization/84224
7304 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7305 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7308 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7311 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7312 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7313 count, remove one USE.
7315 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7317 * doc/install.texi (Configuration): Document the
7318 --with-long-double-format={ibm,ieee} PowerPC configuration
7322 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7323 Convert from define_expand to be define_insn_and_split. Rework
7324 float/double/_Float128 conversions to QI/HI/SImode to work with
7325 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7326 conversions to QI/HImode types did a store and then a load to
7327 truncate the value. For conversions to VSX registers, don't split
7328 the insn, instead emit the code directly. Use the code iterator
7329 any_fix to combine signed and unsigned conversions.
7330 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7331 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7332 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7333 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7334 (fix_<mode>di2_hw): Likewise.
7335 (fixuns_<mode>di2_hw): Likewise.
7336 (fix_<mode>si2_hw): Likewise.
7337 (fixuns_<mode>si2_hw): Likewise.
7338 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7339 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7340 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7341 fix<uns>_trunc<SFDF:mode>si2_p8.
7342 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7344 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7345 (fix<uns>_<mode>_mem): Likewise.
7346 (fctiw<u>z_<mode>_mem): Likewise.
7347 (fix<uns>_<mode>_mem): Likewise.
7348 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7349 the register allocator from doing a direct move to the GPRs to do
7350 a store, and instead use the ISA 3.0 store byte/half-word from
7351 vector register instruction. For IEEE 128-bit floating point,
7352 also optimize stores of 32-bit ints.
7353 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7355 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7357 * genextract.c (push_pathstr_operand): New function to support
7359 (walk_rtx): Call push_pathstr_operand.
7360 (print_path): Support [a-zA-Z].
7362 2018-02-07 Richard Biener <rguenther@suse.de>
7364 PR tree-optimization/84037
7365 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7366 (cse_and_gimplify_to_preheader): Declare.
7367 (vect_get_place_in_interleaving_chain): Likewise.
7368 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7370 (_loop_vec_info::~_loop_vec_info): Delete it.
7371 (cse_and_gimplify_to_preheader): New function.
7372 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7373 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7374 (vectorizable_load): Likewise. For grouped stores always base
7375 the IV on the first element.
7376 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7377 condition before gimplifying.
7379 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7381 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7382 *DIV_EXPR and *MOD_EXPR.
7384 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7387 * config/i386/i386.c (ix86_option_override_internal): Mask out
7388 the CF_SET bit when checking -fcf-protection.
7390 2018-02-07 Tom de Vries <tom@codesourcery.com>
7393 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7396 2018-02-07 Richard Biener <rguenther@suse.de>
7398 PR tree-optimization/84204
7399 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7402 PR tree-optimization/84205
7403 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7404 special-case isl_ast_op_zdiv_r.
7406 PR tree-optimization/84223
7407 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7408 Only add conditions from within the region.
7409 (gather_bbs::after_dom_children): Adjust.
7411 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7414 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7415 * config/avr/avr.md: Only post-reload split REG-REG moves if
7416 either register is GENERAL_REG_P.
7418 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7420 PR tree-optimization/84235
7421 * tree-ssa-scopedtables.c
7422 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7423 if the subtraction is performed in floating point type where NaNs are
7424 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7425 build 1. Formatting fix.
7427 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7430 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7431 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7432 and skip it regardless of bb boundaries. Use CALL_P macro,
7433 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7434 unnecessarily, formatting fix.
7436 2018-02-06 Michael Collison <michael.collison@arm.com>
7438 * config/arm/thumb2.md:
7439 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7440 (*thumb_mov_notscc): Ditto.
7442 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7445 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7446 unsigned_fix, not "s".
7448 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7450 * configure.ac (gcc_fn_eh_frame_ro): New function.
7451 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7452 correct .eh_frame permissions.
7453 * configure: Regenerate.
7455 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7457 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7460 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7462 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7463 Display warning message for -mno-speculate-indirect-jumps.
7465 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7467 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7469 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7471 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7473 PR tree-optimization/84225
7474 * tree-eh.c (find_trapping_overflow): Only call
7475 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7477 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7480 * config/i386/i386.c: Reimplement the check of possible options
7481 -mibt/-mshstk conbination. Change error messages.
7482 * doc/invoke.texi: Fix a typo: remove extra '='.
7484 2018-02-06 Marek Polacek <polacek@redhat.com>
7486 PR tree-optimization/84228
7487 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7489 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7492 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7493 emitted arch directives.
7494 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7495 __ARM_FEATURE_COPROC before changing architectures.
7497 2018-02-06 Richard Biener <rguenther@suse.de>
7499 * config/i386/i386.c (print_reg): Fix typo.
7500 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7502 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7504 * configure: Regenerate.
7506 2018-02-05 Martin Sebor <msebor@redhat.com>
7508 PR tree-optimization/83369
7509 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7512 2018-02-05 Martin Liska <mliska@suse.cz>
7514 * doc/invoke.texi: Cherry-pick upstream r323995.
7516 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7518 * ira.c (ira_init_register_move_cost): Adjust comment.
7520 2018-02-05 Martin Liska <mliska@suse.cz>
7522 PR gcov-profile/84137
7523 * doc/gcov.texi: Fix typo in documentation.
7525 2018-02-05 Martin Liska <mliska@suse.cz>
7527 PR gcov-profile/83879
7528 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7529 using dlopen functionality.
7531 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7533 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7534 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7535 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7536 _mm_maskz_range_round_ss): New intrinsics.
7537 (__builtin_ia32_rangesd128_round)
7538 (__builtin_ia32_rangess128_round): Remove.
7539 (__builtin_ia32_rangesd128_mask_round,
7540 __builtin_ia32_rangess128_mask_round): New builtins.
7541 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7542 __builtin_ia32_rangess128_round): Remove.
7543 (__builtin_ia32_rangesd128_mask_round,
7544 __builtin_ia32_rangess128_mask_round): New builtins.
7545 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7546 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7547 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7548 "<round_saeonly_constraint>")): Changed to ...
7549 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7550 "<round_saeonly_scalar_constraint>")): ... this.
7551 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7552 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7553 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7554 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7555 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7557 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7559 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7561 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7562 Remove all values except native, 8540 and 8548.
7564 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7566 * config/i386/i386.c (ix86_output_function_return): Pass
7567 INVALID_REGNUM, instead of -1, as invalid register number to
7568 indirect_thunk_name and output_indirect_thunk.
7570 2018-02-02 Julia Koval <julia.koval@intel.com>
7572 * config.gcc: Add -march=icelake.
7573 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7574 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7575 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7576 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7577 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7578 (processor_target_table): Add icelake.
7579 (ix86_option_override_internal): Handle new PTAs.
7580 (get_builtin_code_for_version): Handle icelake.
7581 (M_INTEL_COREI7_ICELAKE): New.
7582 (fold_builtin_cpu): Handle icelake.
7583 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7584 * doc/invoke.texi: Add -march=icelake.
7586 2018-02-02 Julia Koval <julia.koval@intel.com>
7588 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7589 to wide_int_bitmask.
7590 * wide-int-bitmask.h: New.
7592 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7595 * config/i386/i386.md: Replace Pmode with word_mode in
7596 builtin_setjmp_setup and builtin_longjmp to support x32.
7598 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7602 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7604 (rs6000_supported_cpu_names): New static variable.
7605 (linux_cpu_translation_table): Likewise.
7606 (elf_platform) <cpu>: Define new static variable and use it.
7607 Translate kernel AT_PLATFORM name to canonical name if needed.
7608 Error if platform name is unknown.
7610 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7613 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7615 2018-02-01 Jeff Law <law@redhat.com>
7618 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7619 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7620 the scratch if RELEASE_VIA_POP is false.
7621 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7622 If we have to save a temporary register, decrement SIZE appropriately.
7623 Pass new arguments to release_scratch_register_on_entry.
7624 (ix86_adjust_stack_and_probe): Likewise.
7625 (ix86_emit_probe_stack_range): Pass new arguments to
7626 release_scratch_register_on_entry.
7628 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7630 PR rtl-optimization/84157
7631 * combine.c (change_zero_ext): Use REG_P predicate in
7632 front of HARD_REGISTER_P predicate.
7634 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7636 * config/avr/avr.c (avr_option_override): Move disabling of
7637 -fdelete-null-pointer-checks to...
7638 * common/config/avr/avr-common.c (avr_option_optimization_table):
7641 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7643 PR tree-optimization/81635
7644 * tree-data-ref.c (split_constant_offset_1): For types that
7645 wrap on overflow, try to use range info to prove that wrapping
7648 2018-02-01 Renlin Li <renlin.li@arm.com>
7651 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7653 (aarch64_register_move_cost): Likewise.
7654 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7656 (REG_CLASS_NAMES): Likewise.
7657 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7658 TAILCALL_ADDR_REGS. Remove IP registers.
7659 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7661 2018-02-01 Richard Biener <rguenther@suse.de>
7663 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7664 for specifying RPO order and allow NULL for that.
7665 * domwalk.c (dom_walker::dom_walker): Likewise.
7666 (dom_walker::walk): Handle NULL RPO order.
7667 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7669 (rewrite_update_dom_walker): Likewise.
7670 (mark_def_dom_walker): Likewise.
7672 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7674 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7675 (aarch64_maybe_expand_sve_subreg_move): Declare.
7676 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7677 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7679 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7680 that are semantically a reverse operation.
7681 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7682 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7683 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7685 (aarch64_can_change_mode_class): For big-endian, forbid changes
7686 between two SVE modes if they have different element sizes.
7688 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7690 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7691 the TImode handling for big-endian targets.
7693 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7695 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7696 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7698 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7699 Remove BSWAP handing for big-endian targets and use the form of
7700 LD1RQ appropariate for the mode.
7702 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7704 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7705 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7708 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7711 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7712 check for operands that need to go through aarch64_sve_reload_be.
7714 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7716 PR tree-optimization/81661
7717 PR tree-optimization/84117
7718 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7719 * tree-eh.c: Include gimplify.h.
7720 (find_trapping_overflow, replace_trapping_overflow,
7721 rewrite_to_non_trapping_overflow): New functions.
7722 * tree-vect-loop.c: Include tree-eh.h.
7723 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7724 * tree-data-ref.c: Include tree-eh.h.
7725 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7727 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7729 PR rtl-optimization/84123
7730 * combine.c (change_zero_ext): Check if hard register satisfies
7731 can_change_dest_mode before calling gen_lowpart_SUBREG.
7733 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7736 * ira.c (ira_init_register_move_cost): Remove assert.
7738 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7740 PR rtl-optimization/84071
7741 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7742 * doc/tm.texi: Regenerate.
7744 2018-01-31 Richard Biener <rguenther@suse.de>
7746 PR tree-optimization/84132
7747 * tree-data-ref.c (analyze_miv_subscript): Properly
7748 check whether evolution_function_is_affine_multivariate_p
7749 before calling gcd_of_steps_may_divide_p.
7751 2018-01-31 Julia Koval <julia.koval@intel.com>
7754 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7755 * config/i386/i386.md (rdpid_rex64) New.
7756 (rdpid): Make 32bit only.
7758 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7761 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7762 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7764 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7767 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7769 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7771 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7773 PR rtl-optimization/84071
7774 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7775 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7777 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7779 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7780 (arc_attribute_table): Add 'aux' attribute.
7781 (arc_in_small_data_p): Consider aux like variables.
7782 (arc_is_aux_reg_p): New function.
7783 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7784 (arc_get_aux_arg): New function.
7785 (prepare_move_operands): Handle aux-register access.
7786 (arc_handle_aux_attribute): New function.
7787 * doc/extend.texi (ARC Variable attributes): Add subsection.
7789 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7791 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7792 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7793 (arc_attribute_table): Add 'uncached' attribute.
7794 (arc_print_operand): Print '.di' flag for uncached memory
7796 (arc_in_small_data_p): Do not consider for small data the uncached
7798 (arc_is_uncached_mem_p): New function.
7799 * config/arc/predicates.md (compact_store_memory_operand): Check
7800 for uncached memory accesses.
7801 (nonvol_nonimm_operand): Likewise.
7802 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7804 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7807 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7808 falign-loops=): Add Optimization flag.
7810 2018-01-30 Jeff Law <law@redhat.com>
7813 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7814 INT_REGISTERS_SAVED. Check it prior to calling
7815 get_scratch_register_on_entry.
7816 (ix86_adjust_stack_and_probe): Similarly.
7817 (ix86_emit_probe_stack_range): Similarly.
7818 (ix86_expand_prologue): Corresponding changes.
7820 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7823 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7824 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7826 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7829 * lra-constraints.c (curr_insn_transform): Process AND in the
7832 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7834 PR rtl-optimization/83986
7835 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7836 dependence against last_pending_memory_flush in addition to
7839 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7841 PR tree-optimization/81611
7842 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7845 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7848 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7851 2018-01-30 Richard Biener <rguenther@suse.de>
7852 Jakub Jelinek <jakub@redhat.com>
7854 PR tree-optimization/84111
7855 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7856 inner loops added during recursion, as they don't have up-to-date
7859 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7862 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7863 (can_inline_edge_by_limits_p): ... here.
7864 (can_early_inline_edge_p, check_callers,
7865 update_caller_keys, update_callee_keys, recursive_inlining,
7866 add_new_edges_to_heap, speculation_useful_p,
7867 inline_small_functions,
7868 inline_small_functions, flatten_function,
7869 inline_to_all_callers_1): Update.
7871 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7873 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7876 2018-01-30 Richard Biener <rguenther@suse.de>
7878 PR tree-optimization/83008
7879 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7880 invariant and constant vector uses in stmts when they need
7883 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7886 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7887 * configure: Regenerate.
7889 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7891 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7893 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7894 Use gen_rtx_REG rather than gen_lowpart.
7896 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7898 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7899 rather than 0 when creating partial subregs.
7901 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7903 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7906 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7909 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7910 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7911 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7912 flags. This restores the settings used before the 2017-07-24.
7913 Turning off pre increment/decrement/modify allows IVOPTS to
7914 optimize DF/SF loops where the index is an int.
7916 2018-01-29 Richard Biener <rguenther@suse.de>
7917 Kelvin Nilsen <kelvin@gcc.gnu.org>
7920 * tree-vect-stmts.c (vectorizable_call): Don't call
7921 targetm.vectorize_builtin_md_vectorized_function if callee is
7924 2018-01-22 Carl Love <cel@us.ibm.com>
7926 * doc/extend.tex: Fix typo in second arg in
7927 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7929 2018-01-29 Richard Biener <rguenther@suse.de>
7931 PR tree-optimization/84086
7932 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7933 (flush_ssaname_freelist): When SSA names were released reset
7934 the SCEV hash table.
7936 2018-01-29 Richard Biener <rguenther@suse.de>
7938 PR tree-optimization/84057
7939 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7940 removed paths when removing edges.
7942 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7944 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7945 -mfunction-return=@var{choice}.
7947 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7950 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7952 (layout::print_source_line): Likewise.
7953 (test_get_line_width_without_trailing_whitespace): Add test cases.
7955 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7958 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7961 2018-01-26 Jim Wilson <jimw@sifive.com>
7963 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7965 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7968 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7970 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7971 and CMP + SUB-immediate -> SUBS.
7973 2018-01-26 Martin Sebor <msebor@redhat.com>
7975 PR tree-optimization/83896
7976 * tree-ssa-strlen.c (get_string_len): Rename...
7977 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7978 Avoid assuming length is constant.
7979 (handle_char_store): Use HOST_WIDE_INT for string length.
7981 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7984 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7985 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7987 2018-01-26 Richard Biener <rguenther@suse.de>
7989 PR rtl-optimization/84003
7990 * dse.c (record_store): Only record redundant stores when
7991 the earlier store aliases at least all accesses the later one does.
7993 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7995 PR rtl-optimization/83985
7996 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7997 REG_CFA_RESTORE insns.
7998 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7999 REG_CFA_RESTORE insns here.
8002 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8003 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8005 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8007 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8008 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8009 (arc_init): Likewise.
8010 (arc_override_options): Likewise.
8011 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8013 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8015 * config/arc/arc.h (TARGET_DBNZ): Define.
8016 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8017 properly set the tune attribute.
8018 (dbnz): Use TARGET_DBNZ guard.
8019 * config/arc/arc.opt (mtune): Add core3 option.
8021 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8023 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8024 recognize new pic like addresses.
8025 (arc_delegitimize_address): Clean up.
8027 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8029 * config/arc/arc-arches.def: Option mrf16 valid for all
8031 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8032 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8033 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8034 * config/arc/arc-tables.opt: Regenerate.
8035 * config/arc/arc.c (arc_conditional_register_usage): Handle
8036 reduced register file case.
8037 (arc_file_start): Set must have build attributes.
8038 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8040 * config/arc/arc.opt (mrf16): Add new option.
8041 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8042 * config/arc/genmultilib.awk: Handle new mrf16 option.
8043 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8044 * config/arc/t-multilib: Regenerate.
8045 * doc/invoke.texi (ARC Options): Document mrf16 option.
8047 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8049 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8050 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8051 (arc_attribute_table): Add 'secure_call' attribute.
8052 (arc_print_operand): Print secure call operand.
8053 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8055 (arc_is_secure_call_p): New function. * config/arc/arc.md
8056 (call_i): Add support for sjli instruction.
8057 (call_value_i): Likewise.
8058 * config/arc/constraints.md (Csc): New constraint.
8060 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8061 John Eric Martin <John.Martin@emmicro-us.com>
8063 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8064 * config/arc/arc.c (_arc_jli_section): New struct.
8065 (arc_jli_section): New type.
8066 (rc_jli_sections): New static variable.
8067 (arc_handle_jli_attribute): New function.
8068 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8069 (arc_file_end): New function.
8070 (TARGET_ASM_FILE_END): Define.
8071 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8072 (arc_add_jli_section): New function.
8073 (jli_call_scan): Likewise.
8074 (arc_reorg): Call jli_call_scan.
8075 (arc_output_addsi): Remove 'S' from printing asm operand.
8076 (arc_is_jli_call_p): New function.
8077 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8079 (movhi_insn): Likewise.
8080 (movsi_insn): Likewise.
8081 (movsi_set_cc_insn): Likewise.
8082 (loadqi_update): Likewise.
8083 (load_zeroextendqisi_update): Likewise.
8084 (load_signextendqisi_update): Likewise.
8085 (loadhi_update): Likewise.
8086 (load_zeroextendhisi_update): Likewise.
8087 (load_signextendhisi_update): Likewise.
8088 (loadsi_update): Likewise.
8089 (loadsf_update): Likewise.
8090 (movsicc_insn): Likewise.
8091 (bset_insn): Likewise.
8092 (bxor_insn): Likewise.
8093 (bclr_insn): Likewise.
8094 (bmsk_insn): Likewise.
8095 (bicsi3_insn): Likewise.
8096 (cmpsi_cc_c_insn): Likewise.
8097 (movsi_ne): Likewise.
8098 (movsi_cond_exec): Likewise.
8099 (clrsbsi2): Likewise.
8110 (call_i): Remove 'S' asm letter, add jli instruction.
8111 (call_value_i): Likewise.
8112 * config/arc/arc.op (mjli-always): New option.
8113 * config/arc/constraints.md (Cji): New constraint.
8114 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8116 (subsf3_fpx): Likewise.
8117 (mulsf3_fpx): Likewise.
8118 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8120 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8122 * doc/invoke.texi (ARC): Document mjli-always option.
8124 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8126 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8127 avoid addition with 0 and use incw and decw where possible.
8129 2018-01-26 Richard Biener <rguenther@suse.de>
8131 PR tree-optimization/81082
8132 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8133 association if it requires casting to unsigned.
8134 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8135 from fold_plusminus_mult_expr to catch important cases late when
8136 range info is available.
8138 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8140 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8141 * configure.ac (hidden_linkonce): New test.
8142 * configure: Regenerate.
8143 * config.in: Regenerate.
8145 2018-01-26 Julia Koval <julia.koval@intel.com>
8147 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8148 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8149 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8150 _mm_mask_bitshuffle_epi64_mask): Fix type.
8151 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8152 USI_FTYPE_V4DI_V4DI_USI): Remove.
8153 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8154 __builtin_ia32_vpshufbitqmb256_mask,
8155 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8156 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8157 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8159 2018-01-26 Alan Modra <amodra@gmail.com>
8162 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8163 UNSPEC_VBPERMQ. Sort other unspecs.
8165 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8167 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8169 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8172 * predict.c (drop_profile): Do not push/pop cfun; update also
8174 (handle_missing_profiles): Fix logic looking for zero profiles.
8176 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8179 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8180 on functions with #pragma omp declare simd or functions with simd
8182 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8183 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8184 Remove trailing \n from warning_at calls.
8186 2018-01-25 Tom de Vries <tom@codesourcery.com>
8189 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8190 for neutered workers.
8192 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8195 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8196 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8198 2018-01-24 Jeff Law <law@redhat.com>
8201 * i386.c (get_probe_interval): Move to earlier point.
8202 (ix86_compute_frame_layout): If -fstack-clash-protection and
8203 the frame is larger than the probe interval, then use pushes
8204 to save registers rather than reg->mem moves.
8205 (ix86_expand_prologue): Remove conditional for int_registers_saved
8208 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8211 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8212 min/max for never referenced object.
8214 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8217 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8219 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8220 attributes from DECL_ATTRIBUTES (decl) without affecting
8221 DECL_ATTRIBUTES (current_function_decl).
8222 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8223 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8225 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8227 PR tree-optimization/83979
8228 * fold-const.c (fold_comparison): Use constant_boolean_node
8229 instead of boolean_{true,false}_node.
8231 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8233 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8236 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8238 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8239 Simplify the clause that sets the length attribute.
8240 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8241 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8242 clause that sets the length attribute.
8243 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8245 2018-01-24 Tom de Vries <tom@codesourcery.com>
8248 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8249 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8250 Add strict parameter.
8251 (prevent_branch_around_nothing): Insert dummy insn between branch to
8252 label and label with no ptx insn inbetween.
8253 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8255 2018-01-24 Tom de Vries <tom@codesourcery.com>
8258 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8259 for neutered threads in warp.
