1 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2 Chung-Ju Wu <jasonwucj@gmail.com>
4 * config/nds32/constants.md: Add TP_REGNUM constant.
5 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
6 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
8 * config/nds32/nds32-doubleword.md: Consider flag_pic.
9 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
10 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
11 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
12 and PIC code generation.
13 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
15 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
17 * config/nds32/nds32.md: Support TLS and PIC.
18 * config/nds32/nds32.c: Support TLS and PIC.
19 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
20 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
23 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
25 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
28 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
29 Chung-Ju Wu <jasonwucj@gmail.com>
31 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
32 * config/nds32/nds32-md-auxiliary.c
33 (symbolic_reference_mentioned_p): New.
34 (nds32_legitimize_ict_address): New.
35 (nds32_expand_ict_move): New.
36 (nds32_indirect_call_referenced_p): New.
37 (nds32_symbol_binds_local_p): Delete.
38 (nds32_long_call_p): Modify.
39 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
40 * config/nds32/nds32-protos.h
41 (symbolic_reference_mentioned_p): Declare.
42 (nds32_legitimize_ict_address): Declare.
43 (nds32_expand_ict_move): Declare.
44 (nds32_indirect_call_referenced_p): Declare.
45 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
46 (nds32_relax_group): Use nds32_ict_const_p as condition.
47 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
48 (nds32_asm_file_start): Output ict_model directive in asm code.
49 (nds32_legitimate_address_p): Consider indirect call.
50 (nds32_print_operand): Consider indirect call.
51 (nds32_print_operand_address): Consider indirect call.
52 (nds32_insert_attributes): Handle "indirect_call" attribute.
53 (TARGET_LEGITIMATE_ADDRESS_P): Define.
54 (TARGET_LEGITIMATE_CONSTANT_P): Define.
55 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
56 (TARGET_DELEGITIMIZE_ADDRESS): Define.
57 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
58 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
59 (TARGET_ICT_MODEL_SMALL): Define.
60 (TARGET_ICT_MODEL_LARGE): Define.
61 * config/nds32/nds32.md (movsi): Consider ict model.
62 (call, call_value): Consider ict model.
63 (sibcall, sibcall_value): Consider ict model.
64 * config/nds32/nds32.opt (mict-model): New option.
65 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
68 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
69 Monk Chiang <sh.chiang04@gmail.com>
70 Jim Wilson <jimw@sifive.com>
72 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
73 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
74 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
75 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
76 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
77 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
78 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
79 compute save_libcall_adjustment properly.
80 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
81 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
82 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
83 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
84 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
85 (ABI_SPEC): Handle mabi=ilp32e.
86 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
88 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
89 <-march>: Add rv32e as an example.
91 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
94 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
95 (intra_create_variable_infos): Handle C++ constructors.
97 2018-05-18 Martin Liska <mliska@suse.cz>
99 * passes.def: Remove a redundant pass.
101 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
104 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
106 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
108 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
110 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
112 (armv2, armv2a, armv3, armv3m): Delete architectures.
113 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
114 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
115 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
117 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
118 (*mulsidi3adddi): Likewise.
119 (mulsidi3): Likewise.
120 (*mulsidi3_nov6): Likewise.
121 (umulsidi3): Likewise.
122 (umulsidi3_nov6): Likewise.
123 (umaddsidi4): Likewise.
124 (*umulsidi3adddi): Likewise.
125 (smulsi3_highpart): Likewise.
126 (*smulsi3_highpart_nov6): Likewise.
127 (umulsi3_highpart): Likewise.
128 (*umulsi3_highpart_nov6): Likewise.
129 * config/arm/arm.h (arm_arch3m): Delete.
130 * config/arm/arm.c (arm_arch3m): Delete.
131 (arm_option_override_internal): Update armv3-related comment.
132 (arm_configure_build_target): Delete use of isa_bit_mode32.
133 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
134 (arm_rtx_costs_internal): Delete check of arm_arch3m.
135 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
138 * config/arm/arm-protos.h (arm_arch3m): Delete.
139 * config/arm/arm-tables.opt: Regenerate.
140 * config/arm/arm-tune.md: Likewise.
141 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
142 deleted architectures.
144 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
146 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
147 (armv5t, armv5te): New features.
148 (ARMv5, ARMv5e): Delete fgroups.
149 (ARMv5t, ARMv5te): Adjust for above changes.
151 (armv5, armv5e): Delete arches.
152 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
154 (*call_reg_arm): Likewise.
155 (*call_value_reg_armv5): Likewise.
156 (*call_value_reg_arm): Likewise.
157 (*call_symbol): Likewise.
158 (*call_value_symbol): Likewise.
159 (*sibcall_insn): Likewise.
160 (*sibcall_value_insn): Likewise.
162 (prefetch): Likewise.
163 (define_split and define_peephole2 dependent on arm_arch5):
165 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
167 (TARGET_ARM_QBIT): Likewise.
168 (TARGET_DSP_MULTIPLY): Likewise.
169 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
170 (arm_arch5, arm_arch5e): Delete.
171 (arm_arch5t, arm_arch5te): Declare.
172 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
173 (arm_arch5t): Declare.
174 (arm_option_reconfigure_globals): Update for the above.
175 (arm_options_perform_arch_sanity_checks): Update comment, replace
176 use of arm_arch5 with arm_arch5t.
177 (use_return_insn): Likewise.
178 (arm_emit_call_insn): Likewise.
179 (output_return_instruction): Likewise.
180 (arm_final_prescan_insn): Likewise.
181 (arm_coproc_builtin_available): Likewise.
182 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
183 arm_arch5e with arm_arch5t and arm_arch5te.
184 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
185 (arm_arch5t, arm_arch5te): Declare.
186 * config/arm/arm-tables.opt: Regenerate.
187 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
188 * config/arm/t-multilib: Likewise.
189 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
190 instead of arm_arch5.
191 (*call_reg_thumb1): Likewise.
192 (*call_value_reg_thumb1_v5): Likewise.
193 (*call_value_reg_thumb1): Likewise.
194 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
196 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
198 2018-05-18 Martin Liska <mliska@suse.cz>
200 PR gcov-profile/84846
201 * doc/gcov.texi: Document -t option of gcov tool.
203 2018-05-18 Martin Liska <mliska@suse.cz>
205 PR gcov-profile/84846
206 * gcov.c (print_usage): Add new -t option.
207 (process_args): Handle the option.
208 (generate_results): Use stdout as output when requested by
211 2018-05-18 Martin Liska <mliska@suse.cz>
213 PR gcov-profile/84846
214 * coverage.c (coverage_init): Write PWD to .gcno file.
215 * doc/gcov.texi: Document how working directory is printed.
216 * gcov-dump.c (dump_gcov_file): Print PWD.
217 * gcov.c (output_intermediate_file): Likewise.
218 (read_graph_file): Read PWD string.
219 (output_lines): Print PWD.
221 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
224 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
225 for retval and return false if all args to phi are zero.
227 2018-05-18 Richard Biener <rguenther@suse.de>
229 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
231 (evrp_dom_walker::before_dom_children): Call it.
233 2018-05-18 Richard Biener <rguenther@suse.de>
235 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
236 results when processing array refs with variable index.
238 2018-05-18 Toon Moene <toon@moene.org>
240 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
241 directly after that of -floop-interchange. Indicate that both
242 options are enabled by default when specifying -O3.
244 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
246 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
247 iterator. Delete separate integer-mode vec_set<mode> expander.
248 (aarch64_simd_vec_setv2di): Delete.
249 (vec_setv2di): Delete.
250 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
251 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
252 the "w, r" alternative.
254 2018-05-18 Martin Liska <mliska@suse.cz>
256 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
257 * tree-pass.h (make_pass_lower_switch_O0): New function.
258 * tree-switch-conversion.c (node_has_low_bound): Remove.
259 (node_has_high_bound): Likewise.
260 (node_is_bounded): Likewise.
261 (class pass_lower_switch): Make it a template type and create
263 (pass_lower_switch::execute): Add template argument.
264 (make_pass_lower_switch): New function.
265 (make_pass_lower_switch_O0): New function.
266 (do_jump_if_equal): Remove.
267 (emit_case_nodes): Simplify to just handle all 3 cases and leave
268 all the hard work to tree optimization passes.
270 2018-05-18 Martin Liska <mliska@suse.cz>
272 * dbgcnt.c (limit_low): Renamed from limit.
273 (limit_high): New variable.
274 (dbg_cnt_is_enabled): Check for upper limit.
275 (dbg_cnt): Adjust dumping.
276 (dbg_cnt_set_limit_by_index): Add new argument for high
278 (dbg_cnt_set_limit_by_name): Likewise.
279 (dbg_cnt_process_single_pair): Parse new format.
280 (dbg_cnt_process_opt): Use strtok.
281 (dbg_cnt_list_all_counters): Remove 'value' and add
283 * doc/invoke.texi: Document changes.
285 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
287 * doc/sourcebuild.texi (scalar_all_fma): Document.
288 * tree.def (FMA_EXPR): Delete.
289 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
290 * internal-fn.c (ternary_direct): New macro.
291 (expand_ternary_optab_fn): Likewise.
292 (direct_ternary_optab_supported_p): Likewise.
293 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
294 * builtins.c (fold_builtin_fma): Delete.
295 (fold_builtin_3): Don't call it.