8260 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8262 2018-01-24 Richard Biener <rguenther@suse.de>
8264 PR tree-optimization/83176
8265 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8268 2018-01-24 Richard Biener <rguenther@suse.de>
8270 PR tree-optimization/82819
8271 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8272 code generating pluses that are no-ops in the target precision.
8274 2018-01-24 Richard Biener <rguenther@suse.de>
8277 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8279 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8281 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8282 to merge probabilities.
8283 * predict.c (probably_never_executed): Also mark as cold functions
8284 with global 0 profile and guessed local profile.
8285 * profile-count.c (profile_probability::combine_with_count): New
8287 * profile-count.h (profile_probability::operator*,
8288 profile_probability::operator*=, profile_probability::operator/,
8289 profile_probability::operator/=): Reduce precision to adjusted
8290 and set value to guessed on contradictory divisions.
8291 (profile_probability::combine_with_freq): Remove.
8292 (profile_probability::combine_wiht_count): Declare.
8293 (profile_count::force_nonzero):: Set to adjusted.
8294 (profile_count::probability_in):: Set quality to adjusted.
8295 * tree-ssa-tail-merge.c (replace_block_by): Use
8298 2018-01-23 Andrew Waterman <andrew@sifive.com>
8299 Jim Wilson <jimw@sifive.com>
8301 * config/riscv/riscv.c (riscv_stack_boundary): New.
8302 (riscv_option_override): Set riscv_stack_boundary. Handle
8303 riscv_preferred_stack_boundary_arg.
8304 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8305 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8306 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8307 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8308 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8309 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8311 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8314 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8315 of struct ix86_frame.
8316 (ix86_expand_epilogue): Likewise. Add a local variable for
8317 the reg_save_offset field in struct ix86_frame.
8319 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8321 PR tree-optimization/82604
8322 * tree-loop-distribution.c (enum partition_kind): New enum item
8323 PKIND_PARTIAL_MEMSET.
8324 (partition_builtin_p): Support above new enum item.
8325 (generate_code_for_partition): Ditto.
8326 (compute_access_range): Differentiate cases that equality can be
8327 proven at all loops, the innermost loops or no loops.
8328 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8329 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8330 (finalize_partitions, distribute_loop): Don't fuse partition of
8331 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8332 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8335 2018-01-23 Martin Liska <mliska@suse.cz>
8337 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8338 order to ignore the predictor.
8339 (PRED_POLYMORPHIC_CALL): Likewise.
8340 (PRED_RECURSIVE_CALL): Likewise.
8342 2018-01-23 Martin Liska <mliska@suse.cz>
8344 * tree-profile.c (tree_profiling): Print function header to
8345 aware reader which function we are working on.
8346 * value-prof.c (gimple_find_values_to_profile): Do not print
8347 not interesting value histograms.
8349 2018-01-23 Martin Liska <mliska@suse.cz>
8351 * profile-count.h (enum profile_quality): Add
8352 profile_uninitialized as the first value. Do not number values
8353 as they are zero based.
8354 (profile_count::verify): Update sanity check.
8355 (profile_probability::verify): Likewise.
8357 2018-01-23 Nathan Sidwell <nathan@acm.org>
8359 * doc/invoke.texi (ffor-scope): Deprecate.
8361 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8363 PR tree-optimization/83510
8364 * domwalk.c (set_all_edges_as_executable): New function.
8365 (dom_walker::dom_walker): Convert bool param
8366 "skip_unreachable_blocks" to enum reachability. Move setup of
8367 edge flags to set_all_edges_as_executable and only do it when
8368 reachability is REACHABLE_BLOCKS.
8369 * domwalk.h (enum dom_walker::reachability): New enum.
8370 (dom_walker::dom_walker): Convert bool param
8371 "skip_unreachable_blocks" to enum reachability.
8372 (set_all_edges_as_executable): New decl.
8373 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8374 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8376 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8377 but converting true to REACHABLE_BLOCKS.
8378 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8380 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8381 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8382 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8384 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8385 if check_all_array_refs will be called.
8387 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8389 * tree.c (selftest::test_location_wrappers): Add more test
8392 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8394 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8395 (selftest::test_bit_in_range): Likewise.
8397 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8400 * doc/sourcebuild.texi (vect_float): Say that the selector
8401 only describes the situation when -funsafe-math-optimizations is on.
8402 (vect_float_strict): Document.
8404 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8406 PR tree-optimization/83965
8407 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8408 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8409 instead of checking only for a reduction.
8410 (vect_recog_widen_sum_pattern): Likewise.
8412 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8414 * predict.c (probably_never_executed): Only use precise profile info.
8415 (compute_function_frequency): Skip after inlining hack since we now
8416 have quality checking.
8418 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8420 * profile-count.h (profile_probability::very_unlikely,
8421 profile_probability::unlikely, profile_probability::even): Set
8422 precision to guessed.
8424 2018-01-23 Richard Biener <rguenther@suse.de>
8426 PR tree-optimization/83963
8427 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8428 Properly terminate dominator walk when crossing the exit edge not
8429 when visiting its source block.
8431 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8434 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8435 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8437 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8439 PR tree-optimization/83957
8440 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8441 semicolon after for body surrounded by braces.
8443 PR tree-optimization/83081
8444 * profile-count.h (profile_probability::split): New method.
8445 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8446 Use profile_probability::split.
8447 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8448 when splitting a single conditional jump into 2.
8450 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8452 PR tree-optimization/69452
8453 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8456 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8458 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8459 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8460 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8462 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8464 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8465 * config/rl78/rl78.md (movdi): New define_expand.
8466 * config/rl78/rl78.c (rl78_split_movdi): New function.
8468 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8471 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8473 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8474 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8475 128-bit to produce an UNSPEC move to get the double word with the
8476 signbit and then a shift directly to do signbit.
8477 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8478 implementation with a new version that just does either a direct
8479 move or a regular move. Move memory interface to separate insns.
8480 Move insns so they are next to the expander.
8481 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8482 with signbit move. Split big and little endian case.
8483 (signbit<mode>2_dm_mem_le): Likewise.
8484 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8485 (signbit<mode>2_dm2): Likewise.
8487 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8489 * config/rl78/rl78.md (anddi3): New define_expand.
8491 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8493 * config/rl78/rl78.md (umindi3): New define_expand.
8495 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8497 * config/rl78/rl78.md (smindi3): New define_expand.
8499 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8501 * config/rl78/rl78.md (smaxdi3): New define_expand.
8503 2018-01-22 Carl Love <cel@us.ibm.com>
8505 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8506 LVX_V1TI): Add macro expansion.
8507 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8508 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8509 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8510 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8511 Change check to determine if the instruction is a byte reversing
8512 entry. Fix typo in comment.
8513 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8514 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8515 Add def_builtin calls for new builtins.
8516 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8517 Add define_insn expansion.
8519 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8521 * config/rl78/rl78.md (umaxdi3): New define_expand.
8523 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8525 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8526 for non-QImode registers.
8528 2018-01-22 Richard Biener <rguenther@suse.de>
8530 PR tree-optimization/83963
8531 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8532 including the loop exit block.
8533 (scop_detection::merge_sese): Likewise.
8534 (scop_detection::add_scop): Do it here instead.
8536 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8538 * doc/sourcebuild.texi (arm_softfloat): Document.
8540 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8543 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8544 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8545 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8547 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8548 David Edelsohn <dje.gcc@gmail.com>
8551 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8552 Change "crset eq" to "crset 2".
8553 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8554 (*call_indirect_aix<mode>_nospec): Likewise.
8555 (*call_value_indirect_aix<mode>_nospec): Likewise.
8556 (*call_indirect_elfv2<mode>_nospec): Likewise.
8557 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8558 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8559 change assembly output from . to $.
8560 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8561 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8562 (*tablejump<mode>_internal1_nospec): Likewise.
8564 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8567 * config/sh/sh_optimize_sett_clrt.cc:
8568 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8570 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8572 PR tree-optimization/83940
8573 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8574 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8575 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8576 use it to pass back the definition type.
8577 (vect_check_store_rhs): Likewise rhs_dt_out.
8578 (vect_build_gather_load_calls): Add a mask_dt argument and use
8579 it instead of a call to vect_is_simple_use.
8580 (vectorizable_store): Update calls to vect_check_load_store_mask
8581 and vect_check_store_rhs. Use the dt returned by the latter instead
8582 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8583 instead of calls to vect_is_simple_use. Pass the scalar rather
8584 than the vector operand to vect_is_simple_use when handling
8585 second and subsequent copies of an rhs value.
8586 (vectorizable_load): Update calls to vect_check_load_store_mask
8587 and vect_build_gather_load_calls. Use the cached mask_dt and
8588 gs_info.offset_dt instead of calls to vect_is_simple_use.
8590 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8593 * tree-emutls.c: Include gimplify.h.
8594 (lower_emutls_2): New function.
8595 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8596 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8597 it before further processing.
8600 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8601 UINTVAL (trueop1) instead of INTVAL (op1).
8603 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8607 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8608 INCOMING_FRAME_SP_OFFSET if not defined.
8609 (scan_trace): Add ENTRY argument. If true and
8610 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8611 emit a note to adjust the CFA offset.
8612 (create_cfi_notes): Adjust scan_trace callers.
8613 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8614 INCOMING_FRAME_SP_OFFSET in the CIE.
8615 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8616 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8618 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8619 * doc/tm.texi: Regenerated.
8621 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8623 PR rtl-optimization/83147
8624 * lra-constraints.c (remove_inheritance_pseudos): Use
8625 lra_substitute_pseudo_within_insn.
8627 2018-01-19 Tom de Vries <tom@codesourcery.com>
8628 Cesar Philippidis <cesar@codesourcery.com>
8631 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8633 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8636 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8637 spaces for function labels.
8639 2018-01-19 Martin Liska <mliska@suse.cz>
8641 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8642 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8643 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8644 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8645 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8646 (PRED_CONST_RETURN): Change from 69 to 65.
8647 (PRED_NULL_RETURN): Change from 91 to 71.
8648 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8649 (PRED_LOOP_GUARD): Change from 66 to 73.
8651 2018-01-19 Martin Liska <mliska@suse.cz>
8653 * predict.c (predict_insn_def): Add new assert.
8654 (struct branch_predictor): Change type to signed integer.
8655 (test_prediction_value_range): Amend test to cover
8657 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8658 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8659 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8660 (PRED_LOOP_IV_COMPARE): Likewise.
8661 * predict.h (PROB_UNINITIALIZED): Define new constant.
8663 2018-01-19 Martin Liska <mliska@suse.cz>
8665 * predict.c (dump_prediction): Add new format for
8666 analyze_brprob.py script which is enabled with -details
8668 * profile-count.h (precise_p): New function.
8670 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8672 PR tree-optimization/83922
8673 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8674 there are no statements that need masking.
8675 (vect_active_double_reduction_p): New function.
8676 (vect_analyze_loop_operations): Use it when handling phis that
8677 are not in the loop header.
8679 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8681 PR tree-optimization/83914
8682 * tree-vect-loop.c (vectorizable_induction): Don't convert
8683 init_expr or apply the peeling adjustment for inductions
8684 that are nested within the vectorized loop.
8686 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8688 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8691 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8695 * function.h (gimplify_parameters): Add gimple_seq * argument.
8696 * function.c: Include gimple.h and options.h.
8697 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8698 for the added local temporaries if needed.
8699 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8700 if there are any parameter cleanups, wrap whole body into a
8701 try/finally with the cleanups.
8703 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8706 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8707 Use GET_MODE_CLASS for scalar floating point.
8709 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8713 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8714 Fix call of call_cgraph_insertion_hooks.
8716 2018-01-18 Martin Sebor <msebor@redhat.com>
8718 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8720 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8723 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8726 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8729 * common.opt: (-ffile-prefix-map): New option.
8730 * opts.c (common_handle_option): Defer it.
8731 * opts-global.c (handle_common_deferred_options): Handle it.
8732 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8733 * file-prefix-map.h: New file.
8734 (remap_debug_filename, add_debug_prefix_map): ...here.
8735 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8736 * final.c (debug_prefix_map, add_debug_prefix_map
8737 remap_debug_filename): Move to...
8738 * file-prefix-map.c: New file.
8739 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8740 generalize, get rid of alloca(), use strrchr() instead of strchr().
8741 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8742 Implement in terms of add_prefix_map().
8743 (remap_macro_filename, remap_debug_filename): Implement in term of
8745 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8746 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8747 * dbxout.c: Include file-prefix-map.h.
8748 * varasm.c: Likewise.
8749 * vmsdbgout.c: Likewise.
8750 * xcoffout.c: Likewise.
8751 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8752 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8753 * doc/invoke.texi (-ffile-prefix-map): Document.
8754 (-fdebug-prefix-map): Update description.
8756 2018-01-18 Martin Liska <mliska@suse.cz>
8758 * config/i386/i386.c (indirect_thunk_name): Document that also
8760 (output_indirect_thunk): Document why both instructions
8761 (pause and lfence) are generated.
8763 2018-01-18 Richard Biener <rguenther@suse.de>
8765 PR tree-optimization/83887
8766 * graphite-scop-detection.c
8767 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8768 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8769 (scop_detection::merge_sese): Re-implement with a flood-fill
8770 algorithm that properly finds a SESE region if it exists.
8772 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8775 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8776 pointer_diff optimizations use view_convert instead of convert.
8778 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8780 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8781 Generate different code for -mno-speculate-indirect-jumps.
8782 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8783 (*call_indirect_aix<mode>): Disable for
8784 -mno-speculate-indirect-jumps.
8785 (*call_indirect_aix<mode>_nospec): New define_insn.
8786 (*call_value_indirect_aix<mode>): Disable for
8787 -mno-speculate-indirect-jumps.
8788 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8789 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8790 -mno-speculate-indirect-jumps.
8791 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8793 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8795 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8796 long double type, set the flags for noting the default long double
8797 type, even if we don't pass or return a long double type.
8799 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8802 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8805 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8807 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8808 support for merge[hl].
8809 (fold_mergehl_helper): New helper function.
8810 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8811 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8812 (altivec_vmrglw_direct): Add xxmrglw insn.
8814 2018-01-17 Andrew Waterman <andrew@sifive.com>
8816 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8817 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8819 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8822 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8823 call the lto_location_cache before reading the
8824 DECL_SOURCE_LOCATION of the types.
8826 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8827 Richard Sandiford <richard.sandiford@linaro.org>
8829 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8830 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8831 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8832 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8833 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8835 * config/aarch64/constraints.md (aarch64_movti_operand):
8837 * config/aarch64/predicates.md (Uti): Add new constraint.
8839 2018-01-17 Carl Love <cel@us.ibm.com>
8841 * config/rs6000/vsx.md (define_expand xl_len_r,
8842 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8843 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8845 (define_expand, define_insn): Move the shift left from the
8846 define_insn to the define_expand for lxvl and stxvl instructions.
8847 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8848 and XL_LEN_R definitions to PURE.
8850 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8852 * config/i386/i386.c (indirect_thunk_name): Declare regno
8853 as unsigned int. Compare regno with INVALID_REGNUM.
8854 (output_indirect_thunk): Ditto.
8855 (output_indirect_thunk_function): Ditto.
8856 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8857 in the call to output_indirect_thunk_function.
8859 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8862 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8863 rather than the size of inner_type to determine the stack slot size
8864 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8866 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8869 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8872 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8874 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8875 endian Linux systems to optionally enable multilibs for selecting
8876 the long double type if the user configured an explicit type.
8877 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8878 have no long double multilibs if not defined.
8879 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8880 warn if the user used -mabi={ieee,ibm}longdouble and we built
8881 multilibs for long double.
8882 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8883 appropriate multilib option.
8884 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8886 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8887 for building long double multilibs.
8888 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8890 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8892 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8895 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8897 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8900 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8903 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8906 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8908 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8909 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8911 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8913 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8914 different rtl trees depending on TARGET_64BIT.
8915 (rs6000_gen_lvx): Likewise.
8917 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8919 * config/visium/visium.md (nop): Tweak comment.
8920 (hazard_nop): Likewise.
8922 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8924 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8925 -mspeculate-indirect-jumps.
8926 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8927 for -mno-speculate-indirect-jumps.
8928 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8929 (*call_value_indirect_elfv2<mode>): Disable for
8930 -mno-speculate-indirect-jumps.
8931 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8932 (indirect_jump): Emit different RTL for
8933 -mno-speculate-indirect-jumps.
8934 (*indirect_jump<mode>): Disable for
8935 -mno-speculate-indirect-jumps.
8936 (*indirect_jump<mode>_nospec): New define_insn.
8937 (tablejump): Emit different RTL for
8938 -mno-speculate-indirect-jumps.
8939 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8940 (tablejumpsi_nospec): New define_expand.
8941 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8942 (tablejumpdi_nospec): New define_expand.
8943 (*tablejump<mode>_internal1): Disable for
8944 -mno-speculate-indirect-jumps.
8945 (*tablejump<mode>_internal1_nospec): New define_insn.
8946 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8949 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8951 * caller-save.c (insert_save): Drop unnecessary parameter. All
8954 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8955 Richard Biener <rguenth@suse.de>
8958 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8959 return early, inline manually is_gimple_sizepos. Make sure if we
8960 call gimplify_expr we don't end up with a gimple constant.
8961 * tree.c (variably_modified_type_p): Don't return true for
8962 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8963 * gimplify.h (is_gimple_sizepos): Remove.
8965 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8967 PR tree-optimization/83857
8968 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8969 vectorizable_live_operation for pure SLP statements.
8970 (vectorizable_live_operation): Handle PHIs.
8972 2018-01-16 Richard Biener <rguenther@suse.de>
8974 PR tree-optimization/83867
8975 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8976 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8978 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8981 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8982 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8983 If off is not INTEGER_CST, issue a may not be aligned warning
8984 rather than isn't aligned. Use isn%'t rather than isn't.
8985 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8987 <case MULT_EXPR>: Improve the case when bottom and one of the
8988 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8989 operand, in that case check if the other operand is multiple of
8990 bottom divided by the INTEGER_CST operand.
8992 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8995 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8996 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8997 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8998 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8999 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9000 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9001 (pa_function_arg_size): New function.
9003 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9005 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9006 in a separate statement.
9008 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9010 PR tree-optimization/83847
9011 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9012 group gathers and scatters.
9014 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9016 PR rtl-optimization/86620
9017 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9019 PR rtl-optimization/83213
9020 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9021 to last if both are JUMP_INSNs.
9023 PR tree-optimization/83843
9024 * gimple-ssa-store-merging.c
9025 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9026 store_immediate_info for bswap/nop orig_stores.
9028 2018-01-15 Andrew Waterman <andrew@sifive.com>
9030 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9032 <UDIV>: Increase cost if !TARGET_DIV.
9034 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9036 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9037 (define_attr "cr_logical_3op"): New.
9038 (cceq_ior_compare): Adjust.
9039 (cceq_ior_compare_complement): Adjust.
9040 (*cceq_rev_compare): Adjust.
9041 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9042 (is_cracked_insn): Adjust.
9043 (insn_must_be_first_in_group): Adjust.
9044 * config/rs6000/40x.md: Adjust.
9045 * config/rs6000/440.md: Adjust.
9046 * config/rs6000/476.md: Adjust.
9047 * config/rs6000/601.md: Adjust.
9048 * config/rs6000/603.md: Adjust.
9049 * config/rs6000/6xx.md: Adjust.
9050 * config/rs6000/7450.md: Adjust.
9051 * config/rs6000/7xx.md: Adjust.
9052 * config/rs6000/8540.md: Adjust.
9053 * config/rs6000/cell.md: Adjust.
9054 * config/rs6000/e300c2c3.md: Adjust.
9055 * config/rs6000/e500mc.md: Adjust.
9056 * config/rs6000/e500mc64.md: Adjust.
9057 * config/rs6000/e5500.md: Adjust.
9058 * config/rs6000/e6500.md: Adjust.
9059 * config/rs6000/mpc.md: Adjust.
9060 * config/rs6000/power4.md: Adjust.
9061 * config/rs6000/power5.md: Adjust.
9062 * config/rs6000/power6.md: Adjust.
9063 * config/rs6000/power7.md: Adjust.
9064 * config/rs6000/power8.md: Adjust.
9065 * config/rs6000/power9.md: Adjust.
9066 * config/rs6000/rs64.md: Adjust.
9067 * config/rs6000/titan.md: Adjust.
9069 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9071 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9072 ix86_indirect_branch_register logic.
9074 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9076 * config/i386/constraints.md (Bs): Update
9077 ix86_indirect_branch_register check. Don't check
9078 ix86_indirect_branch_register with GOT_memory_operand.
9080 * config/i386/predicates.md (GOT_memory_operand): Don't check
9081 ix86_indirect_branch_register here.
9082 (GOT32_symbol_operand): Likewise.
9084 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9086 * config/i386/predicates.md (constant_call_address_operand):
9087 Rewrite ix86_indirect_branch_register logic.
9088 (sibcall_insn_operand): Likewise.
9090 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9092 * config/i386/constraints.md (Bs): Replace
9093 ix86_indirect_branch_thunk_register with
9094 ix86_indirect_branch_register.
9096 * config/i386/i386.md (indirect_jump): Likewise.
9097 (tablejump): Likewise.
9098 (*sibcall_memory): Likewise.
9099 (*sibcall_value_memory): Likewise.
9100 Peepholes of indirect call and jump via memory: Likewise.
9101 * config/i386/i386.opt: Likewise.
9102 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9103 (GOT_memory_operand): Likewise.
9104 (call_insn_operand): Likewise.
9105 (sibcall_insn_operand): Likewise.
9106 (GOT32_symbol_operand): Likewise.
9108 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9111 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9112 type rather than type addr's type points to.
9113 (expand_omp_atomic_mutex): Likewise.
9114 (expand_omp_atomic): Likewise.
9116 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9119 * config/i386/i386.c (output_indirect_thunk_function): Use
9120 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9121 for __x86_return_thunk.
9123 2018-01-15 Richard Biener <rguenther@suse.de>
9126 * expmed.c (extract_bit_field_1): Fix typo.
9128 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9131 * config/arm/iterators.md (VF): New mode iterator.