296 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
297 * expr.c (expand_expr_real_2): Likewise.
298 * fold-const.c (operand_equal_p): Likewise.
299 (fold_ternary_loc): Likewise.
300 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
301 * gimple.c (DEFTREECODE): Likewise.
302 * gimplify.c (gimplify_expr): Likewise.
303 * optabs-tree.c (optab_for_tree_code): Likewise.
304 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
305 * tree-eh.c (operation_could_trap_p): Likewise.
306 (stmt_could_throw_1_p): Likewise.
307 * tree-inline.c (estimate_operator_cost): Likewise.
308 * tree-pretty-print.c (dump_generic_node): Likewise.
309 (op_code_prio): Likewise.
310 * tree-ssa-loop-im.c (stmt_cost): Likewise.
311 * tree-ssa-operands.c (get_expr_operands): Likewise.
312 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
313 * fold-const-call.h (fold_fma): Delete.
314 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
315 CFN_FNMA and CFN_FNMS.
317 * genmatch.c (combined_fn): New enum.
318 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
319 (commutative_op): New function.
320 (commutate): Use it. Handle more than 2 operands.
321 (dt_operand::gen_gimple_expr): Use commutative_op.
322 (parser::parse_expr): Allow :c to be used with non-binary
323 operators if the commutative operand is known.
324 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
325 CFN_FMS, CFN_FNMA and CFN_FNMS.
326 (backprop::process_assign_use): Remove FMA_EXPR handling.
327 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
328 (gen_hsa_fma): New function.
329 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
330 IFN_FNMA and IFN_FNMS.
331 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
332 * gimple-fold.h (follow_all_ssa_edges): Declare.
333 * gimple-fold.c (follow_all_ssa_edges): New function.
334 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
335 gimple_build interface and use follow_all_ssa_edges to fold the result.
336 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
337 instead of checking for optabs directly.
338 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
339 rather than FMA_EXPRs.
340 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
341 call to IFN_FMA instead of an FMA_EXPR.
343 2018-05-17 Jim Wilson <jimw@sifive.com>
345 * expr.c (do_tablejump): When converting index to Pmode, if we have a
346 sign extended promoted subreg, and the range does not have the sign bit
347 set, then do a sign extend.
349 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
350 test, check for sign extended subreg and/or constant operands, and
351 do a sign extend in that case.
353 2018-05-17 Steve Ellcey <sellcey@cavium.com>
355 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
356 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
358 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
359 Change logics_shift_reg to logics_shift_imm.
360 (thunderx2t99_fp_loadpair_basic): Delete.
361 (thunderx2t99_fp_storepair_basic): Delete.
362 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
363 (thunderx2t99_asimd_polynomial): Delete.
364 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
365 and neon_fp_mul_d_scalar_q.
366 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
367 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
368 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
369 (thunderx2t99_asimd_lut): Add missing tbl types.
370 (thunderx2t99_asimd_ext): Delete.
371 (thunderx2t99_asimd_load1_1_mult): Delete.
372 (thunderx2t99_asimd_load1_2_mult): Delete.
373 (thunderx2t99_asimd_load1_ldp): New.
374 (thunderx2t99_asimd_load1): New.
375 (thunderx2t99_asimd_load2): Add missing *load2* types.
376 (thunderx2t99_asimd_load3): New.
377 (thunderx2t99_asimd_load4): New.
378 (thunderx2t99_asimd_store1_1_mult): Delete.
379 (thunderx2t99_asimd_store1_2_mult): Delete.
380 (thunderx2t99_asimd_store2_mult): Delete.
381 (thunderx2t99_asimd_store2_onelane): Delete.
382 (thunderx2t99_asimd_store_stp): New.
383 (thunderx2t99_asimd_store1): New.
384 (thunderx2t99_asimd_store2): New.
385 (thunderx2t99_asimd_store3): New.
386 (thunderx2t99_asimd_store4): New.
388 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
390 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
391 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
393 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
394 Segher Boessenkool <segher@kernel.crashing.org>
397 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
400 2018-05-17 Richard Biener <rguenther@suse.de>
402 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
403 for pruning loop and prune defs feeding only already visited PHIs.
405 2018-05-17 Richard Biener <rguenther@suse.de>
407 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
409 2018-05-17 Bin Cheng <bin.cheng@arm.com>
410 Richard Biener <rguenther@suse.de>
412 PR tree-optimization/85793
413 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
414 for VMAT_ELEMENTWISE.
416 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
418 * internal-fn.h (lookup_internal_fn): Declare
419 * internal-fn.c (lookup_internal_fn): New function.
420 * gimple.c (gimple_build_call_from_tree): Handle calls to
422 * gimple-pretty-print.c (dump_gimple_call): Print "." before
423 internal function names.
424 * tree-pretty-print.c (dump_generic_node): Likewise.
425 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
427 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
429 * gimple-fold.h (gimple_build): Make the function forms take
430 combined_fn rather than built_in_function.
431 (gimple_simplify): Likewise.
432 * gimple-match-head.c (gimple_simplify): Likewise.
433 * gimple-fold.c (gimple_build): Likewise.
434 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
435 rather than gimple_build_call_internal.
436 (get_initial_defs_for_reduction): Likewise.
437 (vect_create_epilog_for_reduction): Likewise.
438 (vectorizable_live_operation): Likewise.
440 2018-05-17 Martin Liska <mliska@suse.cz>
442 * gimple-ssa-sprintf.c (format_directive): Do not use
443 space in between 'G_' and '('.
445 2018-05-17 Jakub Jelinek <jakub@redhat.com>
448 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
449 even if the mask is not all ones.
452 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
454 (ix86_gimple_fold_builtin): Likewise.
457 * config/i386/i386.c: Include tree-vector-builder.h.
458 (ix86_vector_shift_count): New function.
459 (ix86_fold_builtin): Fold shift builtins by scalar count.
460 (ix86_gimple_fold_builtin): Likewise.
462 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
463 _mm512_setzero): New intrinsics.
465 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
466 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
468 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
469 code generation for cases where splatting a value is not useful.
470 * simplify-rtx.c (simplify_ternary_operation): Simplify
471 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
472 mode to a vec_concat.
474 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
476 * config.gcc: Support "goldmont-plus".
477 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
479 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
480 PROCESSOR_GOLDMONT_PLUS.
481 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
482 (processor_target_table): Add "goldmont-plus".
483 (PTA_GOLDMONT_PLUS): Define.
484 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
485 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
486 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
487 (fold_builtin_cpu): Add "goldmont-plus".
488 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
489 (ix86_option_override_internal): Add "goldmont-plus".
490 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
491 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
492 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
493 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
495 2018-05-17 Richard Biener <rguenther@suse.de>
497 PR tree-optimization/85757
498 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
499 remove defs that only feed that PHI from further processing.
501 2018-05-16 Jim Wilson <jimw@sifive.com>
503 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
505 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
507 2018-05-16 Mark Wielaard <mark@klomp.org>
509 * dwarf2out.c (count_index_strings): New function.
510 (output_indirect_strings): Call count_index_strings and generate
511 header for dwarf_version >= 5.
513 2018-05-16 Mark Wielaard <mark@klomp.org>
515 * dwarf2out.c (dwarf_FORM): New function.
516 (set_indirect_string): Use dwarf_FORM.
517 (reset_indirect_string): Likewise.
518 (size_of_die): Likewise.
519 (value_format): Likewise.
520 (output_die): Likewise.
521 (add_skeleton_AT_string): Likewise.
522 (output_macinfo_op): Likewise.
523 (index_string): Likewise.
524 (output_index_string_offset): Likewise.
525 (output_index_string): Likewise.
526 (count_index_strings): Likewise.
528 2018-05-16 Carl Love <cel@us.ibm.com>
530 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
531 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
533 2018-05-16 Martin Jambor <mjambor@suse.cz>
535 * ipa-prop.c (ipa_free_all_edge_args): Remove.
536 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
538 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
540 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
541 (fnma<mode>4): Likewise.
542 (fms<mode>4): Likewise.
543 (fnms<mode>4): Likewise.
544 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
545 (aarch64_fnma<mode>4): Likewise.
546 (aarch64_fms<mode>4): Likewise.
547 (aarch64_fnms<mode>4): Likewise.
548 (aarch64_fnmadd<mode>4): Likewise.
550 2018-05-16 Jason Merrill <jason@redhat.com>
552 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
554 2018-05-16 Richard Biener <rguenther@suse.de>
556 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
557 (dump_stmt_cost): Declare.
558 (add_stmt_cost): Dump cost we add.
559 (add_stmt_costs): New function.
560 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
562 (vect_analyze_stmt): Adjust prototype.
563 (vectorizable_condition): Likewise.
564 (vectorizable_live_operation): Likewise.
565 (vectorizable_reduction): Likewise.
566 (vectorizable_induction): Likewise.
567 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
568 cost vector to pass to vectorizable_ and record afterwards.
569 (vect_model_reduction_cost): Take cost vector argument and adjust.
570 (vect_model_induction_cost): Likewise.
571 (vectorizable_reduction): Likewise.
572 (vectorizable_induction): Likewise.
573 (vectorizable_live_operation): Likewise.
574 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
575 SLP_TREE_NUMBER_OF_VEC_STMTS.
576 (vect_analyze_slp_cost_1): Remove.