9132 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9133 Remove integer-related logic from pattern.
9134 (neon_vabd<mode>_3): Likewise.
9136 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9139 * common.opt (fstrict-overflow): No longer an alias.
9140 (fwrapv-pointer): New option.
9141 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9142 also for pointer types based on flag_wrapv_pointer.
9143 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9144 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9145 opts->x_flag_wrapv got set.
9146 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9147 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9148 POINTER_TYPE_OVERFLOW_UNDEFINED.
9149 * match.pd: Likewise in address comparison pattern.
9150 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9152 2018-01-15 Richard Biener <rguenther@suse.de>
9155 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9156 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9157 Reset type names to their identifier if their TYPE_DECL doesn't
9158 have linkage (and thus is used for ODR and devirt).
9159 (save_debug_info_for_decl): Remove.
9160 (save_debug_info_for_type): Likewise.
9161 (add_tree_to_fld_list): Adjust.
9162 * tree-pretty-print.c (dump_generic_node): Make dumping of
9163 type names more robust.
9165 2018-01-15 Richard Biener <rguenther@suse.de>
9167 * BASE-VER: Bump to 8.0.1.
9169 2018-01-14 Martin Sebor <msebor@redhat.com>
9172 * builtins.c (check_access): Avoid warning when the no-warning bit
9175 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9177 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9178 * ira-color (allocno_hard_regs_compare): Likewise.
9180 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9183 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9184 Use .pushsection/.popsection.
9186 2018-01-14 Martin Sebor <msebor@redhat.com>
9189 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9191 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9193 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9194 entry from extra_headers.
9195 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9196 extra_headers, make the list bitwise identical to the i?86-*-* one.
9198 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9200 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9201 -mcmodel=large with -mindirect-branch=thunk,
9202 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9203 -mfunction-return=thunk-extern.
9204 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9205 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9206 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9208 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9210 * config/i386/i386.c (print_reg): Print the name of the full
9211 integer register without '%'.
9212 (ix86_print_operand): Handle 'V'.
9213 * doc/extend.texi: Document 'V' modifier.
9215 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9217 * config/i386/constraints.md (Bs): Disallow memory operand for
9218 -mindirect-branch-register.
9220 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9221 (GOT_memory_operand): Likewise.
9222 (call_insn_operand): Likewise.
9223 (sibcall_insn_operand): Likewise.
9224 (GOT32_symbol_operand): Likewise.
9225 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9226 for -mindirect-branch-register.
9227 (tablejump): Likewise.
9228 (*sibcall_memory): Likewise.
9229 (*sibcall_value_memory): Likewise.
9230 Disallow peepholes of indirect call and jump via memory for
9231 -mindirect-branch-register.
9232 (*call_pop): Replace m with Bw.
9233 (*call_value_pop): Likewise.
9234 (*sibcall_pop_memory): Replace m with Bs.
9235 * config/i386/i386.opt (mindirect-branch-register): New option.
9236 * doc/invoke.texi: Document -mindirect-branch-register option.
9238 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9240 * config/i386/i386-protos.h (ix86_output_function_return): New.
9241 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9242 set function_return_type.
9243 (indirect_thunk_name): Add ret_p to indicate thunk for function
9245 (output_indirect_thunk_function): Pass false to
9246 indirect_thunk_name.
9247 (ix86_output_indirect_branch_via_reg): Likewise.
9248 (ix86_output_indirect_branch_via_push): Likewise.
9249 (output_indirect_thunk_function): Create alias for function
9250 return thunk if regno < 0.
9251 (ix86_output_function_return): New function.
9252 (ix86_handle_fndecl_attribute): Handle function_return.
9253 (ix86_attribute_table): Add function_return.
9254 * config/i386/i386.h (machine_function): Add
9255 function_return_type.
9256 * config/i386/i386.md (simple_return_internal): Use
9257 ix86_output_function_return.
9258 (simple_return_internal_long): Likewise.
9259 * config/i386/i386.opt (mfunction-return=): New option.
9260 (indirect_branch): Mention -mfunction-return=.
9261 * doc/extend.texi: Document function_return function attribute.
9262 * doc/invoke.texi: Document -mfunction-return= option.
9264 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9266 * config/i386/i386-opts.h (indirect_branch): New.
9267 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9268 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9269 with local indirect jump when converting indirect call and jump.
9270 (ix86_set_indirect_branch_type): New.
9271 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9272 (indirectlabelno): New.
9273 (indirect_thunk_needed): Likewise.
9274 (indirect_thunk_bnd_needed): Likewise.
9275 (indirect_thunks_used): Likewise.
9276 (indirect_thunks_bnd_used): Likewise.
9277 (INDIRECT_LABEL): Likewise.
9278 (indirect_thunk_name): Likewise.
9279 (output_indirect_thunk): Likewise.
9280 (output_indirect_thunk_function): Likewise.
9281 (ix86_output_indirect_branch_via_reg): Likewise.
9282 (ix86_output_indirect_branch_via_push): Likewise.
9283 (ix86_output_indirect_branch): Likewise.
9284 (ix86_output_indirect_jmp): Likewise.
9285 (ix86_code_end): Call output_indirect_thunk_function if needed.
9286 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9288 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9289 (ix86_attribute_table): Add indirect_branch.
9290 * config/i386/i386.h (machine_function): Add indirect_branch_type
9291 and has_local_indirect_jump.
9292 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9294 (tablejump): Likewise.
9295 (*indirect_jump): Use ix86_output_indirect_jmp.
9296 (*tablejump_1): Likewise.
9297 (simple_return_indirect_internal): Likewise.
9298 * config/i386/i386.opt (mindirect-branch=): New option.
9299 (indirect_branch): New.
9302 (thunk-inline): Likewise.
9303 (thunk-extern): Likewise.
9304 * doc/extend.texi: Document indirect_branch function attribute.
9305 * doc/invoke.texi: Document -mindirect-branch= option.
9307 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9310 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9312 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9314 * ipa-inline.c (want_inline_small_function_p): Return false if
9315 inlining has already failed with CIF_FINAL_ERROR.
9316 (update_caller_keys): Call want_inline_small_function_p before
9318 (update_callee_keys): Likewise.
9320 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9322 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9324 (rs6000_quadword_masked_address_p): Likewise.
9325 (quad_aligned_load_p): Likewise.
9326 (quad_aligned_store_p): Likewise.
9327 (const_load_sequence_p): Add comment to describe the outer-most loop.
9328 (mimic_memory_attributes_and_flags): New function.
9329 (rs6000_gen_stvx): Likewise.
9330 (replace_swapped_aligned_store): Likewise.
9331 (rs6000_gen_lvx): Likewise.
9332 (replace_swapped_aligned_load): Likewise.
9333 (replace_swapped_load_constant): Capitalize argument name in
9334 comment describing this function.
9335 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9336 and stores that access quad-word aligned addresses and replace
9337 with stvx or lvx instructions when appropriate.
9338 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9339 New function prototype.
9340 (rs6000_quadword_masked_address_p): Likewise.
9341 (rs6000_gen_lvx): Likewise.
9342 (rs6000_gen_stvx): Likewise.
9343 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9344 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9345 when memory address is aligned.
9346 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9347 this split to select lvx instruction when memory address is aligned.
9348 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9349 instruction when memory address is aligned.
9350 (*vsx_le_perm_load_v16qi): Likewise.
9351 (four unnamed splitters): Modify to select the stvx instruction
9352 when memory is aligned.
9354 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9356 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9357 which appears in the queue multiple times.
9359 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9360 Alan Hayward <alan.hayward@arm.com>
9361 David Sherwood <david.sherwood@arm.com>
9363 * tree-vectorizer.h (vec_lower_bound): New structure.
9364 (_loop_vec_info): Add check_nonzero and lower_bounds.
9365 (LOOP_VINFO_CHECK_NONZERO): New macro.
9366 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9367 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9368 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9369 fields. Make seg_len the distance travelled, not including the
9371 (dr_direction_indicator): Declare.
9372 (dr_zero_step_indicator): Likewise.
9373 (dr_known_forward_stride_p): Likewise.
9374 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9376 (runtime_alias_check_p): Allow runtime alias checks with
9378 (operator ==): Compare access_size and align.
9379 (prune_runtime_alias_test_list): Rework for new distinction between
9380 the access_size and seg_len.
9381 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9383 (get_segment_min_max): New function.
9384 (create_intersect_range_checks): Use it.
9385 (dr_step_indicator): New function.
9386 (dr_direction_indicator): Likewise.
9387 (dr_zero_step_indicator): Likewise.
9388 (dr_known_forward_stride_p): Likewise.
9389 * tree-loop-distribution.c (data_ref_segment_size): Return
9390 DR_STEP * (niters - 1).
9391 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9393 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9394 (vect_preserves_scalar_order_p): New function, split out from...
9395 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9396 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9397 (vect_vfa_access_size): New function.
9398 (vect_vfa_align): Likewise.
9399 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9400 (dump_lower_bound): New function.
9401 (vect_check_lower_bound): Likewise.
9402 (vect_small_gap_p): Likewise.
9403 (vectorizable_with_step_bound_p): Likewise.
9404 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9405 depencies if the vectorization factor is 1. Convert the checks
9406 for nonzero steps into checks on the bounds of DR_STEP. Try using
9407 a bunds check for variable steps if the minimum required step is
9408 relatively small. Update calls to the dr_with_seg_len
9409 constructor and to vect_compile_time_alias.
9410 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9412 (vect_loop_versioning): Call it.
9413 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9415 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9417 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9418 Alan Hayward <alan.hayward@arm.com>
9419 David Sherwood <david.sherwood@arm.com>
9421 * doc/sourcebuild.texi (vect_scatter_store): Document.
9422 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9424 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9426 * genopinit.c (main): Add supports_vec_scatter_store and
9427 supports_vec_scatter_store_cached to target_optabs.
9428 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9429 IFN_MASK_SCATTER_STORE.
9430 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9432 * internal-fn.h (internal_store_fn_p): Declare.
9433 (internal_fn_stored_value_index): Likewise.
9434 * internal-fn.c (scatter_store_direct): New macro.
9435 (expand_scatter_store_optab_fn): New function.
9436 (direct_scatter_store_optab_supported_p): New macro.
9437 (internal_store_fn_p): New function.
9438 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9439 IFN_MASK_SCATTER_STORE.
9440 (internal_fn_mask_index): Likewise.
9441 (internal_fn_stored_value_index): New function.
9442 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9444 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9445 * optabs-query.c (supports_vec_scatter_store_p): New function.
9446 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9447 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9448 true for scatter stores.
9449 (vect_gather_scatter_fn_p): Handle scatter stores too.
9450 (vect_check_gather_scatter): Consider using scatter stores if
9451 supports_vec_scatter_store_p.
9452 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9454 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9455 internal_fn_stored_value_index.
9456 (check_load_store_masking): Handle scatter stores too.
9457 (vect_get_store_rhs): Make public.
9458 (vectorizable_call): Use internal_store_fn_p.
9459 (vectorizable_store): Handle scatter store internal functions.
9460 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9461 when deciding whether the end of the group has been reached.
9462 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9463 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9464 (mask_scatter_store<mode>): New insns.
9466 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9467 Alan Hayward <alan.hayward@arm.com>
9468 David Sherwood <david.sherwood@arm.com>
9470 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9471 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9472 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9474 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9475 Use vect_truncate_gather_scatter_offset if we can't treat the
9476 operation as a normal gather load or scatter store.
9477 (get_group_load_store_type): Take the gather_scatter_info
9478 as argument. Try using a gather load or scatter store for
9479 single-element groups.
9480 (get_load_store_type): Update calls to get_group_load_store_type
9481 and vect_use_strided_gather_scatters_p.
9483 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9484 Alan Hayward <alan.hayward@arm.com>
9485 David Sherwood <david.sherwood@arm.com>
9487 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9488 optional tree argument.
9489 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9491 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9492 but continue to use the current value as a fallback.
9493 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9494 to compare the updates.
9495 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9496 (get_load_store_type): Use it when handling a strided access.
9497 (vect_get_strided_load_store_ops): New function.
9498 (vect_get_data_ptr_increment): Likewise.
9499 (vectorizable_load): Handle strided gather loads. Always pass
9500 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9502 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9503 Alan Hayward <alan.hayward@arm.com>
9504 David Sherwood <david.sherwood@arm.com>
9506 * doc/md.texi (gather_load@var{m}): Document.
9507 (mask_gather_load@var{m}): Likewise.
9508 * genopinit.c (main): Add supports_vec_gather_load and
9509 supports_vec_gather_load_cached to target_optabs.
9510 * optabs-tree.c (init_tree_optimization_optabs): Use
9511 ggc_cleared_alloc to allocate target_optabs.
9512 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9513 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9515 * internal-fn.h (internal_load_fn_p): Declare.
9516 (internal_gather_scatter_fn_p): Likewise.
9517 (internal_fn_mask_index): Likewise.
9518 (internal_gather_scatter_fn_supported_p): Likewise.
9519 * internal-fn.c (gather_load_direct): New macro.
9520 (expand_gather_load_optab_fn): New function.
9521 (direct_gather_load_optab_supported_p): New macro.
9522 (direct_internal_fn_optab): New function.
9523 (internal_load_fn_p): Likewise.
9524 (internal_gather_scatter_fn_p): Likewise.
9525 (internal_fn_mask_index): Likewise.
9526 (internal_gather_scatter_fn_supported_p): Likewise.
9527 * optabs-query.c (supports_at_least_one_mode_p): New function.
9528 (supports_vec_gather_load_p): Likewise.
9529 * optabs-query.h (supports_vec_gather_load_p): Declare.
9530 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9531 and memory_type field.
9532 (NUM_PATTERNS): Bump to 15.
9533 * tree-vect-data-refs.c: Include internal-fn.h.
9534 (vect_gather_scatter_fn_p): New function.
9535 (vect_describe_gather_scatter_call): Likewise.
9536 (vect_check_gather_scatter): Try using internal functions for
9537 gather loads. Recognize existing calls to a gather load function.
9538 (vect_analyze_data_refs): Consider using gather loads if
9539 supports_vec_gather_load_p.
9540 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9541 (vect_get_gather_scatter_offset_type): Likewise.
9542 (vect_convert_mask_for_vectype): Likewise.
9543 (vect_add_conversion_to_patterm): Likewise.
9544 (vect_try_gather_scatter_pattern): Likewise.
9545 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9546 (vect_vect_recog_func_ptrs): Add it.
9547 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9548 internal_fn_mask_index and internal_gather_scatter_fn_p.
9549 (check_load_store_masking): Take the gather_scatter_info as an
9550 argument and handle gather loads.
9551 (vect_get_gather_scatter_ops): New function.
9552 (vectorizable_call): Check internal_load_fn_p.
9553 (vectorizable_load): Likewise. Handle gather load internal
9555 (vectorizable_store): Update call to check_load_store_masking.
9556 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9557 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9558 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9559 (aarch64_gather_scale_operand_d): New predicates.
9560 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9561 (mask_gather_load<mode>): New insns.
9563 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9564 Alan Hayward <alan.hayward@arm.com>
9565 David Sherwood <david.sherwood@arm.com>
9567 * optabs.def (fold_left_plus_optab): New optab.
9568 * doc/md.texi (fold_left_plus_@var{m}): Document.
9569 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9570 * internal-fn.c (fold_left_direct): Define.
9571 (expand_fold_left_optab_fn): Likewise.
9572 (direct_fold_left_optab_supported_p): Likewise.
9573 * fold-const-call.c (fold_const_fold_left): New function.
9574 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9575 * tree-parloops.c (valid_reduction_p): New function.
9576 (gather_scalar_reductions): Use it.
9577 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9578 (vect_finish_replace_stmt): Declare.
9579 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9580 (needs_fold_left_reduction_p): New function, split out from...
9581 (vect_is_simple_reduction): ...here. Accept reductions that
9582 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9583 (vect_force_simple_reduction): Also store the reduction type in
9584 the assignment's STMT_VINFO_REDUC_TYPE.
9585 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9586 (merge_with_identity): New function.
9587 (vect_expand_fold_left): Likewise.
9588 (vectorize_fold_left_reduction): Likewise.
9589 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9590 scalar phi in place for it. Check for target support and reject
9591 cases that would reassociate the operation. Defer the transform
9592 phase to vectorize_fold_left_reduction.
9593 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9594 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9595 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9597 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9599 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9600 call to ifc_temp_var.
9602 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9603 Alan Hayward <alan.hayward@arm.com>
9604 David Sherwood <david.sherwood@arm.com>
9606 * target.def (legitimize_address_displacement): Take the original
9607 offset as a poly_int.
9608 * targhooks.h (default_legitimize_address_displacement): Update
9610 * targhooks.c (default_legitimize_address_displacement): Likewise.
9611 * doc/tm.texi: Regenerate.
9612 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9613 as an argument, moving assert of ad->disp == ad->disp_term to...
9614 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9615 Try calling targetm.legitimize_address_displacement before expanding
9616 the address rather than afterwards, and adjust for the new interface.
9617 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9618 Match the new hook interface. Handle SVE addresses.
9619 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9622 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9624 * Makefile.in (OBJS): Add early-remat.o.
9625 * target.def (select_early_remat_modes): New hook.
9626 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9627 * doc/tm.texi: Regenerate.
9628 * targhooks.h (default_select_early_remat_modes): Declare.
9629 * targhooks.c (default_select_early_remat_modes): New function.
9630 * timevar.def (TV_EARLY_REMAT): New timevar.
9631 * passes.def (pass_early_remat): New pass.
9632 * tree-pass.h (make_pass_early_remat): Declare.
9633 * early-remat.c: New file.
9634 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9636 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9638 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9639 Alan Hayward <alan.hayward@arm.com>
9640 David Sherwood <david.sherwood@arm.com>
9642 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9643 vfm1 with a bound_epilog parameter.
9644 (vect_do_peeling): Update calls accordingly, and move the prologue
9645 call earlier in the function. Treat the base bound_epilog as 0 for
9646 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9647 this base when peeling for gaps.
9648 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9649 with fully-masked loops.
9650 (vect_estimate_min_profitable_iters): Handle the single peeled
9651 iteration in that case.
9653 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9654 Alan Hayward <alan.hayward@arm.com>
9655 David Sherwood <david.sherwood@arm.com>
9657 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9658 single-element interleaving even if the size is not a power of 2.
9659 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9660 accesses for single-element interleaving if the group size is
9663 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9664 Alan Hayward <alan.hayward@arm.com>
9665 David Sherwood <david.sherwood@arm.com>
9667 * doc/md.texi (fold_extract_last_@var{m}): Document.
9668 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9669 * optabs.def (fold_extract_last_optab): New optab.
9670 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9671 * internal-fn.c (fold_extract_direct): New macro.
9672 (expand_fold_extract_optab_fn): Likewise.
9673 (direct_fold_extract_optab_supported_p): Likewise.
9674 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9675 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9676 EXTRACT_LAST_REDUCTION.
9677 (get_initial_def_for_reduction): Do not create an initial vector
9678 for EXTRACT_LAST_REDUCTION reductions.
9679 (vectorizable_reduction): Leave the scalar phi in place for
9680 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9681 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9682 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9683 transform phase to vectorizable_condition.
9684 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9686 (vect_finish_stmt_generation): ...here.
9687 (vect_finish_replace_stmt): New function.
9688 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9689 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9691 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9693 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9694 Alan Hayward <alan.hayward@arm.com>
9695 David Sherwood <david.sherwood@arm.com>
9697 * doc/md.texi (extract_last_@var{m}): Document.
9698 * optabs.def (extract_last_optab): New optab.
9699 * internal-fn.def (EXTRACT_LAST): New internal function.
9700 * internal-fn.c (cond_unary_direct): New macro.
9701 (expand_cond_unary_optab_fn): Likewise.
9702 (direct_cond_unary_optab_supported_p): Likewise.
9703 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9704 loops using EXTRACT_LAST.
9705 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9706 (extract_last_<mode>): ...this optab.
9707 (vec_extract<mode><Vel>): Update accordingly.
9709 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9710 Alan Hayward <alan.hayward@arm.com>
9711 David Sherwood <david.sherwood@arm.com>
9713 * target.def (empty_mask_is_expensive): New hook.
9714 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9715 * doc/tm.texi: Regenerate.
9716 * targhooks.h (default_empty_mask_is_expensive): Declare.
9717 * targhooks.c (default_empty_mask_is_expensive): New function.
9718 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9719 if the target says that empty masks are expensive.
9720 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9722 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9724 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9725 Alan Hayward <alan.hayward@arm.com>
9726 David Sherwood <david.sherwood@arm.com>
9728 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9729 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9730 (vect_use_loop_mask_for_alignment_p): New function.
9731 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9732 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9733 niters_skip argument. Make sure that the first niters_skip elements
9734 of the first iteration are inactive.
9735 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9736 Update call to vect_set_loop_masks_directly.
9737 (get_misalign_in_elems): New function, split out from...
9738 (vect_gen_prolog_loop_niters): ...here.
9739 (vect_update_init_of_dr): Take a code argument that specifies whether
9740 the adjustment should be added or subtracted.
9741 (vect_update_init_of_drs): Likewise.
9742 (vect_prepare_for_masked_peels): New function.
9743 (vect_do_peeling): Skip prologue peeling if we're using a mask
9744 instead. Update call to vect_update_inits_of_drs.
9745 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9747 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9748 alignment. Do not include the number of peeled iterations in
9749 the minimum threshold in that case.
9750 (vectorizable_induction): Adjust the start value down by
9751 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9752 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9753 Take the number of skipped iterations into account when calculating
9755 * tree-vect-stmts.c (vect_gen_while_not): New function.
9757 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9758 Alan Hayward <alan.hayward@arm.com>
9759 David Sherwood <david.sherwood@arm.com>
9761 * doc/sourcebuild.texi (vect_fully_masked): Document.
9762 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9764 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9766 (vect_analyze_loop_2): ...here. Don't check the vectorization
9767 factor against the number of loop iterations if the loop is
9770 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9771 Alan Hayward <alan.hayward@arm.com>
9772 David Sherwood <david.sherwood@arm.com>
9774 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9775 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9776 (dump_groups): Update accordingly.