577 (vect_analyze_slp_cost): Likewise.
578 (vect_slp_analyze_node_operations): Take visited args and
579 a target cost vector. Avoid processing already visited stmt sets.
580 (vect_slp_analyze_operations): Use a local cost vector to gather
581 costs and register those of non-discarded instances.
582 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
583 (vect_schedule_slp_instance): Remove copying of
584 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
586 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
587 adding cost. Record cost entry location.
588 (vect_prologue_cost_for_slp_op): Function to compute cost of
589 a constant or invariant generated for SLP vect in the prologue,
590 split out from vect_analyze_slp_cost_1.
591 (vect_model_simple_cost): Make static. Adjust for SLP costing.
592 (vect_model_promotion_demotion_cost): Likewise.
593 (vect_model_store_cost): Likewise, make static.
594 (vect_model_load_cost): Likewise.
595 (vectorizable_bswap): Add cost vector arg and adjust.
596 (vectorizable_call): Likewise.
597 (vectorizable_simd_clone_call): Likewise.
598 (vectorizable_conversion): Likewise.
599 (vectorizable_assignment): Likewise.
600 (vectorizable_shift): Likewise.
601 (vectorizable_operation): Likewise.
602 (vectorizable_store): Likewise.
603 (vectorizable_load): Likewise.
604 (vectorizable_condition): Likewise.
605 (vectorizable_comparison): Likewise.
606 (can_vectorize_live_stmts): Likewise.
607 (vect_analyze_stmt): Likewise.
608 (vect_transform_stmt): Adjust calls to vectorizable_*.
609 * tree-vectorizer.c: Include gimple-pretty-print.h.
610 (dump_stmt_cost): New function.
612 2018-05-16 Richard Biener <rguenther@suse.de>
614 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
615 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
616 * tree-ssa-dse.c: Include tree-ssa-loop.h.
617 (check_name): New callback.
618 (dse_classify_store): Track cycles via a visited bitmap of PHI
619 defs and simplify handling of in-loop and across loop dead stores
620 and properly fail for loop-variant refs. Handle byte-tracking with
621 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
624 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
626 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
627 (vect_get_mask_type_for_stmt): Likewise.
628 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
630 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
631 to determine the statement's vector type and the vector type that
632 should be used for calculating nunits. Deal with cases in which
633 the type has to be deferred.
634 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
635 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
636 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
637 (vect_determine_vf_for_stmt): New functions, split out from...
638 (vect_determine_vectorization_factor): ...here.
639 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
640 (vect_get_mask_type_for_stmt): New functions, split out from
641 vect_determine_vectorization_factor.
643 2018-05-16 Richard Biener <rguenther@suse.de>
645 * tree-cfg.c (verify_gimple_assign_ternary): Properly
646 verify the [VEC_]COND_EXPR embedded comparison.
648 2018-05-15 Martin Sebor <msebor@redhat.com>
650 PR tree-optimization/85753
651 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
652 RECORD_TYPE in addition to ARRAY_TYPE.
654 2018-05-15 Martin Sebor <msebor@redhat.com>
657 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
659 2018-05-15 Richard Biener <rguenther@suse.de>
661 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
662 add by_clobber_p one. Change algorithm to collect all defs
663 representing uses we need to walk and try reducing them to
664 a single one before failing.
665 (dse_dom_walker::dse_optimize_stmt): Adjust.
667 2018-05-13 Mark Wielaard <mark@klomp.org>
669 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
670 (size_of_loc_descr): Likewise.
671 (output_loc_operands): Likewise.
672 (output_loc_operands_raw): Likewise.
673 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
674 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
675 (hash_loc_operands): Likewise.
676 (compare_loc_operands): Likewise.
678 2018-05-14 Mark Wielaard <mark@klomp.org>
680 * dwarf2out.c (count_index_addrs): New function.
681 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
683 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
685 PR tree-optimization/83648
686 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
687 return value as malloc candidate.
689 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
692 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
693 as true in call to suggest_attribute.
695 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
697 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
700 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
702 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
704 (aarch64_crypto_aesd_fused): Likewise.
706 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
708 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
709 (movsi_aarch64): Likewise.
710 (load_pairsi): Likewise.
711 (load_pairdi): Likewise.
712 (store_pairsi): Likewise.
713 (store_pairdi): Likewise.
714 (load_pairsf): Likewise.
715 (load_pairdf): Likewise.
716 (store_pairsf): Likewise.
717 (store_pairdf): Likewise.
718 (zero_extend): Likewise.
719 (trunc): Swap alternatives.
720 (fcvt_target): Add '?' to prefer w over r.
722 2018-05-14 Jakub Jelinek <jakub@redhat.com>
725 * config/i386/i386.md: Disallow non-commutative arithmetics in
726 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
727 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
728 in the peephole2 before it.