9777 (iv_use::mem_type): New member variable.
9778 (address_p): New function.
9779 (record_use): Add a mem_type argument and initialize the new
9781 (record_group_use): Add a mem_type argument. Use address_p.
9782 Remove obsolete null checks of base_object. Update call to record_use.
9783 (find_interesting_uses_op): Update call to record_group_use.
9784 (find_interesting_uses_cond): Likewise.
9785 (find_interesting_uses_address): Likewise.
9786 (get_mem_type_for_internal_fn): New function.
9787 (find_address_like_use): Likewise.
9788 (find_interesting_uses_stmt): Try find_address_like_use before
9789 calling find_interesting_uses_op.
9790 (addr_offset_valid_p): Use the iv mem_type field as the type
9791 of the addressed memory.
9792 (add_autoinc_candidates): Likewise.
9793 (get_address_cost): Likewise.
9794 (split_small_address_groups_p): Use address_p.
9795 (split_address_groups): Likewise.
9796 (add_iv_candidate_for_use): Likewise.
9797 (autoinc_possible_for_pair): Likewise.
9798 (rewrite_groups): Likewise.
9799 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9800 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9801 (get_alias_ptr_type_for_ptr_address): New function.
9802 (rewrite_use_address): Rewrite address uses in calls that were
9803 identified by find_address_like_use.
9805 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9806 Alan Hayward <alan.hayward@arm.com>
9807 David Sherwood <david.sherwood@arm.com>
9809 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9811 * gimple-expr.h (is_gimple_addressable: Likewise.
9812 * gimple-expr.c (is_gimple_address): Likewise.
9813 * internal-fn.c (expand_call_mem_ref): New function.
9814 (expand_mask_load_optab_fn): Use it.
9815 (expand_mask_store_optab_fn): Likewise.
9817 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9818 Alan Hayward <alan.hayward@arm.com>
9819 David Sherwood <david.sherwood@arm.com>
9821 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9822 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9823 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9824 (cond_umax@var{mode}): Document.
9825 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9826 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9827 (cond_umin_optab, cond_umax_optab): New optabs.
9828 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9829 (COND_IOR, COND_XOR): New internal functions.
9830 * internal-fn.h (get_conditional_internal_fn): Declare.
9831 * internal-fn.c (cond_binary_direct): New macro.
9832 (expand_cond_binary_optab_fn): Likewise.
9833 (direct_cond_binary_optab_supported_p): Likewise.
9834 (get_conditional_internal_fn): New function.
9835 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9836 Cope with reduction statements that are vectorized as calls rather
9838 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9839 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9840 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9841 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9842 (UNSPEC_COND_EOR): New unspecs.
9843 (optab): Add mappings for them.
9844 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9845 (sve_int_op, sve_fp_op): New int attributes.
9847 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9848 Alan Hayward <alan.hayward@arm.com>
9849 David Sherwood <david.sherwood@arm.com>
9851 * optabs.def (while_ult_optab): New optab.
9852 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9853 * internal-fn.def (WHILE_ULT): New internal function.
9854 * internal-fn.h (direct_internal_fn_supported_p): New override
9855 that takes two types as argument.
9856 * internal-fn.c (while_direct): New macro.
9857 (expand_while_optab_fn): New function.
9858 (convert_optab_supported_p): Likewise.
9859 (direct_while_optab_supported_p): New macro.
9860 * wide-int.h (wi::udiv_ceil): New function.
9861 * tree-vectorizer.h (rgroup_masks): New structure.
9862 (vec_loop_masks): New typedef.
9863 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9865 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9866 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9867 (vect_max_vf): New function.
9868 (slpeel_make_loop_iterate_ntimes): Delete.
9869 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9870 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9871 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9872 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9873 internal-fn.h, stor-layout.h and optabs-query.h.
9874 (vect_set_loop_mask): New function.
9875 (add_preheader_seq): Likewise.
9876 (add_header_seq): Likewise.
9877 (interleave_supported_p): Likewise.
9878 (vect_maybe_permute_loop_masks): Likewise.
9879 (vect_set_loop_masks_directly): Likewise.
9880 (vect_set_loop_condition_masked): Likewise.
9881 (vect_set_loop_condition_unmasked): New function, split out from
9882 slpeel_make_loop_iterate_ntimes.
9883 (slpeel_make_loop_iterate_ntimes): Rename to..
9884 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9885 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9886 (vect_do_peeling): Update call accordingly.
9887 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9889 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9890 mask_compare_type, can_fully_mask_p and fully_masked_p.
9891 (release_vec_loop_masks): New function.
9892 (_loop_vec_info): Use it to free the loop masks.
9893 (can_produce_all_loop_masks_p): New function.
9894 (vect_get_max_nscalars_per_iter): Likewise.
9895 (vect_verify_full_masking): Likewise.
9896 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9897 retries, and free the mask rgroups before retrying. Check loop-wide
9898 reasons for disallowing fully-masked loops. Make the final decision
9899 about whether use a fully-masked loop or not.
9900 (vect_estimate_min_profitable_iters): Do not assume that peeling
9901 for the number of iterations will be needed for fully-masked loops.
9902 (vectorizable_reduction): Disable fully-masked loops.
9903 (vectorizable_live_operation): Likewise.
9904 (vect_halve_mask_nunits): New function.
9905 (vect_double_mask_nunits): Likewise.
9906 (vect_record_loop_mask): Likewise.
9907 (vect_get_loop_mask): Likewise.
9908 (vect_transform_loop): Handle the case in which the final loop
9909 iteration might handle a partial vector. Call vect_set_loop_condition
9910 instead of slpeel_make_loop_iterate_ntimes.
9911 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9912 (check_load_store_masking): New function.
9913 (prepare_load_store_mask): Likewise.
9914 (vectorizable_store): Handle fully-masked loops.
9915 (vectorizable_load): Likewise.
9916 (supportable_widening_operation): Use vect_halve_mask_nunits for
9918 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9919 (vect_gen_while): New function.
9920 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9921 (aarch64_uqdec<mode>): New insn.
9923 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9924 Alan Hayward <alan.hayward@arm.com>
9925 David Sherwood <david.sherwood@arm.com>
9927 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9928 (reduc_xor_scal_optab): New optabs.
9929 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9930 (reduc_xor_scal_@var{m}): Document.
9931 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9932 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9934 * fold-const-call.c (fold_const_call): Handle them.
9935 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9936 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9937 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9938 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9939 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9940 (UNSPEC_XORV): New unspecs.
9941 (optab): Add entries for them.
9942 (BITWISEV): New int iterator.
9943 (bit_reduc_op): New int attributes.
9945 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9946 Alan Hayward <alan.hayward@arm.com>
9947 David Sherwood <david.sherwood@arm.com>
9949 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9950 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9951 * optabs.def (vec_shl_insert_optab): New optab.
9952 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9953 (duplicate_and_interleave): Likewise.
9954 * tree-vect-loop.c: Include internal-fn.h.
9955 (neutral_op_for_slp_reduction): New function, split out from
9956 get_initial_defs_for_reduction.
9957 (get_initial_def_for_reduction): Handle option 2 for variable-length
9958 vectors by loading the neutral value into a vector and then shifting
9959 the initial value into element 0.
9960 (get_initial_defs_for_reduction): Replace the code argument with
9961 the neutral value calculated by neutral_op_for_slp_reduction.
9962 Use gimple_build_vector for constant-length vectors.
9963 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9964 but the first group_size elements have a neutral value.
9965 Use duplicate_and_interleave otherwise.
9966 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9967 Update call to get_initial_defs_for_reduction. Handle SLP
9968 reductions for variable-length vectors by creating one vector
9969 result for each scalar result, with the elements associated
9970 with other scalar results stubbed out with the neutral value.
9971 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9972 Require IFN_VEC_SHL_INSERT for double reductions on
9973 variable-length vectors, or SLP reductions that have
9974 a neutral value. Require can_duplicate_and_interleave_p
9975 support for variable-length unchained SLP reductions if there
9976 is no neutral value, such as for MIN/MAX reductions. Also require
9977 the number of vector elements to be a multiple of the number of
9978 SLP statements when doing variable-length unchained SLP reductions.
9979 Update call to vect_create_epilog_for_reduction.
9980 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9981 and remove initial values.
9982 (duplicate_and_interleave): Make public.
9983 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9984 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9986 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9987 Alan Hayward <alan.hayward@arm.com>
9988 David Sherwood <david.sherwood@arm.com>
9990 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9991 (can_duplicate_and_interleave_p): New function.
9992 (vect_get_and_check_slp_defs): Take the vector of statements
9993 rather than just the current one. Remove excess parentheses.
9994 Restriction rejectinon of vect_constant_def and vect_external_def
9995 for variable-length vectors to boolean types, or types for which
9996 can_duplicate_and_interleave_p is false.
9997 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9998 (duplicate_and_interleave): New function.
9999 (vect_get_constant_vectors): Use gimple_build_vector for
10000 constant-length vectors and suitable variable-length constant
10001 vectors. Use duplicate_and_interleave for other variable-length
10002 vectors. Don't defer the update when inserting new statements.
10004 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10005 Alan Hayward <alan.hayward@arm.com>
10006 David Sherwood <david.sherwood@arm.com>
10008 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10009 min_profitable_iters doesn't go negative.
10011 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10012 Alan Hayward <alan.hayward@arm.com>
10013 David Sherwood <david.sherwood@arm.com>
10015 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10016 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10017 * optabs.def (vec_mask_load_lanes_optab): New optab.
10018 (vec_mask_store_lanes_optab): Likewise.
10019 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10020 (MASK_STORE_LANES): Likewise.
10021 * internal-fn.c (mask_load_lanes_direct): New macro.
10022 (mask_store_lanes_direct): Likewise.
10023 (expand_mask_load_optab_fn): Handle masked operations.
10024 (expand_mask_load_lanes_optab_fn): New macro.
10025 (expand_mask_store_optab_fn): Handle masked operations.
10026 (expand_mask_store_lanes_optab_fn): New macro.
10027 (direct_mask_load_lanes_optab_supported_p): Likewise.
10028 (direct_mask_store_lanes_optab_supported_p): Likewise.
10029 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10031 (vect_load_lanes_supported): Likewise.
10032 * tree-vect-data-refs.c (strip_conversion): New function.
10033 (can_group_stmts_p): Likewise.
10034 (vect_analyze_data_ref_accesses): Use it instead of checking
10035 for a pair of assignments.
10036 (vect_store_lanes_supported): Take a masked_p parameter.
10037 (vect_load_lanes_supported): Likewise.
10038 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10039 vect_store_lanes_supported and vect_load_lanes_supported.
10040 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10041 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10042 parameter. Don't allow gaps for masked accesses.
10043 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10044 and vect_load_lanes_supported.
10045 (get_load_store_type): Take a masked_p parameter and update
10046 call to get_group_load_store_type.
10047 (vectorizable_store): Update call to get_load_store_type.
10048 Handle IFN_MASK_STORE_LANES.
10049 (vectorizable_load): Update call to get_load_store_type.
10050 Handle IFN_MASK_LOAD_LANES.
10052 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10053 Alan Hayward <alan.hayward@arm.com>
10054 David Sherwood <david.sherwood@arm.com>
10056 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10058 * config/aarch64/aarch64-protos.h
10059 (aarch64_sve_struct_memory_operand_p): Declare.
10060 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10061 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10062 (VPRED, vpred): Handle SVE structure modes.
10063 * config/aarch64/constraints.md (Utx): New constraint.
10064 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10065 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10066 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10067 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10068 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10069 structure modes. Split into pieces after RA.
10070 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10071 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10073 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10074 SVE structure modes.
10075 (aarch64_classify_address): Likewise.
10076 (sizetochar): Move earlier in file.
10077 (aarch64_print_operand): Handle SVE register lists.
10078 (aarch64_array_mode): New function.
10079 (aarch64_sve_struct_memory_operand_p): Likewise.
10080 (TARGET_ARRAY_MODE): Redefine.
10082 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10083 Alan Hayward <alan.hayward@arm.com>
10084 David Sherwood <david.sherwood@arm.com>
10086 * target.def (array_mode): New target hook.
10087 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10088 * doc/tm.texi: Regenerate.
10089 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10090 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10091 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10092 targetm.array_mode.
10093 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10096 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10097 Alan Hayward <alan.hayward@arm.com>
10098 David Sherwood <david.sherwood@arm.com>
10100 * fold-const.c (fold_binary_loc): Check the argument types
10101 rather than the result type when testing for a vector operation.
10103 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10105 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10106 * doc/tm.texi: Regenerate.
10108 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10109 Alan Hayward <alan.hayward@arm.com>
10110 David Sherwood <david.sherwood@arm.com>
10112 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10113 (sve): Document new AArch64 extension.
10114 * doc/md.texi (w): Extend the description of the AArch64
10115 constraint to include SVE vectors.
10116 (Upl, Upa): Document new AArch64 predicate constraints.
10117 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10119 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10120 (msve-vector-bits=): New option.
10121 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10122 SVE when these are disabled.
10123 (sve): New extension.
10124 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10125 modes. Adjust their number of units based on aarch64_sve_vg.
10126 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10127 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10128 aarch64_addr_query_type.
10129 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10130 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10131 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10132 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10133 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10134 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10135 (aarch64_simd_imm_zero_p): Delete.
10136 (aarch64_check_zero_based_sve_index_immediate): Declare.
10137 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10138 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10139 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10140 (aarch64_sve_float_mul_immediate_p): Likewise.
10141 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10142 rather than an rtx.
10143 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10144 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10145 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10146 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10147 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10148 (aarch64_regmode_natural_size): Likewise.
10149 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10150 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10152 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10153 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10154 for VG and the SVE predicate registers.
10155 (V_ALIASES): Add a "z"-prefixed alias.
10156 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10157 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10158 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10159 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10160 (REG_CLASS_NAMES): Add entries for them.
10161 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10162 and the predicate registers.
10163 (aarch64_sve_vg): Declare.
10164 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10165 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10166 (REGMODE_NATURAL_SIZE): Define.
10167 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10169 * config/aarch64/aarch64.c: Include cfgrtl.h.
10170 (simd_immediate_info): Add a constructor for series vectors,
10171 and an associated step field.
10172 (aarch64_sve_vg): New variable.
10173 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10174 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10175 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10176 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10177 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10178 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10179 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10180 (aarch64_get_mask_mode): New functions.
10181 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10182 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10183 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10184 predicate modes and predicate registers. Explicitly restrict
10185 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10186 to store a vector mode if it is recognized by
10187 aarch64_classify_vector_mode.
10188 (aarch64_regmode_natural_size): New function.
10189 (aarch64_hard_regno_caller_save_mode): Return the original mode
10191 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10192 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10193 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10194 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10196 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10197 does not overlap dest if the function is frame-related. Handle
10199 (aarch64_split_add_offset): New function.
10200 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10201 them aarch64_add_offset.
10202 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10203 and update call to aarch64_sub_sp.
10204 (aarch64_add_cfa_expression): New function.
10205 (aarch64_expand_prologue): Pass extra temporary registers to the
10206 functions above. Handle the case in which we need to emit new
10207 DW_CFA_expressions for registers that were originally saved
10208 relative to the stack pointer, but now have to be expressed
10209 relative to the frame pointer.
10210 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10212 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10213 IP0 and IP1 values for SVE frames.
10214 (aarch64_expand_vec_series): New function.
10215 (aarch64_expand_sve_widened_duplicate): Likewise.
10216 (aarch64_expand_sve_const_vector): Likewise.
10217 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10218 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10219 into the register, rather than emitting a SET directly.
10220 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10221 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10222 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10223 (offset_9bit_signed_scaled_p): New functions.
10224 (aarch64_replicate_bitmask_imm): New function.
10225 (aarch64_bitmask_imm): Use it.
10226 (aarch64_cannot_force_const_mem): Reject expressions involving
10227 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10228 (aarch64_classify_index): Handle SVE indices, by requiring
10229 a plain register index with a scale that matches the element size.
10230 (aarch64_classify_address): Handle SVE addresses. Assert that
10231 the mode of the address is VOIDmode or an integer mode.
10232 Update call to aarch64_classify_symbol.
10233 (aarch64_classify_symbolic_expression): Update call to
10234 aarch64_classify_symbol.
10235 (aarch64_const_vec_all_in_range_p): New function.
10236 (aarch64_print_vector_float_operand): Likewise.
10237 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10238 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10239 and the FP immediates 1.0 and 0.5.
10240 (aarch64_print_address_internal): Handle SVE addresses.
10241 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10242 (aarch64_regno_regclass): Handle predicate registers.
10243 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10245 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10246 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10247 (aarch64_convert_sve_vector_bits): New function.
10248 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10249 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10250 rather than an rtx.
10251 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10252 Handle SVE vector and predicate modes. Accept VL-based constants
10253 that need only one temporary register, and VL offsets that require
10254 no temporary registers.
10255 (aarch64_conditional_register_usage): Mark the predicate registers
10256 as fixed if SVE isn't available.
10257 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10258 Return true for SVE vector and predicate modes.
10259 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10260 rather than an unsigned int. Handle SVE modes.
10261 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10263 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10265 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10266 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10267 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10268 (aarch64_sve_float_mul_immediate_p): New functions.
10269 (aarch64_sve_valid_immediate): New function.
10270 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10271 Explicitly reject structure modes. Check for INDEX constants.
10272 Handle PTRUE and PFALSE constants.
10273 (aarch64_check_zero_based_sve_index_immediate): New function.
10274 (aarch64_simd_imm_zero_p): Delete.
10275 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10276 vector modes. Accept constants in the range of CNT[BHWD].
10277 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10278 ask for an Advanced SIMD mode.
10279 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10280 (aarch64_simd_vector_alignment): Handle SVE predicates.
10281 (aarch64_vectorize_preferred_vector_alignment): New function.
10282 (aarch64_simd_vector_alignment_reachable): Use it instead of
10284 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10285 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10287 (MAX_VECT_LEN): Delete.
10288 (expand_vec_perm_d): Add a vec_flags field.
10289 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10290 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10291 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10293 (aarch64_evpc_rev): Rename to...
10294 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10295 (aarch64_evpc_rev_global): New function.
10296 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10297 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10299 (aarch64_evpc_sve_tbl): New function.
10300 (aarch64_expand_vec_perm_const_1): Update after rename of
10301 aarch64_evpc_rev. Handle SVE permutes too, trying
10302 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10303 than aarch64_evpc_tbl.
10304 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10305 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10306 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10307 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10308 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10309 (aarch64_expand_sve_vcond): New functions.
10310 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10311 of aarch64_vector_mode_p.
10312 (aarch64_dwarf_poly_indeterminate_value): New function.
10313 (aarch64_compute_pressure_classes): Likewise.
10314 (aarch64_can_change_mode_class): Likewise.
10315 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10316 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10317 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10318 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10319 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10320 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10321 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10322 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10324 (Dn, Dl, Dr): Accept const as well as const_vector.
10325 (Dz): Likewise. Compare against CONST0_RTX.
10326 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10327 of "vector" where appropriate.
10328 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10329 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10330 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10331 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10332 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10333 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10334 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10335 (v_int_equiv): Extend to SVE modes.
10336 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10338 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10339 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10340 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10341 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10342 (SVE_COND_FP_CMP): New int iterators.
10343 (perm_hilo): Handle the new unpack unspecs.
10344 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10346 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10347 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10348 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10349 (aarch64_equality_operator, aarch64_constant_vector_operand)
10350 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10351 (aarch64_sve_nonimmediate_operand): Likewise.
10352 (aarch64_sve_general_operand): Likewise.
10353 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10354 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10355 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10356 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10357 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10358 (aarch64_sve_float_arith_immediate): Likewise.
10359 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10360 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10361 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10362 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10363 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10364 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10365 (aarch64_sve_float_arith_operand): Likewise.
10366 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10367 (aarch64_sve_float_mul_operand): Likewise.
10368 (aarch64_sve_vec_perm_operand): Likewise.
10369 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10370 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10371 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10372 as well as const_vector.
10373 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10374 in file. Use CONST0_RTX and CONSTM1_RTX.
10375 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10376 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10377 Use aarch64_simd_imm_zero.
10378 * config/aarch64/aarch64-sve.md: New file.
10379 * config/aarch64/aarch64.md: Include it.
10380 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10381 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10382 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10383 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10384 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10385 (sve): New attribute.
10386 (enabled): Disable instructions with the sve attribute unless
10388 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10389 aarch64_expand_mov_immediate.
10390 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10391 CNT[BHSD] immediates.
10392 (movti): Split CONST_POLY_INT moves into two halves.
10393 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10394 Split additions that need a temporary here if the destination
10395 is the stack pointer.
10396 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10397 (*add<mode>3_poly_1): New instruction.
10398 (set_clobber_cc): New expander.
10400 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10402 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10403 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10404 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10405 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10406 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10407 Change innermode from fixed_mode_size to machine_mode.
10408 (simplify_subreg): Update call accordingly. Handle a constant-sized
10409 subreg of a variable-length CONST_VECTOR.
10411 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10412 Alan Hayward <alan.hayward@arm.com>
10413 David Sherwood <david.sherwood@arm.com>
10415 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10416 (add_offset_to_base): New function, split out from...
10417 (create_mem_ref): ...here. When handling a scale other than 1,
10418 check first whether the address is valid without the offset.
10419 Add it into the base if so, leaving the index and scale as-is.
10421 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10424 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10425 fold_for_warn before checking if arg2 is INTEGER_CST.
10427 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10429 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10430 (store_multiple_operation): Delete.
10431 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10432 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10433 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10434 guarded by TARGET_STRING.
10435 (rs6000_output_load_multiple): Delete.
10436 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10437 OPTION_MASK_STRING / TARGET_STRING handling.
10438 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10439 (const rs6000_opt_masks) <"string">: Change mask to 0.
10440 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10441 (MASK_STRING): Delete.
10442 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10444 (load_multiple): Delete.
10451 (store_multiple): Delete.
10458 (movmemsi_8reg): Delete.
10459 (corresponding unnamed define_insn): Delete.
10460 (movmemsi_6reg): Delete.
10461 (corresponding unnamed define_insn): Delete.
10462 (movmemsi_4reg): Delete.
10463 (corresponding unnamed define_insn): Delete.
10464 (movmemsi_2reg): Delete.
10465 (corresponding unnamed define_insn): Delete.
10466 (movmemsi_1reg): Delete.
10467 (corresponding unnamed define_insn): Delete.
10468 * config/rs6000/rs6000.opt (mno-string): New.
10469 (mstring): Replace by deprecation warning stub.
10470 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10472 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10474 * regrename.c (regrename_do_replace): If replacing the same
10475 reg multiple times, try to reuse last created gen_raw_REG.
10478 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10479 main to workaround a bug in GDB.
10481 2018-01-12 Tom de Vries <tom@codesourcery.com>
10484 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10486 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10488 PR rtl-optimization/80481
10489 * ira-color.c (get_cap_member): New function.
10490 (allocnos_conflict_by_live_ranges_p): Use it.
10491 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10492 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10494 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10497 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10498 (*saddl_se_1): Ditto.
10499 (*ssubsi_1): Ditto.
10500 (*ssubl_se_1): Ditto.
10502 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10504 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10505 rather than wi::to_widest for DR_INITs.
10506 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10507 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10508 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10510 (vect_analyze_group_access_1): Note that here.
10512 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10514 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10515 polynomial type sizes.
10517 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10519 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10520 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10521 (gimple_add_tmp_var): Likewise.
10523 2018-01-12 Martin Liska <mliska@suse.cz>
10525 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10526 (gimple_alloc_sizes): Likewise.
10527 (dump_gimple_statistics): Use PRIu64 in printf format.
10528 * gimple.h: Change uint64_t to int.
10530 2018-01-12 Martin Liska <mliska@suse.cz>
10532 * tree-core.h: Use uint64_t instead of int.
10533 * tree.c (tree_node_counts): Likewise.
10534 (tree_node_sizes): Likewise.
10535 (dump_tree_statistics): Use PRIu64 in printf format.
10537 2018-01-12 Martin Liska <mliska@suse.cz>
10539 * Makefile.in: As qsort_chk is implemented in vec.c, add
10540 vec.o to linkage of gencfn-macros.
10541 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10542 passing the info to record_node_allocation_statistics.
10543 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10545 * ggc-common.c (struct ggc_usage): Add operator== and use
10546 it in operator< and compare function.
10547 * mem-stats.h (struct mem_usage): Likewise.
10548 * vec.c (struct vec_usage): Remove operator< and compare
10549 function. Can be simply inherited.
10551 2018-01-12 Martin Jambor <mjambor@suse.cz>
10554 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10555 * tree-ssa-math-opts.c: Include domwalk.h.
10556 (convert_mult_to_fma_1): New function.
10557 (fma_transformation_info): New type.
10558 (fma_deferring_state): Likewise.
10559 (cancel_fma_deferring): New function.
10560 (result_of_phi): Likewise.
10561 (last_fma_candidate_feeds_initial_phi): Likewise.
10562 (convert_mult_to_fma): Added deferring logic, split actual
10563 transformation to convert_mult_to_fma_1.
10564 (math_opts_dom_walker): New type.
10565 (math_opts_dom_walker::after_dom_children): New method, body moved
10566 here from pass_optimize_widening_mul::execute, added deferring logic
10568 (pass_optimize_widening_mul::execute): Moved most of code to
10569 math_opts_dom_walker::after_dom_children.
10570 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10571 * config/i386/i386.c (ix86_option_override_internal): Added
10572 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10574 2018-01-12 Richard Biener <rguenther@suse.de>
10577 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10578 inline instance vars.
10580 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10583 * config/rx/rx.c (rx_is_restricted_memory_address):
10584 Handle SUBREG case.
10586 2018-01-12 Richard Biener <rguenther@suse.de>
10588 PR tree-optimization/80846
10589 * target.def (split_reduction): New target hook.
10590 * targhooks.c (default_split_reduction): New function.
10591 * targhooks.h (default_split_reduction): Declare.
10592 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10593 target requests first reduce vectors by combining low and high
10595 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10596 (get_vectype_for_scalar_type_and_size): Export.
10597 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10598 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10599 * doc/tm.texi: Regenerate.
10600 * config/i386/i386.c (ix86_split_reduction): Implement
10601 TARGET_VECTORIZE_SPLIT_REDUCTION.
10603 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10606 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10607 in PIC mode except for TARGET_VXWORKS_RTP.
10608 * config/sparc/sparc.c: Include cfgrtl.h.
10609 (TARGET_INIT_PIC_REG): Define.
10610 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10611 (sparc_pic_register_p): New predicate.
10612 (sparc_legitimate_address_p): Use it.
10613 (sparc_legitimize_pic_address): Likewise.
10614 (sparc_delegitimize_address): Likewise.
10615 (sparc_mode_dependent_address_p): Likewise.
10616 (gen_load_pcrel_sym): Remove 4th parameter.
10617 (load_got_register): Adjust call to above. Remove obsolete stuff.
10618 (sparc_expand_prologue): Do not call load_got_register here.
10619 (sparc_flat_expand_prologue): Likewise.
10620 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10621 (sparc_use_pseudo_pic_reg): New function.
10622 (sparc_init_pic_reg): Likewise.
10623 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10624 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10626 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10628 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10629 Add item for branch_cost.
10631 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10633 PR rtl-optimization/83565
10634 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10635 not extend the result to a larger mode for rotate operations.
10636 (num_sign_bit_copies1): Likewise.
10638 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10641 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10643 Use values-Xc.o for -pedantic.
10644 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10646 2018-01-12 Martin Liska <mliska@suse.cz>
10649 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10651 (possible_polymorphic_call_targets): Use it.
10652 (ipa_devirt): Likewise.
10654 2018-01-12 Martin Liska <mliska@suse.cz>
10656 * profile-count.h (enum profile_quality): Use 0 as invalid
10657 enum value of profile_quality.
10659 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10661 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10662 -mext-string options.
10664 2018-01-12 Richard Biener <rguenther@suse.de>
10666 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10667 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10668 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10670 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10672 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10674 * configure.ac (--with-long-double-format): Add support for the
10675 configuration option to change the default long double format on
10677 * config.gcc (powerpc*-linux*-*): Likewise.
10678 * configure: Regenerate.
10679 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10680 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10681 used without modification.
10683 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10685 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10686 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10687 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10688 MISC_BUILTIN_SPEC_BARRIER.
10689 (rs6000_init_builtins): Likewise.
10690 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10692 (speculation_barrier): New define_insn.
10693 * doc/extend.texi: Document __builtin_speculation_barrier.
10695 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10698 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10699 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10700 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10702 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10703 integral modes instead of "ss" and "sd".
10704 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10705 vectors with 32-bit and 64-bit elements.
10706 (vecdupssescalarmodesuffix): New mode attribute.
10707 (vec_dup<mode>): Use it.
10709 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10712 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10713 frame if argument is passed on stack.
10715 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10718 * ree.c (combine_reaching_defs): Optimize also
10719 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10720 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10722 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10724 PR middle-end/83189
10725 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10727 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10729 PR middle-end/83718
10730 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10731 after they are computed.
10733 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10735 PR tree-optimization/83695
10736 * gimple-loop-linterchange.cc
10737 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10738 reset cached scev information after interchange.
10739 (pass_linterchange::execute): Remove call to scev_reset_htab.
10741 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10743 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10744 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10745 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10746 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10747 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10748 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10749 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10750 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10751 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10752 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10753 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10754 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10755 (V_lane_reg): Likewise.
10756 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10758 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10759 (vfmal_lane_low<mode>_intrinsic,
10760 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10761 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10762 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10763 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10764 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10765 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10767 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10769 * config/arm/arm-cpus.in (fp16fml): New feature.
10770 (ALL_SIMD): Add fp16fml.
10771 (armv8.2-a): Add fp16fml as an option.
10772 (armv8.3-a): Likewise.
10773 (armv8.4-a): Add fp16fml as part of fp16.
10774 * config/arm/arm.h (TARGET_FP16FML): Define.
10775 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10777 * config/arm/arm-modes.def (V2HF): Define.
10778 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10779 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10780 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10781 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10782 vfmsl_low, vfmsl_high): New set of builtins.
10783 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10784 (vfml_op): New code attribute.
10785 (VFMLHALVES): New int iterator.
10786 (VFML, VFMLSEL): New mode attributes.
10787 (V_reg): Define mapping for V2HF.
10788 (V_hi, V_lo): New mode attributes.
10789 (VF_constraint): Likewise.
10790 (vfml_half, vfml_half_selector): New int attributes.
10791 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10793 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10794 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10796 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10797 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10798 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10799 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10801 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10802 Document new effective target and option set.
10804 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10806 * config/arm/arm-cpus.in (armv8_4): New feature.
10807 (ARMv8_4a): New fgroup.
10808 (armv8.4-a): New arch.
10809 * config/arm/arm-tables.opt: Regenerate.
10810 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10811 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10812 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10813 Add matching rules for -march=armv8.4-a and extensions.
10814 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10816 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10819 * config/rx/rx.md (BW): New mode attribute.
10820 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10822 2018-01-11 Richard Biener <rguenther@suse.de>
10824 PR tree-optimization/83435
10825 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10826 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10827 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10829 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10830 Alan Hayward <alan.hayward@arm.com>
10831 David Sherwood <david.sherwood@arm.com>
10833 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10835 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10836 (aarch64_print_address_internal): Use it to check for a zero offset.
10838 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10839 Alan Hayward <alan.hayward@arm.com>
10840 David Sherwood <david.sherwood@arm.com>
10842 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10843 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10844 Return a poly_int64 rather than a HOST_WIDE_INT.
10845 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10846 rather than a HOST_WIDE_INT.
10847 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10848 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10849 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10850 final_offset from HOST_WIDE_INT to poly_int64.
10851 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10852 to_constant when getting the number of units in an Advanced SIMD
10854 (aarch64_builtin_vectorized_function): Check for a constant number
10856 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10858 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10859 attribute instead of GET_MODE_NUNITS.
10860 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10861 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10862 GET_MODE_SIZE for fixed-size registers.
10863 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10864 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10865 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10866 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10867 (aarch64_print_operand, aarch64_print_address_internal)
10868 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10869 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10870 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10871 Handle polynomial GET_MODE_SIZE.
10872 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10873 wider than SImode without modification.
10874 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10875 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10876 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10877 passing and returning SVE modes.
10878 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10879 rather than GEN_INT.
10880 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10881 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10882 (aarch64_allocate_and_probe_stack_space): Likewise.
10883 (aarch64_layout_frame): Cope with polynomial offsets.
10884 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10885 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10886 polynomial offsets.
10887 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10888 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10889 poly_int64 rather than a HOST_WIDE_INT.
10890 (aarch64_get_separate_components, aarch64_process_components)
10891 (aarch64_expand_prologue, aarch64_expand_epilogue)
10892 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10893 (aarch64_anchor_offset): New function, split out from...
10894 (aarch64_legitimize_address): ...here.
10895 (aarch64_builtin_vectorization_cost): Handle polynomial
10896 TYPE_VECTOR_SUBPARTS.
10897 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10899 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10900 number of elements from the PARALLEL rather than the mode.
10901 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10902 rather than GET_MODE_BITSIZE.
10903 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10904 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10905 (aarch64_expand_vec_perm_const_1): Handle polynomial
10906 d->perm.length () and d->perm elements.
10907 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10908 Apply to_constant to d->perm elements.
10909 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10910 polynomial CONST_VECTOR_NUNITS.
10911 (aarch64_move_pointer): Take amount as a poly_int64 rather
10913 (aarch64_progress_pointer): Avoid temporary variable.
10914 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10915 the mode attribute instead of GET_MODE.
10917 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10918 Alan Hayward <alan.hayward@arm.com>
10919 David Sherwood <david.sherwood@arm.com>
10921 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10922 x exists before using it.
10923 (aarch64_add_constant_internal): Rename to...
10924 (aarch64_add_offset_1): ...this. Replace regnum with separate
10925 src and dest rtxes. Handle the case in which they're different,
10926 including when the offset is zero. Replace scratchreg with an rtx.
10927 Use 2 additions if there is no spare register into which we can
10928 move a 16-bit constant.
10929 (aarch64_add_constant): Delete.
10930 (aarch64_add_offset): Replace reg with separate src and dest
10931 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10932 Use aarch64_add_offset_1.
10933 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10934 an rtx rather than an int. Take the delta as a poly_int64
10935 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10936 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10937 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10938 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10939 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10940 and aarch64_add_sp.
10941 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10942 aarch64_add_constant.
10944 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10946 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10947 Use scalar_float_mode.
10949 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10951 * config/aarch64/aarch64-simd.md
10952 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10953 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10954 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10955 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10956 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10957 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10958 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10959 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10960 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10961 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10963 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10966 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10967 targ_options->x_arm_arch_string is non NULL.
10969 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10971 * config/aarch64/aarch64.h
10972 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10974 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10977 * expmed.c (emit_store_flag_force): Swap if const op0
10978 and change VOIDmode to mode of op0.
10980 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10982 PR rtl-optimization/83761
10983 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10984 than bytes to mode_for_size.
10986 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10988 PR middle-end/83189
10989 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10990 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10993 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10995 PR middle-end/83575
10996 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10997 when in layout mode.
10998 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10999 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11002 2018-01-10 Michael Collison <michael.collison@arm.com>
11004 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11005 * config/aarch64/aarch64-option-extension.def: Add
11006 AARCH64_OPT_EXTENSION of 'fp16fml'.
11007 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11008 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11009 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11010 * config/aarch64/constraints.md (Ui7): New constraint.
11011 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11012 (VFMLA_SEL_W): Ditto.
11015 (VFMLA16_LOW): New int iterator.
11016 (VFMLA16_HIGH): Ditto.
11017 (UNSPEC_FMLAL): New unspec.
11018 (UNSPEC_FMLSL): Ditto.
11019 (UNSPEC_FMLAL2): Ditto.
11020 (UNSPEC_FMLSL2): Ditto.
11021 (f16mac): New code attribute.
11022 * config/aarch64/aarch64-simd-builtins.def
11023 (aarch64_fmlal_lowv2sf): Ditto.
11024 (aarch64_fmlsl_lowv2sf): Ditto.
11025 (aarch64_fmlalq_lowv4sf): Ditto.
11026 (aarch64_fmlslq_lowv4sf): Ditto.
11027 (aarch64_fmlal_highv2sf): Ditto.
11028 (aarch64_fmlsl_highv2sf): Ditto.
11029 (aarch64_fmlalq_highv4sf): Ditto.
11030 (aarch64_fmlslq_highv4sf): Ditto.
11031 (aarch64_fmlal_lane_lowv2sf): Ditto.
11032 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11033 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11034 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11035 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11036 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11037 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11038 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11039 (aarch64_fmlal_lane_highv2sf): Ditto.
11040 (aarch64_fmlsl_lane_highv2sf): Ditto.
11041 (aarch64_fmlal_laneq_highv2sf): Ditto.
11042 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11043 (aarch64_fmlalq_lane_highv4sf): Ditto.
11044 (aarch64_fmlsl_lane_highv4sf): Ditto.
11045 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11046 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11047 * config/aarch64/aarch64-simd.md:
11048 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11049 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11050 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11051 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11052 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11053 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11054 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11055 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11056 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11057 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11058 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11059 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11060 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11061 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11062 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11063 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11064 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11065 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11066 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11067 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11068 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11069 (vfmlsl_low_u32): Ditto.
11070 (vfmlalq_low_u32): Ditto.
11071 (vfmlslq_low_u32): Ditto.
11072 (vfmlal_high_u32): Ditto.
11073 (vfmlsl_high_u32): Ditto.
11074 (vfmlalq_high_u32): Ditto.
11075 (vfmlslq_high_u32): Ditto.
11076 (vfmlal_lane_low_u32): Ditto.
11077 (vfmlsl_lane_low_u32): Ditto.
11078 (vfmlal_laneq_low_u32): Ditto.
11079 (vfmlsl_laneq_low_u32): Ditto.
11080 (vfmlalq_lane_low_u32): Ditto.
11081 (vfmlslq_lane_low_u32): Ditto.
11082 (vfmlalq_laneq_low_u32): Ditto.
11083 (vfmlslq_laneq_low_u32): Ditto.
11084 (vfmlal_lane_high_u32): Ditto.
11085 (vfmlsl_lane_high_u32): Ditto.
11086 (vfmlal_laneq_high_u32): Ditto.
11087 (vfmlsl_laneq_high_u32): Ditto.
11088 (vfmlalq_lane_high_u32): Ditto.
11089 (vfmlslq_lane_high_u32): Ditto.
11090 (vfmlalq_laneq_high_u32): Ditto.
11091 (vfmlslq_laneq_high_u32): Ditto.
11092 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11093 (AARCH64_FL_FOR_ARCH8_4): New.
11094 (AARCH64_ISA_F16FML): New ISA flag.
11095 (TARGET_F16FML): New feature flag for fp16fml.
11096 (doc/invoke.texi): Document new fp16fml option.
11098 2018-01-10 Michael Collison <michael.collison@arm.com>
11100 * config/aarch64/aarch64-builtins.c:
11101 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11102 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11103 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11104 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11105 (AARCH64_ISA_SHA3): New ISA flag.
11106 (TARGET_SHA3): New feature flag for sha3.
11107 * config/aarch64/iterators.md (sha512_op): New int attribute.
11108 (CRYPTO_SHA512): New int iterator.
11109 (UNSPEC_SHA512H): New unspec.
11110 (UNSPEC_SHA512H2): Ditto.
11111 (UNSPEC_SHA512SU0): Ditto.
11112 (UNSPEC_SHA512SU1): Ditto.
11113 * config/aarch64/aarch64-simd-builtins.def
11114 (aarch64_crypto_sha512hqv2di): New builtin.
11115 (aarch64_crypto_sha512h2qv2di): Ditto.
11116 (aarch64_crypto_sha512su0qv2di): Ditto.
11117 (aarch64_crypto_sha512su1qv2di): Ditto.
11118 (aarch64_eor3qv8hi): Ditto.
11119 (aarch64_rax1qv2di): Ditto.
11120 (aarch64_xarqv2di): Ditto.
11121 (aarch64_bcaxqv8hi): Ditto.
11122 * config/aarch64/aarch64-simd.md:
11123 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11124 (aarch64_crypto_sha512su0qv2di): Ditto.
11125 (aarch64_crypto_sha512su1qv2di): Ditto.
11126 (aarch64_eor3qv8hi): Ditto.
11127 (aarch64_rax1qv2di): Ditto.
11128 (aarch64_xarqv2di): Ditto.
11129 (aarch64_bcaxqv8hi): Ditto.
11130 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11131 (vsha512h2q_u64): Ditto.
11132 (vsha512su0q_u64): Ditto.
11133 (vsha512su1q_u64): Ditto.
11134 (veor3q_u16): Ditto.
11135 (vrax1q_u64): Ditto.
11136 (vxarq_u64): Ditto.
11137 (vbcaxq_u16): Ditto.
11138 * config/arm/types.md (crypto_sha512): New type attribute.
11139 (crypto_sha3): Ditto.
11140 (doc/invoke.texi): Document new sha3 option.
11142 2018-01-10 Michael Collison <michael.collison@arm.com>
11144 * config/aarch64/aarch64-builtins.c:
11145 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11146 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11147 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11148 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11149 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11150 (AARCH64_ISA_SM4): New ISA flag.
11151 (TARGET_SM4): New feature flag for sm4.
11152 * config/aarch64/aarch64-simd-builtins.def
11153 (aarch64_sm3ss1qv4si): Ditto.
11154 (aarch64_sm3tt1aq4si): Ditto.
11155 (aarch64_sm3tt1bq4si): Ditto.
11156 (aarch64_sm3tt2aq4si): Ditto.
11157 (aarch64_sm3tt2bq4si): Ditto.
11158 (aarch64_sm3partw1qv4si): Ditto.
11159 (aarch64_sm3partw2qv4si): Ditto.
11160 (aarch64_sm4eqv4si): Ditto.
11161 (aarch64_sm4ekeyqv4si): Ditto.
11162 * config/aarch64/aarch64-simd.md:
11163 (aarch64_sm3ss1qv4si): Ditto.
11164 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11165 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11166 (aarch64_sm4eqv4si): Ditto.
11167 (aarch64_sm4ekeyqv4si): Ditto.
11168 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11169 (sm3part_op): Ditto.
11170 (CRYPTO_SM3TT): Ditto.
11171 (CRYPTO_SM3PART): Ditto.
11172 (UNSPEC_SM3SS1): New unspec.
11173 (UNSPEC_SM3TT1A): Ditto.
11174 (UNSPEC_SM3TT1B): Ditto.
11175 (UNSPEC_SM3TT2A): Ditto.
11176 (UNSPEC_SM3TT2B): Ditto.
11177 (UNSPEC_SM3PARTW1): Ditto.
11178 (UNSPEC_SM3PARTW2): Ditto.
11179 (UNSPEC_SM4E): Ditto.
11180 (UNSPEC_SM4EKEY): Ditto.
11181 * config/aarch64/constraints.md (Ui2): New constraint.
11182 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11183 * config/arm/types.md (crypto_sm3): New type attribute.
11184 (crypto_sm4): Ditto.
11185 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11186 (vsm3tt1aq_u32): Ditto.
11187 (vsm3tt1bq_u32): Ditto.
11188 (vsm3tt2aq_u32): Ditto.
11189 (vsm3tt2bq_u32): Ditto.
11190 (vsm3partw1q_u32): Ditto.
11191 (vsm3partw2q_u32): Ditto.
11192 (vsm4eq_u32): Ditto.
11193 (vsm4ekeyq_u32): Ditto.
11194 (doc/invoke.texi): Document new sm4 option.
11196 2018-01-10 Michael Collison <michael.collison@arm.com>
11198 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11199 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11200 (AARCH64_FL_FOR_ARCH8_4): New.
11201 (AARCH64_FL_V8_4): New flag.
11202 (doc/invoke.texi): Document new armv8.4-a option.
11204 2018-01-10 Michael Collison <michael.collison@arm.com>
11206 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11207 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11208 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11209 * config/aarch64/aarch64-option-extension.def: Add
11210 AARCH64_OPT_EXTENSION of 'sha2'.
11211 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11212 (crypto): Disable sha2 and aes if crypto disabled.