730 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
732 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
733 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
734 (ix86_handle_option): Handle -mcldemote.
735 * config.gcc: New header.
736 * config/i386/cldemoteintrin.h: New file.
737 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
738 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
740 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
741 OPTION_MASK_ISA_CLDEMOTE.
742 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
743 (ix86_valid_target_attribute_inner_p): Ditto.
744 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
745 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
746 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
747 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
748 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
750 * config/i386/i386.opt: Add -mcldemote.
751 * config/i386/x86intrin.h: New header.
752 * doc/invoke.texi: Add -mcldemote.
754 2018-05-14 Richard Biener <rguenther@suse.de>
756 * doc/match-and-simplify.texi: Adjust :s documentation.
758 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
760 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
761 intended memcpy size.
762 (REORDER_45): Likewise.
764 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
767 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
768 * vec.c (qsort_chk): Use gcc_qsort.
769 * Makefile.in (OBJS-libcommon): Add sort.o.
770 (build/sort.o): New target. Use it...
771 (BUILD_RTL): ... here, and...
772 (build/gencfn-macros): ... here, and...
773 (build/genmatch): ... here.
775 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
776 Chung-Ju Wu <jasonwucj@gmail.com>
778 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
779 * config/nds32/nds32-graywolf.md: New file.
780 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
781 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
783 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
784 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
785 * config/nds32/nds32.md (pipeline_model): Add graywolf.
786 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
787 * config/nds32/pipelines.md: Include n15 settings.
789 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
790 Chung-Ju Wu <jasonwucj@gmail.com>
792 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
793 * config/nds32/nds32-n13.md: New file.
794 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
795 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
797 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
798 * config/nds32/nds32.md (pipeline_model): Add n13.
799 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
800 * config/nds32/pipelines.md: Include n13 settings.
802 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
803 Chung-Ju Wu <jasonwucj@gmail.com>
805 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
806 * config/nds32/nds32-n10.md: New file.
807 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
808 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
810 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
811 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
812 * config/nds32/nds32.md (pipeline_model): Add n10.
813 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
814 * config/nds32/pipelines.md: Include n10 settings.
816 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
817 Kito Cheng <kito.cheng@gmail.com>
818 Chung-Ju Wu <jasonwucj@gmail.com>
820 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
821 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
822 Add enum values for DSP extension instructions.
823 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
825 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
826 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
828 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
829 * config/nds32/nds32-dspext.md: New file for DSP implementation.
830 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
831 * config/nds32/nds32-intrinsic.md: Likewise.
832 * config/nds32/nds32_intrinsic.h: Likewise.
833 * config/nds32/nds32-md-auxiliary.c: Likewise.
834 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
835 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
836 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
837 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
838 * config/nds32/nds32-protos.h: New declarations for DSP extension.
839 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
840 TYPE_DMAC in switch statement.
841 * config/nds32/nds32.c: New checking and implementation for DSP
842 extension instructions.
843 * config/nds32/nds32.h: Likewise.
844 * config/nds32/nds32.md: Likewise.
845 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
846 * config/nds32/predicates.md: Implement new predicates for DSP
849 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
851 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
852 Reformat alternatives and attributes so it is easier to identify
853 which constraints/attributes go with which instruction.
854 (mov<mode>_hardfloat32, FMOVE64): Likewise.
855 (mov<mode>_softfloat32, FMOVE64): Likewise.
856 (mov<mode>_hardfloat64, FMOVE64): Likewise.
857 (mov<mode>_softfloat64, FMOVE64): Likewise.
859 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
861 * doc/extend.texi (PowerPC Built-in Functions): Rename this
863 (Basic PowerPC Built-in Functions): The new name of the
864 subsection previously known as "PowerPC Built-in Functions".
865 (Basic PowerPC Built-in Functions Available on all Configurations):
867 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
868 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
869 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
870 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
872 2018-05-11 Martin Jambor <mjambor@suse.cz>
875 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
878 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
881 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
883 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
885 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
886 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
887 (ix86_handle_option): Handle -mwaitpkg.
888 * config.gcc: New header.
889 * config/i386/cpuid.h (bit_WAITPKG): New bit.
890 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
891 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
893 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
894 OPTION_MASK_ISA_WAITPKG.
895 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
896 (ix86_option_override_internal): Add PTA_WAITPKG.
897 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
898 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
900 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
901 __builtin_ia32_umwait and __builtin_ia32_tpause.
902 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
903 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
904 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
905 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
906 UNSPECV_TPAUSE): New.
907 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
908 * config/i386/i386.opt: Add -mwaitpkg.
909 * config/i386/waitpkgintrin.h: New file.