11213 (crypto): Enable aes and sha2 if enabled.
11214 (simd): Disable sha2 and aes if simd disabled.
11215 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11217 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11218 (TARGET_SHA2): New feature flag for sha2.
11219 (TARGET_AES): New feature flag for aes.
11220 * config/aarch64/aarch64-simd.md:
11221 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11222 conditional on TARGET_AES.
11223 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11224 (aarch64_crypto_sha1hsi): Make pattern conditional
11226 (aarch64_crypto_sha1hv4si): Ditto.
11227 (aarch64_be_crypto_sha1hv4si): Ditto.
11228 (aarch64_crypto_sha1su1v4si): Ditto.
11229 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11230 (aarch64_crypto_sha1su0v4si): Ditto.
11231 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11232 (aarch64_crypto_sha256su0v4si): Ditto.
11233 (aarch64_crypto_sha256su1v4si): Ditto.
11234 (doc/invoke.texi): Document new aes and sha2 options.
11236 2018-01-10 Martin Sebor <msebor@redhat.com>
11238 PR tree-optimization/83781
11239 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11242 2018-01-11 Martin Sebor <msebor@gmail.com>
11243 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11245 PR tree-optimization/83501
11246 PR tree-optimization/81703
11248 * tree-ssa-strlen.c (get_string_cst): Rename...
11249 (get_string_len): ...to this. Handle global constants.
11250 (handle_char_store): Adjust.
11252 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11253 Jim Wilson <jimw@sifive.com>
11255 * config/riscv/riscv-protos.h (riscv_output_return): New.
11256 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11257 (riscv_attribute_table, riscv_output_return),
11258 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11259 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11260 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11262 (riscv_expand_prologue): Add early return for naked function.
11263 (riscv_expand_epilogue): Likewise.
11264 (riscv_function_ok_for_sibcall): Return false for naked function.
11265 (riscv_set_current_function): New.
11266 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11267 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11268 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11269 * doc/extend.texi (RISC-V Function Attributes): New.
11271 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11273 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11274 check for 128-bit long double before checking TCmode.
11275 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11276 128-bit long doubles before checking TFmode or TCmode.
11277 (FLOAT128_IBM_P): Likewise.
11279 2018-01-10 Martin Sebor <msebor@redhat.com>
11281 PR tree-optimization/83671
11282 * builtins.c (c_strlen): Unconditionally return zero for the empty
11284 Use -Warray-bounds for warnings.
11285 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11286 for non-constant array indices with COMPONENT_REF, arrays of
11287 arrays, and pointers to arrays.
11288 (gimple_fold_builtin_strlen): Determine and set length range for
11289 non-constant character arrays.
11291 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11293 PR middle-end/81897
11294 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11297 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11299 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11301 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11304 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11305 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11306 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11307 indexed_or_indirect_operand predicate.
11308 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11309 (*vsx_le_perm_load_v8hi): Likewise.
11310 (*vsx_le_perm_load_v16qi): Likewise.
11311 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11312 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11313 (*vsx_le_perm_store_v8hi): Likewise.
11314 (*vsx_le_perm_store_v16qi): Likewise.
11315 (eight unnamed splitters): Likewise.
11317 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11319 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11320 * config/rs6000/emmintrin.h: Likewise.
11321 * config/rs6000/mmintrin.h: Likewise.
11322 * config/rs6000/xmmintrin.h: Likewise.
11324 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11327 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11329 * tree.c (tree_nop_conversion): Return true for location wrapper
11331 (maybe_wrap_with_location): New function.
11332 (selftest::check_strip_nops): New function.
11333 (selftest::test_location_wrappers): New function.
11334 (selftest::tree_c_tests): Call it.
11335 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11336 (maybe_wrap_with_location): New decl.
11337 (EXPR_LOCATION_WRAPPER_P): New macro.
11338 (location_wrapper_p): New inline function.
11339 (tree_strip_any_location_wrapper): New inline function.
11341 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11344 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11345 stack_realign_offset for the largest alignment of stack slot
11347 (ix86_find_max_used_stack_alignment): New function.
11348 (ix86_finalize_stack_frame_flags): Use it. Set
11349 max_used_stack_alignment if we don't realign stack.
11350 * config/i386/i386.h (machine_function): Add
11351 max_used_stack_alignment.
11353 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11355 * config/arm/arm.opt (-mbranch-cost): New option.
11356 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11359 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11362 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11363 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11365 2018-01-10 Richard Biener <rguenther@suse.de>
11368 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11369 early out so it also covers the case where we have a non-NULL
11372 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11374 PR tree-optimization/83753
11375 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11376 for non-strided grouped accesses if the number of elements is 1.
11378 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11381 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11382 * i386.h (TARGET_USE_GATHER): Define.
11383 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11385 2018-01-10 Martin Liska <mliska@suse.cz>
11388 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11389 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11391 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11392 CLEANUP_NO_PARTITIONING is not set.
11394 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11396 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11397 for vectors, as a partial revert of r254296.
11398 * rtl.h (const_vec_p): Delete.
11399 (const_vec_duplicate_p): Don't test for vector CONSTs.
11400 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11401 * expmed.c (make_tree): Likewise.
11404 * common.md (E, F): Use CONSTANT_P instead of checking for
11406 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11407 checking for CONST_VECTOR.
11409 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11411 PR middle-end/83575
11412 * predict.c (force_edge_cold): Handle in more sane way edges
11413 with no prediction.
11415 2018-01-09 Carl Love <cel@us.ibm.com>
11417 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11419 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11420 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11421 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11422 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11423 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11424 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11425 * config/rs6000/rs6000-protos.h: Add extern defition for
11426 rs6000_generate_float2_double_code.
11427 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11429 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11430 (float2_v2df): Add define_expand.
11432 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11435 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11436 op_mode in the force_to_mode call.
11438 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11440 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11441 instead of checking each element individually.
11442 (aarch64_evpc_uzp): Likewise.
11443 (aarch64_evpc_zip): Likewise.
11444 (aarch64_evpc_ext): Likewise.
11445 (aarch64_evpc_rev): Likewise.
11446 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11447 instead of checking each element individually. Return true without
11449 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11450 whether all selected elements come from the same input, instead of
11451 checking each element individually. Remove calls to gen_rtx_REG,
11452 start_sequence and end_sequence and instead assert that no rtl is
11455 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11457 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11458 order of HIGH and CONST checks.
11460 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11462 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11463 if the destination isn't an SSA_NAME.
11465 2018-01-09 Richard Biener <rguenther@suse.de>
11467 PR tree-optimization/83668
11468 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11470 (canonicalize_loop_form): ... here, renamed from ...
11471 (canonicalize_loop_closed_ssa_form): ... this and amended to
11472 swap successor edges for loop exit blocks to make us use
11473 the RPO order we need for initial schedule generation.
11475 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11477 PR tree-optimization/64811
11478 * match.pd: When optimizing comparisons with Inf, avoid
11479 introducing or losing exceptions from comparisons with NaN.
11481 2018-01-09 Martin Liska <mliska@suse.cz>
11484 * asan.c (shadow_mem_size): Add gcc_assert.
11486 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11488 Don't save registers in main().
11491 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11492 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11493 * config/avr/avr.c (avr_set_current_function): Don't error if
11494 naked, OS_task or OS_main are specified at the same time.
11495 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11497 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11499 * common/config/avr/avr-common.c (avr_option_optimization_table):
11500 Switch on -mmain-is-OS_task for optimizing compilations.
11502 2018-01-09 Richard Biener <rguenther@suse.de>
11504 PR tree-optimization/83572
11505 * graphite.c: Include cfganal.h.
11506 (graphite_transform_loops): Connect infinite loops to exit
11507 and remove fake edges at the end.
11509 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11511 * ipa-inline.c (edge_badness): Revert accidental checkin.
11513 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11516 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11517 symbols; not inline clones.
11519 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11522 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11523 hard registers. Formatting fixes.
11525 PR preprocessor/83722
11526 * gcc.c (try_generate_repro): Pass
11527 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11528 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11531 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11532 Kito Cheng <kito.cheng@gmail.com>
11534 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11535 (riscv_leaf_function_p): Delete.
11536 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11538 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11540 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11542 (do_ifelse): New function.
11543 (do_isel): New function.
11544 (do_sub3): New function.
11545 (do_add3): New function.
11546 (do_load_mask_compare): New function.
11547 (do_overlap_load_compare): New function.
11548 (expand_compare_loop): New function.
11549 (expand_block_compare): Call expand_compare_loop() when appropriate.
11550 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11551 option description.
11552 (-mblock-compare-inline-loop-limit): New option.
11554 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11557 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11558 Reverse order of second and third operands in first alternative.
11559 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11560 of first and second elements in UNSPEC_VPERMR vector.
11561 (altivec_expand_vec_perm_le): Likewise.
11563 2018-01-08 Jeff Law <law@redhat.com>
11565 PR rtl-optimizatin/81308
11566 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11567 (process_switch): If group_case_labels makes a change, then set
11569 (pass_convert_switch::execute): If a switch is converted, then
11570 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11572 PR rtl-optimization/81308
11573 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11576 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11578 PR target/83663 - Revert r255946
11579 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11580 generation for cases where splatting a value is not useful.
11581 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11582 across a vec_duplicate and a paradoxical subreg forming a vector
11583 mode to a vec_concat.
11585 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11587 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11588 -march=armv8.3-a variants.
11589 * config/arm/t-multilib: Likewise.
11590 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11592 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11594 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11596 (cceq_ior_compare_complement): Give it a name so I can use it, and
11597 change boolean_or_operator predicate to boolean_operator so it can
11598 be used to generate a crand.
11599 (eqne): New code iterator.
11600 (bd/bd_neg): New code_attrs.
11601 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11602 a single define_insn.
11603 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11604 decrement (bdnzt/bdnzf/bdzt/bdzf).
11605 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11606 with the new names of the branch decrement patterns, and added the
11607 names of the branch decrement conditional patterns.
11609 2018-01-08 Richard Biener <rguenther@suse.de>
11611 PR tree-optimization/83563
11612 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11615 2018-01-08 Richard Biener <rguenther@suse.de>
11617 PR middle-end/83713
11618 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11620 2018-01-08 Richard Biener <rguenther@suse.de>
11622 PR tree-optimization/83685
11623 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11624 references to abnormals.
11626 2018-01-08 Richard Biener <rguenther@suse.de>
11629 * dwarf2out.c (output_indirect_strings): Handle empty
11630 skeleton_debug_str_hash.
11631 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11633 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11635 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11636 (emit_store_direct): Likewise.
11637 (arc_trampoline_adjust_address): Likewise.
11638 (arc_asm_trampoline_template): New function.
11639 (arc_initialize_trampoline): Use asm_trampoline_template.
11640 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11641 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11642 * config/arc/arc.md (flush_icache): Delete pattern.
11644 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11646 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11647 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11650 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11653 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11654 by not USED_FOR_TARGET.
11655 (make_pass_resolve_sw_modes): Likewise.
11657 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11659 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11662 2018-01-08 Richard Biener <rguenther@suse.de>
11664 PR middle-end/83580
11665 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11667 2018-01-08 Richard Biener <rguenther@suse.de>
11669 PR middle-end/83517
11670 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11672 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11674 PR middle-end/81897
11675 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11676 basic blocks with a small number of successors.
11677 (convert_control_dep_chain_into_preds): Improve handling of
11679 (dump_predicates): Split apart into...
11680 (dump_pred_chain): ...here...
11681 (dump_pred_info): ...and here.
11682 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11683 (can_chain_union_be_invalidated_p): Improve check for invalidation
11685 (uninit_uses_cannot_happen): Avoid unnecessary if
11686 convert_control_dep_chain_into_preds yielded nothing.
11688 2018-01-06 Martin Sebor <msebor@redhat.com>
11690 PR tree-optimization/83640
11691 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11692 subtracting negative offset from size.
11693 (builtin_access::overlap): Adjust offset bounds of the access to fall
11694 within the size of the object if possible.
11696 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11698 PR rtl-optimization/83699
11699 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11700 extract_bit_field_as_subreg to cases in which the extracted
11701 value is also a vector.
11703 * lra-constraints.c (process_alt_operands): Test for the equivalence
11704 substitutions when detecting a possible reload cycle.
11706 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11709 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11710 by default if flag_selective_schedling{,2}. Formatting fixes.
11712 PR rtl-optimization/83682
11713 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11714 if it has non-VECTOR_MODE element mode.
11715 (vec_duplicate_p): Likewise.
11717 PR middle-end/83694
11718 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11719 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11721 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11724 * config/i386/i386-builtin.def
11725 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11726 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11727 Require also OPTION_MASK_ISA_AVX512F in addition to
11728 OPTION_MASK_ISA_GFNI.
11729 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11730 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11731 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11732 to OPTION_MASK_ISA_GFNI.
11733 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11734 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11735 OPTION_MASK_ISA_AVX512BW.
11736 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11737 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11738 addition to OPTION_MASK_ISA_GFNI.
11739 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11740 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11741 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11742 to OPTION_MASK_ISA_GFNI.
11743 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11744 a requirement for all ISAs rather than any of them with a few
11746 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11748 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11749 bitmasks to be enabled with 3 exceptions, instead of requiring any
11750 enabled ISA with lots of exceptions.
11751 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11752 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11753 Change avx512bw in isa attribute to avx512f.
11754 * config/i386/sgxintrin.h: Add license boilerplate.
11755 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11756 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11757 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11758 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11760 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11761 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11762 temporarily sse2 rather than sse if not enabled already.
11765 * config/i386/sse.md (VI248_VLBW): Rename to ...
11766 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11767 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11768 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11769 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11770 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11771 mode iterator instead of VI248_VLBW.
11773 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11775 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11776 (record_modified): Skip clobbers; add debug output.
11777 (param_change_prob): Use sreal frequencies.
11779 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11781 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11782 punt for user-aligned variables.
11784 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11786 * tree-chrec.c (chrec_contains_symbols): Return true for
11789 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11792 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11793 of (x|y) == x for BICS pattern.
11795 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11797 PR tree-optimization/83605
11798 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11799 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11802 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11804 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11805 * config/epiphany/rtems.h: New file.
11807 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11808 Uros Bizjak <ubizjak@gmail.com>
11811 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11812 QIreg_operand instead of register_operand predicate.
11813 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11814 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11815 comments instead of -fmitigate[-_]rop.
11817 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11820 * cgraphunit.c (symbol_table::compile): Switch to text_section
11821 before calling assembly_start debug hook.
11822 * run-rtl-passes.c (run_rtl_passes): Likewise.
11825 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11827 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11828 range_int_cst_p rather than !symbolic_range_p before calling
11829 extract_range_from_multiplicative_op_1.
11831 2018-01-04 Jeff Law <law@redhat.com>
11833 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11834 redundant test in assertion.
11836 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11838 * doc/rtl.texi: Document machine_mode wrapper classes.
11840 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11842 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11843 using tree_to_uhwi.
11845 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11847 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11848 the VEC_PERM_EXPR fold to fail.
11850 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11853 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11854 to switched_sections.
11856 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11859 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11860 test for d.testing.
11862 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11865 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11866 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11868 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11871 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11872 is BLKmode and bitpos not zero or mode change is needed.
11874 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11877 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11880 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11883 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11884 instead of MULT rtx. Update all corresponding splitters.
11885 (*saddl_se): Ditto.
11886 (*ssub<modesuffix>): Ditto.
11887 (*ssubl_se): Ditto.
11888 (*cmp_sadd_di): Update split patterns.
11889 (*cmp_sadd_si): Ditto.
11890 (*cmp_sadd_sidi): Ditto.
11891 (*cmp_ssub_di): Ditto.
11892 (*cmp_ssub_si): Ditto.
11893 (*cmp_ssub_sidi): Ditto.
11894 * config/alpha/predicates.md (const23_operand): New predicate.
11895 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11896 Look for ASHIFT, not MULT inner operand.
11897 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11899 2018-01-04 Martin Liska <mliska@suse.cz>
11901 PR gcov-profile/83669
11902 * gcov.c (output_intermediate_file): Add version to intermediate
11904 * doc/gcov.texi: Document new field 'version' in intermediate
11905 file format. Fix location of '-k' option of gcov command.
11907 2018-01-04 Martin Liska <mliska@suse.cz>
11910 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11912 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11914 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11916 2018-01-03 Martin Sebor <msebor@redhat.com>
11918 PR tree-optimization/83655
11919 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11920 checking calls with invalid arguments.
11922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11924 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11925 (vectorizable_mask_load_store): Delete.
11926 (vectorizable_call): Return false for masked loads and stores.
11927 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11928 instead of gimple_assign_rhs1.
11929 (vectorizable_load): Handle IFN_MASK_LOAD.
11930 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11932 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11934 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11936 (vectorizable_mask_load_store): ...here.
11937 (vectorizable_load): ...and here.
11939 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11941 * tree-vect-stmts.c (vect_build_all_ones_mask)
11942 (vect_build_zero_merge_argument): New functions, split out from...
11943 (vectorizable_load): ...here.
11945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11947 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11949 (vectorizable_mask_load_store): ...here.
11950 (vectorizable_store): ...and here.
11952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11954 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11956 (vectorizable_mask_load_store): ...here.
11958 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11960 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11961 (vect_model_store_cost): Take a vec_load_store_type instead of a
11963 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11964 (vect_model_store_cost): Take a vec_load_store_type instead of a
11966 (vectorizable_mask_load_store): Update accordingly.
11967 (vectorizable_store): Likewise.
11968 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11972 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11973 IFN_MASK_LOAD calls here rather than...
11974 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11976 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11977 Alan Hayward <alan.hayward@arm.com>
11978 David Sherwood <david.sherwood@arm.com>
11980 * expmed.c (extract_bit_field_1): For vector extracts,
11981 fall back to extract_bit_field_as_subreg if vec_extract
11984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11985 Alan Hayward <alan.hayward@arm.com>
11986 David Sherwood <david.sherwood@arm.com>
11988 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11989 they are variable or constant sized.
11990 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11991 slots for constant-sized data.
11993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11994 Alan Hayward <alan.hayward@arm.com>
11995 David Sherwood <david.sherwood@arm.com>
11997 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11998 handling COND_EXPRs with boolean comparisons, try to find a better
11999 basis for the mask type than the boolean itself.
12001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12003 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12004 is calculated and how it can be overridden.
12005 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12006 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12008 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12012 Alan Hayward <alan.hayward@arm.com>
12013 David Sherwood <david.sherwood@arm.com>
12015 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12016 Remove the mode argument.
12017 (aarch64_simd_valid_immediate): Remove the mode and inverse
12019 * config/aarch64/iterators.md (bitsize): New iterator.
12020 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12021 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12022 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12023 aarch64_simd_valid_immediate.
12024 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12025 (aarch64_reg_or_bic_imm): Likewise.
12026 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12027 with an insn_type enum and msl with a modifier_type enum.
12028 Replace element_width with a scalar_mode. Change the shift
12029 to unsigned int. Add constructors for scalar_float_mode and
12030 scalar_int_mode elements.
12031 (aarch64_vect_float_const_representable_p): Delete.
12032 (aarch64_can_const_movi_rtx_p)
12033 (aarch64_simd_scalar_immediate_valid_for_move)
12034 (aarch64_simd_make_constant): Update call to
12035 aarch64_simd_valid_immediate.
12036 (aarch64_advsimd_valid_immediate_hs): New function.
12037 (aarch64_advsimd_valid_immediate): Likewise.
12038 (aarch64_simd_valid_immediate): Remove mode and inverse
12039 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12040 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12041 and aarch64_float_const_representable_p on the result.
12042 (aarch64_output_simd_mov_immediate): Remove mode argument.
12043 Update call to aarch64_simd_valid_immediate and use of
12044 simd_immediate_info.
12045 (aarch64_output_scalar_simd_mov_immediate): Update call
12048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12049 Alan Hayward <alan.hayward@arm.com>
12050 David Sherwood <david.sherwood@arm.com>
12052 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12053 (mode_nunits): Likewise CONST_MODE_NUNITS.
12054 * machmode.def (ADJUST_NUNITS): Document.
12055 * genmodes.c (mode_data::need_nunits_adj): New field.
12056 (blank_mode): Update accordingly.
12057 (adj_nunits): New variable.
12058 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12060 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12061 listed in adj_nunits.
12062 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12063 listed in adj_nunits. Don't emit case statements for such modes.
12064 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12065 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12066 nothing if adj_nunits is nonnull.
12067 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12068 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12069 (emit_mode_fbit): Update use of print_maybe_const_decl.
12070 (emit_move_size): Likewise. Treat the array as non-const
12072 (emit_mode_adjustments): Handle adj_nunits.
12074 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12076 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12077 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12078 (VECTOR_MODES): Use it.
12079 (make_vector_modes): Take the prefix as an argument.
12081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12082 Alan Hayward <alan.hayward@arm.com>
12083 David Sherwood <david.sherwood@arm.com>
12085 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12086 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12087 for MODE_VECTOR_BOOL.
12088 * machmode.def (VECTOR_BOOL_MODE): Document.
12089 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12090 (make_vector_bool_mode): New function.
12091 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12093 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12094 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12096 * stor-layout.c (int_mode_for_mode): Likewise.
12097 * tree.c (build_vector_type_for_mode): Likewise.
12098 * varasm.c (output_constant_pool_2): Likewise.
12099 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12100 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12101 for MODE_VECTOR_BOOL.
12102 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12103 of mode class checks.
12104 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12105 instead of a list of mode class checks.
12106 (expand_vector_scalar_condition): Likewise.
12107 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12110 Alan Hayward <alan.hayward@arm.com>
12111 David Sherwood <david.sherwood@arm.com>
12113 * machmode.h (mode_size): Change from unsigned short to
12115 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12116 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12117 or if measurement_type is not polynomial.
12118 (fixed_size_mode::includes_p): Check for constant-sized modes.
12119 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12120 return a poly_uint16 rather than an unsigned short.
12121 (emit_mode_size): Change the type of mode_size from unsigned short
12122 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12123 (emit_mode_adjustments): Cope with polynomial vector sizes.