910 * config/i386/x86intrin.h: New header.
911 * doc/invoke.texi: Add -mwaitpkg.
913 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
916 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
918 (cortex-m0): Use armv6s-m isa.
919 (cortex-m0plus): Likewise.
920 (cortex-m1): Likewise.
921 (cortex-m0.small-multiply): Likewise.
922 (cortex-m0plus.small-multiply): Likewise.
923 (cortex-m1.small-multiply): Likewise.
925 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
926 Jakub Jelinek <jakub@redhat.com>
928 PR tree-optimization/85692
929 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
930 source permute as well.
932 2018-05-11 Martin Liska <mliska@suse.cz>
935 * doc/extend.texi: Document LLVM style format for no_sanitize
938 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
940 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
941 mode_supports_vsx_dform_quad to mode_supports_dq_form.
942 (mode_supports_vsx_dform_quad): Likewise.
943 (mode_supports_vmx_dform): Move these functions to be next to the
944 other mode_supports functions.
945 (mode_supports_dq_form): Likewise.
946 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
947 mode_supports_dq_form.
948 (reg_offset_addressing_ok_p): Likewise.
949 (offsettable_ok_by_alignment): Likewise.
950 (rs6000_legitimate_offset_address_p): Likewise.
951 (legitimate_lo_sum_address_p): Likewise.
952 (rs6000_legitimize_address): Likewise.
953 (rs6000_legitimize_reload_address): Likewise.
954 (rs6000_secondary_reload_inner): Likewise.
955 (rs6000_preferred_reload_class): Likewise.
956 (rs6000_output_move_128bit): Likewise.
958 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
960 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
961 Generate SImode target register for null target.
962 <case IX86_BUILTIN_XGETBV>: Ditto.
963 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
964 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
966 2018-05-10 Carl Love <cel@us.ibm.com>
968 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
969 dcbtt and dcbtstt if operands[2] is 0.
971 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
974 * config/i386/sse.md (usadv64qi): New expander.
976 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
978 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
979 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
980 -maltivec=be support.
981 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
982 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
983 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
984 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
985 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
986 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
987 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
988 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
989 altivec_vsumsws): Adjust.
990 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
991 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
992 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
994 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
995 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
996 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
997 (altivec_lve<VI_char>x): Delete expand.
998 (*altivec_lve<VI_char>x_internal): Rename to...
999 (altivec_lve<VI_char>x): ... this.
1000 (altivec_lvxl_<mode>): Delete expand.
1001 (*altivec_lvxl_<mode>_internal): Rename to ...
1002 (altivec_lvxl_<mode>): ... this.
1003 (altivec_stvxl_<mode>): Delete expand.
1004 (*altivec_stvxl_<mode>_internal): Rename to ...
1005 (altivec_stvxl_<mode>): ... this.
1006 (altivec_stve<VI_char>x): Delete expand.
1007 (*altivec_stve<VI_char>x_internal): Rename to ...
1008 (altivec_stve<VI_char>x): ... this.
1009 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1010 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1011 reduc_plus_scal_<mode>): Adjust.
1012 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1014 (rs6000_cpu_cpp_builtins): Adjust.
1015 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1016 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1017 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1018 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1019 -maltivec=be support.
1020 (rs6000_split_vec_extract_var): Adjust.
1021 (rs6000_split_v4si_init): Adjust.
1022 (swap_selector_for_mode): Delete.
1023 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1024 altivec_expand_stvex_be): Delete.
1025 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1026 -maltivec=be support.
1027 (rs6000_gimple_fold_builtin): Ditto.
1028 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1030 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1031 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1032 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1033 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1034 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1035 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1036 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1037 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1038 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1039 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1040 anonymous split): Adjust.
1041 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1042 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1044 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1046 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1047 when --with-gxx-include-dir is also specified.
1048 * configure: Regenerate.
1050 2018-05-09 Jim Wilson <jimw@sifive.com>
1053 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1054 * config/riscv/t-withmultilib: New.
1055 * config/riscv/withmultilib.h: New.
1056 * doc/install.texi: Document RISC-V --with-multilib-list support.
1058 2018-05-09 Richard Biener <rguenther@suse.de>
1060 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1062 (vect_bb_vectorization_profitable_p): Adjust. Compute
1063 actual scalar cost using the cost vector and the add_stmt_cost
1066 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1068 PR rtl-optimization/85645
1069 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1070 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1072 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1074 PR rtl-optimization/85645
1075 * shrink-wrap.c (spread_components): Return a boolean saying if
1076 anything was changed.
1077 (try_shrink_wrapping_separate): Iterate spread_components until
1078 nothing changes anymore.
1080 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1082 PR rtl-optimization/85645
1083 * regrename.c (build_def_use): Also kill the chains that include the
1084 destination of a REG_CFA_REGISTER note.