12124 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12126 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12128 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12129 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12130 * caller-save.c (setup_save_areas): Likewise.
12131 (replace_reg_with_saved_mem): Likewise.
12132 * calls.c (emit_library_call_value_1): Likewise.
12133 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12134 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12135 (gen_lowpart_for_combine): Likewise.
12136 * convert.c (convert_to_integer_1): Likewise.
12137 * cse.c (equiv_constant, cse_insn): Likewise.
12138 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12139 (cselib_subst_to_values): Likewise.
12140 * dce.c (word_dce_process_block): Likewise.
12141 * df-problems.c (df_word_lr_mark_ref): Likewise.
12142 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12143 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12144 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12145 (rtl_for_decl_location): Likewise.
12146 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12147 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12148 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12149 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12150 (expand_expr_real_1): Likewise.
12151 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12152 (pad_below): Likewise.
12153 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12154 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12155 * ira.c (get_subreg_tracking_sizes): Likewise.
12156 * ira-build.c (ira_create_allocno_objects): Likewise.
12157 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12158 (ira_sort_regnos_for_alter_reg): Likewise.
12159 * ira-costs.c (record_operand_costs): Likewise.
12160 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12161 (resolve_simple_move): Likewise.
12162 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12163 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12164 (lra_constraints): Likewise.
12165 (CONST_POOL_OK_P): Reject variable-sized modes.
12166 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12167 (add_pseudo_to_slot, lra_spill): Likewise.
12168 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12169 * optabs-query.c (get_best_extraction_insn): Likewise.
12170 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12171 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12172 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12173 * recog.c (offsettable_address_addr_space_p): Likewise.
12174 * regcprop.c (maybe_mode_change): Likewise.
12175 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12176 * regrename.c (build_def_use): Likewise.
12177 * regstat.c (dump_reg_info): Likewise.
12178 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12179 (find_reloads, find_reloads_subreg_address): Likewise.
12180 * reload1.c (eliminate_regs_1): Likewise.
12181 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12182 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12183 (simplify_binary_operation_1, simplify_subreg): Likewise.
12184 * targhooks.c (default_function_arg_padding): Likewise.
12185 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12186 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12187 (verify_gimple_assign_ternary): Likewise.
12188 * tree-inline.c (estimate_move_cost): Likewise.
12189 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12190 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12191 (get_address_cost_ainc): Likewise.
12192 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12193 (vect_supportable_dr_alignment): Likewise.
12194 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12195 (vectorizable_reduction): Likewise.
12196 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12197 (vectorizable_operation, vectorizable_load): Likewise.
12198 * tree.c (build_same_sized_truth_vector_type): Likewise.
12199 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12200 * var-tracking.c (emit_note_insn_var_location): Likewise.
12201 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12202 (ADDR_VEC_ALIGN): Likewise.
12204 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12205 Alan Hayward <alan.hayward@arm.com>
12206 David Sherwood <david.sherwood@arm.com>
12208 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12210 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12211 or if measurement_type is polynomial.
12212 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12213 * combine.c (make_extraction): Likewise.
12214 * dse.c (find_shift_sequence): Likewise.
12215 * dwarf2out.c (mem_loc_descriptor): Likewise.
12216 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12217 (extract_bit_field, extract_low_bits): Likewise.
12218 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12219 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12220 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12221 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12222 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12223 * reload.c (find_reloads): Likewise.
12224 * reload1.c (alter_reg): Likewise.
12225 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12226 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12227 * tree-if-conv.c (predicate_mem_writes): Likewise.
12228 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12229 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12230 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12231 * valtrack.c (dead_debug_insert_temp): Likewise.
12232 * varasm.c (mergeable_constant_section): Likewise.
12233 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12236 Alan Hayward <alan.hayward@arm.com>
12237 David Sherwood <david.sherwood@arm.com>
12239 * expr.c (expand_assignment): Cope with polynomial mode sizes
12240 when assigning to a CONCAT.
12242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12243 Alan Hayward <alan.hayward@arm.com>
12244 David Sherwood <david.sherwood@arm.com>
12246 * machmode.h (mode_precision): Change from unsigned short to
12248 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12250 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12251 or if measurement_type is not polynomial.
12252 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12253 in which the mode is already known to be a scalar_int_mode.
12254 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12255 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12257 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12258 for GET_MODE_PRECISION.
12259 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12260 for GET_MODE_PRECISION.
12261 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12263 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12264 (expand_field_assignment, make_extraction): Likewise.
12265 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12266 (get_last_value): Likewise.
12267 * convert.c (convert_to_integer_1): Likewise.
12268 * cse.c (cse_insn): Likewise.
12269 * expr.c (expand_expr_real_1): Likewise.
12270 * lra-constraints.c (simplify_operand_subreg): Likewise.
12271 * optabs-query.c (can_atomic_load_p): Likewise.
12272 * optabs.c (expand_atomic_load): Likewise.
12273 (expand_atomic_store): Likewise.
12274 * ree.c (combine_reaching_defs): Likewise.
12275 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12276 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12277 * tree.h (type_has_mode_precision_p): Likewise.
12278 * ubsan.c (instrument_si_overflow): Likewise.
12280 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12281 Alan Hayward <alan.hayward@arm.com>
12282 David Sherwood <david.sherwood@arm.com>
12284 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12285 polynomial numbers of units.
12286 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12287 (valid_vector_subparts_p): New function.
12288 (build_vector_type): Remove temporary shim and take the number
12289 of units as a poly_uint64 rather than an int.
12290 (build_opaque_vector_type): Take the number of units as a
12291 poly_uint64 rather than an int.
12292 * tree.c (build_vector_from_ctor): Handle polynomial
12293 TYPE_VECTOR_SUBPARTS.
12294 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12295 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12296 (build_vector_from_val): If the number of units is variable,
12297 use build_vec_duplicate_cst for constant operands and
12298 VEC_DUPLICATE_EXPR otherwise.
12299 (make_vector_type): Remove temporary is_constant ().
12300 (build_vector_type, build_opaque_vector_type): Take the number of
12301 units as a poly_uint64 rather than an int.
12302 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12304 * cfgexpand.c (expand_debug_expr): Likewise.
12305 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12306 (store_constructor, expand_expr_real_1): Likewise.
12307 (const_scalar_mask_from_tree): Likewise.
12308 * fold-const-call.c (fold_const_reduction): Likewise.
12309 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12310 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12311 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12312 (fold_relational_const): Likewise.
12313 (native_interpret_vector): Likewise. Change the size from an
12314 int to an unsigned int.
12315 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12316 TYPE_VECTOR_SUBPARTS.
12317 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12318 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12319 duplicating a non-constant operand into a variable-length vector.
12320 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12321 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12322 * ipa-icf.c (sem_variable::equals): Likewise.
12323 * match.pd: Likewise.
12324 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12325 * print-tree.c (print_node): Likewise.
12326 * stor-layout.c (layout_type): Likewise.
12327 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12328 * tree-cfg.c (verify_gimple_comparison): Likewise.
12329 (verify_gimple_assign_binary): Likewise.
12330 (verify_gimple_assign_ternary): Likewise.
12331 (verify_gimple_assign_single): Likewise.
12332 * tree-pretty-print.c (dump_generic_node): Likewise.
12333 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12334 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12335 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12336 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12337 (vect_shift_permute_load_chain): Likewise.
12338 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12339 (expand_vector_condition, optimize_vector_constructor): Likewise.
12340 (lower_vec_perm, get_compute_type): Likewise.
12341 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12342 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12343 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12344 (vect_recog_mask_conversion_pattern): Likewise.
12345 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12346 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12347 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12348 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12349 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12350 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12351 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12352 (supportable_widening_operation): Likewise.
12353 (supportable_narrowing_operation): Likewise.
12354 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12356 * varasm.c (output_constant): Likewise.
12358 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12359 Alan Hayward <alan.hayward@arm.com>
12360 David Sherwood <david.sherwood@arm.com>
12362 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12363 so that both the length == 3 and length != 3 cases set up their
12364 own permute vectors. Add comments explaining why we know the
12365 number of elements is constant.
12366 (vect_permute_load_chain): Likewise.
12368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12369 Alan Hayward <alan.hayward@arm.com>
12370 David Sherwood <david.sherwood@arm.com>
12372 * machmode.h (mode_nunits): Change from unsigned char to
12374 (ONLY_FIXED_SIZE_MODES): New macro.
12375 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12376 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12377 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12379 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12380 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12381 or if measurement_type is not polynomial.
12382 * genmodes.c (ZERO_COEFFS): New macro.
12383 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12385 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12386 Use ZERO_COEFFS when emitting initializers.
12387 * data-streamer.h (bp_pack_poly_value): New function.
12388 (bp_unpack_poly_value): Likewise.
12389 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12390 for GET_MODE_NUNITS.
12391 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12392 for GET_MODE_NUNITS.
12393 * tree.c (make_vector_type): Remove temporary shim and make
12394 the real function take the number of units as a poly_uint64
12395 rather than an int.
12396 (build_vector_type_for_mode): Handle polynomial nunits.
12397 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12398 * emit-rtl.c (const_vec_series_p_1): Likewise.
12399 (gen_rtx_CONST_VECTOR): Likewise.
12400 * fold-const.c (test_vec_duplicate_folding): Likewise.
12401 * genrecog.c (validate_pattern): Likewise.
12402 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12403 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12404 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12405 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12406 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12407 * rtlanal.c (subreg_get_info): Likewise.
12408 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12409 (vect_grouped_load_supported): Likewise.
12410 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12411 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12412 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12413 (simplify_const_unary_operation, simplify_binary_operation_1)
12414 (simplify_const_binary_operation, simplify_ternary_operation)
12415 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12416 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12417 instead of CONST_VECTOR_NUNITS.
12418 * varasm.c (output_constant_pool_2): Likewise.
12419 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12420 explicit-encoded elements in the XVEC for variable-length vectors.
12422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12424 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12427 Alan Hayward <alan.hayward@arm.com>
12428 David Sherwood <david.sherwood@arm.com>
12430 * coretypes.h (fixed_size_mode): Declare.
12431 (fixed_size_mode_pod): New typedef.
12432 * builtins.h (target_builtins::x_apply_args_mode)
12433 (target_builtins::x_apply_result_mode): Change type to
12434 fixed_size_mode_pod.
12435 * builtins.c (apply_args_size, apply_result_size, result_vector)
12436 (expand_builtin_apply_args_1, expand_builtin_apply)
12437 (expand_builtin_return): Update accordingly.
12439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12441 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12442 * cselib.c (cselib_hash_rtx): Likewise.
12443 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12444 CONST_VECTOR encoding.
12446 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12447 Jeff Law <law@redhat.com>
12450 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12451 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12452 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12453 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12456 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12457 explicitly probe *sp in a noreturn function if there were any callee
12458 register saves or frame pointer is needed.
12460 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12463 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12464 BLKmode for ternary, binary or unary expressions.
12467 * var-tracking.c (delete_vta_debug_insn): New inline function.
12468 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12469 insns from get_insns () to NULL instead of each bb separately.
12470 Use delete_vta_debug_insn. No longer static.
12471 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12472 delete_vta_debug_insns callers.
12473 * rtl.h (delete_vta_debug_insns): Declare.
12474 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12475 instead of variable_tracking_main.
12477 2018-01-03 Martin Sebor <msebor@redhat.com>
12479 PR tree-optimization/83603
12480 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12481 arguments past the endof the argument list in functions declared
12482 without a prototype.
12483 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12484 Avoid checking when arguments are null.
12486 2018-01-03 Martin Sebor <msebor@redhat.com>
12489 * doc/extend.texi (attribute const): Fix a typo.
12490 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12491 issuing -Wsuggest-attribute for void functions.
12493 2018-01-03 Martin Sebor <msebor@redhat.com>
12495 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12496 offset_int::from instead of wide_int::to_shwi.
12497 (maybe_diag_overlap): Remove assertion.
12498 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12499 * gimple-ssa-sprintf.c (format_directive): Same.
12500 (parse_directive): Same.
12501 (sprintf_dom_walker::compute_format_length): Same.
12502 (try_substitute_return_value): Same.
12504 2018-01-03 Jeff Law <law@redhat.com>
12506 PR middle-end/83654
12507 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12508 non-constant residual for zero at runtime and avoid probing in
12509 that case. Reorganize code for trailing problem to mirror handling
12512 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12514 PR tree-optimization/83501
12515 * tree-ssa-strlen.c (get_string_cst): New.
12516 (handle_char_store): Call get_string_cst.
12518 2018-01-03 Martin Liska <mliska@suse.cz>
12520 PR tree-optimization/83593
12521 * tree-ssa-strlen.c: Include tree-cfg.h.
12522 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12523 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12524 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12526 (strlen_dom_walker::before_dom_children): Call
12527 gimple_purge_dead_eh_edges. Dump tranformation with details
12529 (strlen_dom_walker::before_dom_children): Update call by adding
12530 new argument cleanup_eh.
12531 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12533 2018-01-03 Martin Liska <mliska@suse.cz>
12536 * cif-code.def (VARIADIC_THUNK): New enum value.
12537 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12540 2018-01-03 Jan Beulich <jbeulich@suse.com>
12542 * sse.md (mov<mode>_internal): Tighten condition for when to use
12543 vmovdqu<ssescalarsize> for TI and OI modes.
12545 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12547 Update copyright years.
12549 2018-01-03 Martin Liska <mliska@suse.cz>
12552 * ipa-visibility.c (function_and_variable_visibility): Skip
12553 functions with noipa attribure.
12555 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12557 * gcc.c (process_command): Update copyright notice dates.
12558 * gcov-dump.c (print_version): Ditto.
12559 * gcov.c (print_version): Ditto.
12560 * gcov-tool.c (print_version): Ditto.
12561 * gengtype.c (create_file): Ditto.
12562 * doc/cpp.texi: Bump @copying's copyright year.
12563 * doc/cppinternals.texi: Ditto.
12564 * doc/gcc.texi: Ditto.
12565 * doc/gccint.texi: Ditto.
12566 * doc/gcov.texi: Ditto.
12567 * doc/install.texi: Ditto.
12568 * doc/invoke.texi: Ditto.
12570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12572 * vector-builder.h (vector_builder::m_full_nelts): Change from
12573 unsigned int to poly_uint64.
12574 (vector_builder::full_nelts): Update prototype accordingly.
12575 (vector_builder::new_vector): Likewise.
12576 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12577 (vector_builder::operator ==): Likewise.
12578 (vector_builder::finalize): Likewise.
12579 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12580 Take the number of elements as a poly_uint64 rather than an
12582 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12583 from unsigned int to poly_uint64.
12584 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12585 (vec_perm_indices::new_vector): Likewise.
12586 (vec_perm_indices::length): Likewise.
12587 (vec_perm_indices::nelts_per_input): Likewise.
12588 (vec_perm_indices::input_nelts): Likewise.
12589 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12590 number of elements per input as a poly_uint64 rather than an
12591 unsigned int. Use the original encoding for variable-length
12592 vectors, rather than clamping each individual element.
12593 For the second and subsequent elements in each pattern,
12594 clamp the step and base before clamping their sum.
12595 (vec_perm_indices::series_p): Handle polynomial element counts.
12596 (vec_perm_indices::all_in_range_p): Likewise.
12597 (vec_perm_indices_to_tree): Likewise.
12598 (vec_perm_indices_to_rtx): Likewise.
12599 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12600 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12601 (tree_vector_builder::new_binary_operation): Handle polynomial
12602 element counts. Return false if we need to know the number
12603 of elements at compile time.
12604 * fold-const.c (fold_vec_perm): Punt if the number of elements
12605 isn't known at compile time.
12607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12609 * vec-perm-indices.h (vec_perm_builder): Change element type
12610 from HOST_WIDE_INT to poly_int64.
12611 (vec_perm_indices::element_type): Update accordingly.
12612 (vec_perm_indices::clamp): Handle polynomial element_types.
12613 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12614 (vec_perm_indices::all_in_range_p): Likewise.
12615 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12617 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12618 polynomial vec_perm_indices element types.
12619 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12620 * fold-const.c (fold_vec_perm): Likewise.
12621 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12622 * tree-vect-generic.c (lower_vec_perm): Likewise.
12623 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12624 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12625 element type to HOST_WIDE_INT.
12627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12628 Alan Hayward <alan.hayward@arm.com>
12629 David Sherwood <david.sherwood@arm.com>
12631 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12632 rather than an int. Use plus_constant.
12633 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12634 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12637 Alan Hayward <alan.hayward@arm.com>
12638 David Sherwood <david.sherwood@arm.com>
12640 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12641 a HOST_WIDE_INT to a poly_int64.
12643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12644 Alan Hayward <alan.hayward@arm.com>
12645 David Sherwood <david.sherwood@arm.com>
12647 * calls.c (load_register_parameters): Cope with polynomial
12648 mode sizes. Require a constant size for BLKmode parameters
12649 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12650 forces a parameter to be padded at the lsb end in order to
12651 fill a complete number of words, require the parameter size
12652 to be ordered wrt UNITS_PER_WORD.
12654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12655 Alan Hayward <alan.hayward@arm.com>
12656 David Sherwood <david.sherwood@arm.com>
12658 * reload1.c (spill_stack_slot_width): Change element type
12659 from unsigned int to poly_uint64_pod.
12660 (alter_reg): Treat mode sizes as polynomial.
12662 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12663 Alan Hayward <alan.hayward@arm.com>
12664 David Sherwood <david.sherwood@arm.com>
12666 * reload.c (complex_word_subreg_p): New function.
12667 (reload_inner_reg_of_subreg, push_reload): Use it.
12669 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12670 Alan Hayward <alan.hayward@arm.com>
12671 David Sherwood <david.sherwood@arm.com>
12673 * lra-constraints.c (process_alt_operands): Reject matched
12674 operands whose sizes aren't ordered.
12675 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12682 that the mode size is in the set {1, 2, 4, 8, 16}.
12684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12685 Alan Hayward <alan.hayward@arm.com>
12686 David Sherwood <david.sherwood@arm.com>
12688 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12689 Use plus_constant instead of gen_rtx_PLUS.
12691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12692 Alan Hayward <alan.hayward@arm.com>
12693 David Sherwood <david.sherwood@arm.com>
12695 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12696 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12697 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12698 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12699 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12700 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12701 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12702 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12703 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12704 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12706 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12707 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12708 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12709 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12710 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12711 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12712 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12713 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12714 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12715 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12717 * expr.c (emit_move_resolve_push): Treat the input and result
12718 of PUSH_ROUNDING as a poly_int64.
12719 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12720 (emit_push_insn): Likewise.
12721 * lra-eliminations.c (mark_not_eliminable): Likewise.
12722 * recog.c (push_operand): Likewise.
12723 * reload1.c (elimination_effects): Likewise.
12724 * rtlanal.c (nonzero_bits1): Likewise.
12725 * calls.c (store_one_arg): Likewise. Require the padding to be
12726 known at compile time.
12728 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12729 Alan Hayward <alan.hayward@arm.com>
12730 David Sherwood <david.sherwood@arm.com>
12732 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12733 Use plus_constant instead of gen_rtx_PLUS.
12735 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12736 Alan Hayward <alan.hayward@arm.com>
12737 David Sherwood <david.sherwood@arm.com>
12739 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12740 rather than an int.
12742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12743 Alan Hayward <alan.hayward@arm.com>
12744 David Sherwood <david.sherwood@arm.com>
12746 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12747 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12748 via stack temporaries. Treat the mode size as polynomial too.
12750 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12751 Alan Hayward <alan.hayward@arm.com>
12752 David Sherwood <david.sherwood@arm.com>
12754 * expr.c (expand_expr_real_2): When handling conversions involving
12755 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12756 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12757 as a poly_uint64 too.
12759 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12760 Alan Hayward <alan.hayward@arm.com>
12761 David Sherwood <david.sherwood@arm.com>
12763 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
12770 REGMODE_NATURAL_SIZE.
12771 * expmed.c (store_bit_field_1): Likewise.
12772 * expr.c (store_constructor): Likewise.
12773 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12774 and polynomial REGMODE_NATURAL_SIZE.
12775 (gen_lowpart_common): Likewise.
12776 * reginfo.c (record_subregs_of_mode): Likewise.
12777 * rtlanal.c (read_modify_subreg_p): Likewise.
12779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12780 Alan Hayward <alan.hayward@arm.com>
12781 David Sherwood <david.sherwood@arm.com>
12783 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12784 numbers of elements.
12786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12787 Alan Hayward <alan.hayward@arm.com>
12788 David Sherwood <david.sherwood@arm.com>
12790 * match.pd: Cope with polynomial numbers of vector elements.
12792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12793 Alan Hayward <alan.hayward@arm.com>
12794 David Sherwood <david.sherwood@arm.com>
12796 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12797 in a POINTER_PLUS_EXPR.
12799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12800 Alan Hayward <alan.hayward@arm.com>
12801 David Sherwood <david.sherwood@arm.com>
12803 * omp-simd-clone.c (simd_clone_subparts): New function.
12804 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12805 (ipa_simd_modify_function_body): Likewise.
12807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12808 Alan Hayward <alan.hayward@arm.com>
12809 David Sherwood <david.sherwood@arm.com>
12811 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12812 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12813 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12814 (expand_vector_condition, vector_element): Likewise.
12815 (subparts_gt): New function.
12816 (get_compute_type): Use subparts_gt.
12817 (count_type_subparts): Delete.
12818 (expand_vector_operations_1): Use subparts_gt instead of
12819 count_type_subparts.
12821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12822 Alan Hayward <alan.hayward@arm.com>
12823 David Sherwood <david.sherwood@arm.com>
12825 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12826 (vect_compile_time_alias): ...this new function. Do the calculation
12827 on poly_ints rather than trees.
12828 (vect_prune_runtime_alias_test_list): Update call accordingly.
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-slp.c (vect_build_slp_tree_1): Handle polynomial
12836 (vect_schedule_slp_instance): Likewise.
12838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12839 Alan Hayward <alan.hayward@arm.com>
12840 David Sherwood <david.sherwood@arm.com>
12842 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12843 constant and extern definitions for variable-length vectors.
12844 (vect_get_constant_vectors): Note that the number of units
12845 is known to be constant.
12847 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12848 Alan Hayward <alan.hayward@arm.com>
12849 David Sherwood <david.sherwood@arm.com>
12851 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12852 of units as polynomial. Choose between WIDE and NARROW based
12855 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12856 Alan Hayward <alan.hayward@arm.com>
12857 David Sherwood <david.sherwood@arm.com>
12859 * tree-vect-stmts.c (simd_clone_subparts): New function.
12860 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12862 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12863 Alan Hayward <alan.hayward@arm.com>
12864 David Sherwood <david.sherwood@arm.com>
12866 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12867 vectors as polynomial. Use build_index_vector for
12868 IFN_GOMP_SIMD_LANE.
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 (get_load_store_type): Treat the number of
12875 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12876 for variable-length vectors.
12877 (vectorizable_mask_load_store): Treat the number of units as
12878 polynomial, asserting that it is constant if the condition has
12879 already been enforced.
12880 (vectorizable_store, vectorizable_load): Likewise.
12882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12883 Alan Hayward <alan.hayward@arm.com>
12884 David Sherwood <david.sherwood@arm.com>
12886 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12887 of units as polynomial. Punt if we can't tell at compile time
12888 which vector contains the final result.
12890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12891 Alan Hayward <alan.hayward@arm.com>
12892 David Sherwood <david.sherwood@arm.com>
12894 * tree-vect-loop.c (vectorizable_induction): Treat the number
12895 of units as polynomial. Punt on SLP inductions. Use an integer
12896 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12897 cast of such a series for variable-length floating-point
12900 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12901 Alan Hayward <alan.hayward@arm.com>
12902 David Sherwood <david.sherwood@arm.com>
12904 * tree.h (build_index_vector): Declare.
12905 * tree.c (build_index_vector): New function.
12906 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12907 of units as polynomial, forcibly converting it to a constant if
12908 vectorizable_reduction has already enforced the condition.
12909 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12910 to create a {1,2,3,...} vector.
12911 (vectorizable_reduction): Treat the number of units as polynomial.
12912 Choose vectype_in based on the largest scalar element size rather
12913 than the smallest number of units. Enforce the restrictions
12916 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12917 Alan Hayward <alan.hayward@arm.com>
12918 David Sherwood <david.sherwood@arm.com>
12920 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12921 number of units as polynomial.
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 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12928 * target.def (autovectorize_vector_sizes): Return the vector sizes
12929 by pointer, using vector_sizes rather than a bitmask.
12930 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12931 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12932 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12934 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12935 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12936 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12937 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12938 * omp-general.c (omp_max_vf): Likewise.
12939 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12940 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12941 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12942 * tree-vect-slp.c (vect_slp_bb): Likewise.
12943 * doc/tm.texi: Regenerate.
12944 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12946 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12947 the vector size as a poly_uint64 rather than an unsigned int.
12948 (current_vector_size): Change from an unsigned int to a poly_uint64.
12949 (get_vectype_for_scalar_type): Update accordingly.
12950 * tree.h (build_truth_vector_type): Take the size and number of
12951 units as a poly_uint64 rather than an unsigned int.
12952 (build_vector_type): Add a temporary overload that takes
12953 the number of units as a poly_uint64 rather than an unsigned int.
12954 * tree.c (make_vector_type): Likewise.
12955 (build_truth_vector_type): Take the number of units as a poly_uint64
12956 rather than an unsigned int.
12958 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12959 Alan Hayward <alan.hayward@arm.com>
12960 David Sherwood <david.sherwood@arm.com>
12962 * target.def (get_mask_mode): Take the number of units and length
12963 as poly_uint64s rather than unsigned ints.
12964 * targhooks.h (default_get_mask_mode): Update accordingly.
12965 * targhooks.c (default_get_mask_mode): Likewise.
12966 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12967 * doc/tm.texi: Regenerate.
12969 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12970 Alan Hayward <alan.hayward@arm.com>
12971 David Sherwood <david.sherwood@arm.com>
12973 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12974 * omp-general.c (omp_max_vf): Likewise.
12975 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12976 (expand_omp_simd): Handle polynomial safelen.
12977 * omp-low.c (omplow_simd_context): Add a default constructor.
12978 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12979 (lower_rec_simd_input_clauses): Update accordingly.
12980 (lower_rec_input_clauses): Likewise.
12982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12983 Alan Hayward <alan.hayward@arm.com>
12984 David Sherwood <david.sherwood@arm.com>
12986 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12987 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12988 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12989 (vect_analyze_slp_cost): Likewise.
12990 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12991 (vect_model_load_cost): Likewise.
12993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12994 Alan Hayward <alan.hayward@arm.com>
12995 David Sherwood <david.sherwood@arm.com>
12997 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12998 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12999 from an unsigned int * to a poly_uint64_pod *.
13000 (calculate_unrolling_factor): New function.
13001 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13004 Alan Hayward <alan.hayward@arm.com>
13005 David Sherwood <david.sherwood@arm.com>
13007 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13008 from an unsigned int to a poly_uint64.
13009 (_loop_vec_info::slp_unrolling_factor): Likewise.
13010 (_loop_vec_info::vectorization_factor): Change from an int
13012 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13013 (vect_get_num_vectors): New function.
13014 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13015 (vect_get_num_copies): Use vect_get_num_vectors.
13016 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13017 to an unsigned int *.
13018 (vect_analyze_data_refs): Change min_vf from an int * to a
13020 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13021 than an unsigned HOST_WIDE_INT.
13022 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13023 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13024 to an unsigned int *.
13025 (vect_analyze_data_ref_dependences): Likewise.
13026 (vect_compute_data_ref_alignment): Handle polynomial vf.
13027 (vect_enhance_data_refs_alignment): Likewise.
13028 (vect_prune_runtime_alias_test_list): Likewise.
13029 (vect_shift_permute_load_chain): Likewise.
13030 (vect_supportable_dr_alignment): Likewise.
13031 (dependence_distance_ge_vf): Take the vectorization factor as a
13032 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13033 (vect_analyze_data_refs): Change min_vf from an int * to a
13035 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13036 vfm1 as a poly_uint64 rather than an int. Make the same change
13037 for the returned bound_scalar.
13038 (vect_gen_vector_loop_niters): Handle polynomial vf.
13039 (vect_do_peeling): Likewise. Update call to
13040 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13041 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13043 * tree-vect-loop.c (vect_determine_vectorization_factor)
13044 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13045 (vect_get_known_peeling_cost): Likewise.
13046 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13047 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13048 (vect_transform_loop): Likewise. Use the lowest possible VF when
13049 updating the upper bounds of the loop.
13050 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13051 rather than an int.
13052 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13053 polynomial unroll factors.
13054 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13055 (vect_make_slp_decision): Likewise.
13056 (vect_supported_load_permutation_p): Likewise, and polynomial
13058 (vect_analyze_slp_cost): Handle polynomial vf.
13059 (vect_slp_analyze_node_operations): Likewise.
13060 (vect_slp_analyze_bb_1): Likewise.
13061 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13062 than an unsigned HOST_WIDE_INT.
13063 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13064 (vectorizable_load): Handle polynomial vf.
13065 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13067 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13070 Alan Hayward <alan.hayward@arm.com>
13071 David Sherwood <david.sherwood@arm.com>
13073 * match.pd: Handle bit operations involving three constants
13074 and try to fold one pair.
13076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13078 * tree-vect-loop-manip.c: Include gimple-fold.h.
13079 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13080 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13081 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13082 Add a path that uses a step of VF instead of 1, but disable it
13084 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13085 and niters_no_overflow parameters. Update calls to
13086 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13087 Create a new SSA name if the latter choses to use a ste other
13088 than zero, and return it via niters_vector_mult_vf_var.
13089 * tree-vect-loop.c (vect_transform_loop): Update calls to
13090 vect_do_peeling, vect_gen_vector_loop_niters and
13091 slpeel_make_loop_iterate_ntimes.
13092 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13093 (vect_gen_vector_loop_niters): Update declarations after above changes.
13095 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13097 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13098 128-bit round to integer instructions.
13099 (ceil<mode>2): Likewise.
13100 (btrunc<mode>2): Likewise.
13101 (round<mode>2): Likewise.
13103 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13105 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13106 unaligned VSX load/store on P8/P9.
13107 (expand_block_clear): Allow the use of unaligned VSX
13108 load/store on P8/P9.
13110 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13112 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13114 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13115 swap associated with both a load and a store.
13117 2018-01-02 Andrew Waterman <andrew@sifive.com>
13119 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13120 * config/riscv/riscv.md (clear_cache): Use it.
13122 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13124 * web.c: Remove out-of-date comment.
13126 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13128 * expr.c (fixup_args_size_notes): Check that any existing
13129 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13130 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13131 (emit_single_push_insn): ...here.
13133 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13135 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13136 (const_vector_encoded_nelts): New function.
13137 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13138 (const_vector_int_elt, const_vector_elt): Declare.
13139 * emit-rtl.c (const_vector_int_elt_1): New function.
13140 (const_vector_elt): Likewise.
13141 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13142 of CONST_VECTOR_ELT.
13144 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13146 * expr.c: Include rtx-vector-builder.h.
13147 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13148 directly on the tree encoding.
13149 (const_vector_from_tree): Likewise.
13150 * optabs.c: Include rtx-vector-builder.h.
13151 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13152 sequence of "u" values.
13153 * vec-perm-indices.c: Include rtx-vector-builder.h.
13154 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13155 directly on the vec_perm_indices encoding.
13157 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13159 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13160 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13161 * rtx-vector-builder.h: New file.
13162 * rtx-vector-builder.c: Likewise.
13163 * rtl.h (rtx_def::u2): Add a const_vector field.
13164 (CONST_VECTOR_NPATTERNS): New macro.
13165 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13166 (CONST_VECTOR_DUPLICATE_P): Likewise.
13167 (CONST_VECTOR_STEPPED_P): Likewise.
13168 (CONST_VECTOR_ENCODED_ELT): Likewise.
13169 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13170 (unwrap_const_vec_duplicate): Likewise.
13171 (const_vec_series_p): Check for a non-duplicated vector encoding.
13172 Say that the function only returns true for integer vectors.
13173 * emit-rtl.c: Include rtx-vector-builder.h.
13174 (gen_const_vec_duplicate_1): Delete.
13175 (gen_const_vector): Call gen_const_vec_duplicate instead of
13176 gen_const_vec_duplicate_1.
13177 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13178 (gen_const_vec_duplicate): Use rtx_vector_builder.
13179 (gen_const_vec_series): Likewise.
13180 (gen_rtx_CONST_VECTOR): Likewise.
13181 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13182 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13183 Build a new vector rather than modifying a CONST_VECTOR in-place.
13184 (handle_special_swappables): Update call accordingly.
13185 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13186 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13187 Build a new vector rather than modifying a CONST_VECTOR in-place.
13188 (handle_special_swappables): Update call accordingly.
13190 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13192 * simplify-rtx.c (simplify_const_binary_operation): Use
13193 CONST_VECTOR_ELT instead of XVECEXP.
13195 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13197 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13198 the selector elements to be different from the data elements
13199 if the selector is a VECTOR_CST.
13200 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13201 ssizetype for the selector.
13203 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13205 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13206 before testing each element individually.
13207 * tree-vect-generic.c (lower_vec_perm): Likewise.
13209 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13211 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13212 * selftest-run-tests.c (selftest::run_tests): Call it.
13213 * vector-builder.h (vector_builder::operator ==): New function.
13214 (vector_builder::operator !=): Likewise.
13215 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13216 (vec_perm_indices::all_from_input_p): New function.
13217 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13218 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13219 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13220 instead of reading the VECTOR_CST directly. Detect whether both
13221 vector inputs are the same before constructing the vec_perm_indices,
13222 and update the number of inputs argument accordingly. Use the
13223 utility functions added above. Only construct sel2 if we need to.
13225 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13227 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13228 the broadcast of the low byte.
13229 (expand_mult_highpart): Use an explicit encoding for the permutes.
13230 * optabs-query.c (can_mult_highpart_p): Likewise.
13231 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13232 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13233 (vectorizable_bswap): Likewise.
13234 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13235 explicit encoding for the power-of-2 permutes.
13236 (vect_permute_store_chain): Likewise.
13237 (vect_grouped_load_supported): Likewise.
13238 (vect_permute_load_chain): Likewise.
13240 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13242 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13243 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13244 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13245 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13246 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13247 (vect_gen_perm_mask_any): Likewise.
13249 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13251 * int-vector-builder.h: New file.
13252 * vec-perm-indices.h: Include int-vector-builder.h.
13253 (vec_perm_indices): Redefine as an int_vector_builder.
13254 (auto_vec_perm_indices): Delete.
13255 (vec_perm_builder): Redefine as a stand-alone class.
13256 (vec_perm_indices::vec_perm_indices): New function.
13257 (vec_perm_indices::clamp): Likewise.
13258 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13259 (vec_perm_indices::new_vector): New function.
13260 (vec_perm_indices::new_expanded_vector): Update for new
13261 vec_perm_indices class.
13262 (vec_perm_indices::rotate_inputs): New function.
13263 (vec_perm_indices::all_in_range_p): Operate directly on the
13264 encoded form, without computing elided elements.
13265 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13266 encoding. Update for new vec_perm_indices class.
13267 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13268 the given vec_perm_builder.
13269 (expand_vec_perm_var): Update vec_perm_builder constructor.
13270 (expand_mult_highpart): Use vec_perm_builder instead of
13271 auto_vec_perm_indices.
13272 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13273 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13274 or double series encoding as appropriate.
13275 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13276 vec_perm_indices instead of auto_vec_perm_indices.
13277 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13278 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13279 (vect_permute_store_chain): Likewise.
13280 (vect_grouped_load_supported): Likewise.
13281 (vect_permute_load_chain): Likewise.
13282 (vect_shift_permute_load_chain): Likewise.
13283 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13284 (vect_transform_slp_perm_load): Likewise.
13285 (vect_schedule_slp_instance): Likewise.
13286 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13287 (vectorizable_mask_load_store): Likewise.
13288 (vectorizable_bswap): Likewise.
13289 (vectorizable_store): Likewise.
13290 (vectorizable_load): Likewise.
13291 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13292 vec_perm_indices instead of auto_vec_perm_indices. Use
13293 tree_to_vec_perm_builder to read the vector from a tree.
13294 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13295 vec_perm_builder instead of a vec_perm_indices.
13296 (have_whole_vector_shift): Use vec_perm_builder and
13297 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13298 truncation to calc_vec_perm_mask_for_shift.
13299 (vect_create_epilog_for_reduction): Likewise.
13300 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13301 from auto_vec_perm_indices to vec_perm_indices.
13302 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13303 instead of changing individual elements.
13304 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13305 the vector in d.perm.
13306 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13307 from auto_vec_perm_indices to vec_perm_indices.
13308 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13309 instead of changing individual elements.
13310 (arm_vectorize_vec_perm_const): Use new_vector to install
13311 the vector in d.perm.
13312 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13313 Update vec_perm_builder constructor.
13314 (rs6000_expand_interleave): Likewise.
13315 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13316 (rs6000_expand_interleave): Likewise.
13318 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13320 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13321 to qimode could truncate the indices.
13322 * optabs.c (expand_vec_perm_var): Likewise.
13324 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13326 * Makefile.in (OBJS): Add vec-perm-indices.o.
13327 * vec-perm-indices.h: New file.
13328 * vec-perm-indices.c: Likewise.
13329 * target.h (vec_perm_indices): Replace with a forward class
13331 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13332 * optabs.h: Include vec-perm-indices.h.
13333 (expand_vec_perm): Delete.
13334 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13335 (expand_vec_perm_const): Declare.
13336 * target.def (vec_perm_const_ok): Replace with...
13337 (vec_perm_const): ...this new hook.
13338 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13339 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13340 * doc/tm.texi: Regenerate.
13341 * optabs.def (vec_perm_const): Delete.
13342 * doc/md.texi (vec_perm_const): Likewise.
13343 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13344 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13345 expand_vec_perm for constant permutation vectors. Assert that
13346 the mode of variable permutation vectors is the integer equivalent
13347 of the mode that is being permuted.
13348 * optabs-query.h (selector_fits_mode_p): Declare.
13349 * optabs-query.c: Include vec-perm-indices.h.
13350 (selector_fits_mode_p): New function.
13351 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13352 is defined, instead of checking whether the vec_perm_const_optab
13353 exists. Use targetm.vectorize.vec_perm_const instead of
13354 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13355 fit in the vector mode before using a variable permute.
13356 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13357 vec_perm_indices instead of an rtx.
13358 (expand_vec_perm): Replace with...
13359 (expand_vec_perm_const): ...this new function. Take the selector
13360 as a vec_perm_indices rather than an rtx. Also take the mode of
13361 the selector. Update call to shift_amt_for_vec_perm_mask.
13362 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13363 Use vec_perm_indices::new_expanded_vector to expand the original
13364 selector into bytes. Check whether the indices fit in the vector
13365 mode before using a variable permute.
13366 (expand_vec_perm_var): Make global.
13367 (expand_mult_highpart): Use expand_vec_perm_const.
13368 * fold-const.c: Includes vec-perm-indices.h.
13369 * tree-ssa-forwprop.c: Likewise.
13370 * tree-vect-data-refs.c: Likewise.
13371 * tree-vect-generic.c: Likewise.
13372 * tree-vect-loop.c: Likewise.
13373 * tree-vect-slp.c: Likewise.
13374 * tree-vect-stmts.c: Likewise.
13375 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13377 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13378 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13379 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13380 (aarch64_vectorize_vec_perm_const): ...this new function.
13381 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13382 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13383 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13384 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13385 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13386 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13387 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13389 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13390 check for NEON modes.
13391 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13392 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13393 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13394 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13396 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13397 the old VEC_PERM_CONST conditions.
13398 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13399 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13400 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13401 (ia64_vectorize_vec_perm_const_ok): Merge into...
13402 (ia64_vectorize_vec_perm_const): ...this new function.
13403 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13404 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13405 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13406 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13407 * config/mips/mips.c (mips_expand_vec_perm_const)
13408 (mips_vectorize_vec_perm_const_ok): Merge into...
13409 (mips_vectorize_vec_perm_const): ...this new function.
13410 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13411 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13412 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13413 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13414 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13415 (rs6000_expand_vec_perm_const): Delete.
13416 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13418 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13419 (altivec_expand_vec_perm_const_le): Take each operand individually.
13420 Operate on constant selectors rather than rtxes.
13421 (altivec_expand_vec_perm_const): Likewise. Update call to
13422 altivec_expand_vec_perm_const_le.
13423 (rs6000_expand_vec_perm_const): Delete.
13424 (rs6000_vectorize_vec_perm_const_ok): Delete.
13425 (rs6000_vectorize_vec_perm_const): New function.
13426 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13427 an element count and rtx array.
13428 (rs6000_expand_extract_even): Update call accordingly.
13429 (rs6000_expand_interleave): Likewise.
13430 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13431 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13432 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13433 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13434 (rs6000_expand_vec_perm_const): Delete.
13435 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13436 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13437 (altivec_expand_vec_perm_const_le): Take each operand individually.
13438 Operate on constant selectors rather than rtxes.
13439 (altivec_expand_vec_perm_const): Likewise. Update call to
13440 altivec_expand_vec_perm_const_le.
13441 (rs6000_expand_vec_perm_const): Delete.
13442 (rs6000_vectorize_vec_perm_const_ok): Delete.
13443 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13444 reference to the SPE evmerge intructions.
13445 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13446 an element count and rtx array.
13447 (rs6000_expand_extract_even): Update call accordingly.
13448 (rs6000_expand_interleave): Likewise.
13449 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13450 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13452 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13454 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13456 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13457 vector mode and that that mode matches the mode of the data
13459 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13460 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13461 directly using expand_vec_perm_1 when forcing selectors into
13463 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13465 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13467 * optabs-query.h (can_vec_perm_p): Delete.
13468 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13469 * optabs-query.c (can_vec_perm_p): Split into...
13470 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13471 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13472 particular selector is valid.
13473 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13474 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13475 (vect_grouped_load_supported): Likewise.
13476 (vect_shift_permute_load_chain): Likewise.
13477 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13478 (vect_transform_slp_perm_load): Likewise.
13479 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13480 (vectorizable_bswap): Likewise.
13481 (vect_gen_perm_mask_checked): Likewise.
13482 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13483 implementations of variable permutation vectors into account
13484 when deciding which selector to use.
13485 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13486 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13487 with a false third argument.
13488 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13489 to test whether the constant selector is valid and can_vec_perm_var_p
13490 to test whether a variable selector is valid.
13492 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13494 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13495 * optabs-query.c (can_vec_perm_p): Likewise.
13496 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13497 instead of vec_perm_indices.
13498 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13499 (vect_gen_perm_mask_checked): Likewise,
13500 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13501 (vect_gen_perm_mask_checked): Likewise,
13503 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13505 * optabs-query.h (qimode_for_vec_perm): Declare.
13506 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13507 (qimode_for_vec_perm): ...this new function.
13508 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13510 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13512 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13513 does not have a conditional at the top.
13515 2018-01-02 Richard Biener <rguenther@suse.de>
13517 * ipa-inline.c (big_speedup_p): Fix expression.
13519 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13522 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13525 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13529 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13530 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13531 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13532 cond_taken_branch_cost 3->4.
13534 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13536 PR tree-optimization/83581
13537 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13538 TODO_cleanup_cfg if any changes have been made.
13540 PR middle-end/83608
13541 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13542 convert_modes if target mode has the right side, but different mode
13545 PR middle-end/83609
13546 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13547 last argument when extracting from CONCAT. If either from_real or
13548 from_imag is NULL, use expansion through memory. If result is not
13549 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13550 the parts directly to inner mode, if even that fails, use expansion
13553 PR middle-end/83623
13554 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13555 check for bswap in mode rather than HImode and use that in expand_unop
13558 Copyright (C) 2018 Free Software Foundation, Inc.
13560 Copying and distribution of this file, with or without modification,
13561 are permitted in any medium without royalty provided the copyright
13562 notice and this notice are preserved.