1086 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1088 PR rtl-optimization/85645
1089 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1090 insn that has a REG_CFA_REGISTER note.
1092 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1094 * cfgexpand.c (expand_clobber): New function.
1095 (expand_gimple_stmt_1): Use it.
1096 * tree-vect-stmts.c (vect_clobber_variable): New function,
1098 (vectorizable_simd_clone_call): ...here.
1099 (vectorizable_store): Emit a clobber either side of an
1100 IFN_STORE_LANES sequence.
1101 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1103 2018-05-09 Tom de Vries <tom@codesourcery.com>
1106 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1107 (define_insn "trap_if_false"): Add exit after trap.
1109 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1111 PR rtl-optimization/85638
1112 * bb-reorder.c: Include common/common-target.h.
1113 (create_forwarder_block): New function extracted from...
1114 (fix_up_crossing_landing_pad): ...here. Rename into...
1115 (dw2_fix_up_crossing_landing_pad): ...this.
1116 (sjlj_fix_up_crossing_landing_pad): New function.
1117 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1118 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1119 from both partitions and exit the loop after one iteration.
1121 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1124 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1126 (Basic PowerPC Built-in Functions): The new name of the
1127 subsection previously known as "PowerPC Built-in Functions".
1128 (Basic PowerPC Built-in Functions Available on all Configurations):
1130 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1132 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1134 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1136 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1139 2018-05-08 Jim Wilson <jimw@sifive.com>
1141 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1142 (LD_EMUL_SUFFIX): New.
1143 (LINK_SPEC): Use it.
1145 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1147 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1149 (Basic PowerPC Built-in Functions): The new name of the
1150 subsection previously known as "PowerPC Built-in Functions".
1151 (Basic PowerPC Built-in Functions Available on all Configurations):
1153 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1155 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1157 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1159 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1162 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1165 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1166 after cmpelim optimization.
1168 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1170 * config.gcc: Support "goldmont".
1171 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1172 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1174 * config/i386/i386.c (m_GOLDMONT): Define.
1175 (processor_target_table): Add "goldmont".
1176 (PTA_GOLDMONT): Define.
1177 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1178 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1179 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1180 (fold_builtin_cpu): Add "goldmont".
1181 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1182 (ix86_option_override_internal): Add "goldmont".
1183 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1184 (processor_type): Add PROCESSOR_GOLDMONT.
1185 * config/i386/i386.md: Add CPU "glm".
1186 * config/i386/glm.md: New file.
1187 * config/i386/x86-tune.def: Add m_GOLDMONT.
1188 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1190 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1193 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1195 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1196 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1197 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1200 * config/i386/i386.c (ix86_fold_builtin): Handle
1201 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1204 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1205 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1207 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1210 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1211 (check_arch): Likewise.
1212 (check_fpu): Return the result rather than printing it.
1213 (end arch): Fix operator precedence.
1214 (end cpu): Likewise.
1215 (END): Print the result from check_fpu.
1217 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1218 Alan Hayward <alan.hayward@arm.com>
1219 David Sherwood <david.sherwood@arm.com>
1221 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1222 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1223 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1224 (*fcmuo<mode>_and): New patterns.
1226 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1228 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1229 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1230 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1231 (cmp_op, sve_imm_con): New code attributes.
1232 (SVE_COND_INT_CMP, imm_con): Delete.
1233 (cmp_op): Remove above unspecs from int attribute.
1234 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1236 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1237 comparison-specific unspecs.
1238 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1239 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1240 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1241 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1242 (*vec_fcm<cmp_op><mode>): Rename to...
1243 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1244 (*vec_fcmuo<mode>): Rename to...
1245 (*fcmuo<mode>): ...this and adjust likewise.
1246 (*pred_fcm<cmp_op><mode>): New pattern.
1247 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1248 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1250 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1252 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1253 (aarch64_emit_sve_predicated_cond): New function.
1254 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1255 (aarch64_emit_unspec_cond_or): Replace with...
1256 (aarch64_emit_sve_or_conds): ...this new function. Use
1257 aarch64_emit_sve_ptrue_op for the individual comparisons and
1258 aarch64_emit_binop to OR them together.
1259 (aarch64_emit_inverted_unspec_cond): Replace with...
1260 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1261 aarch64_emit_sve_ptrue_op for the comparison and
1262 aarch64_emit_unop to invert the result.
1263 (aarch64_expand_sve_vec_cmp_float): Update after the above
1264 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1266 2018-05-07 Nathan Sidwell <nathan@acm.org>
1268 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1269 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1270 (Backwards Compatibility): Likewise.
1272 2018-05-07 Luis Machado <luis.machado@linaro.org>
1276 2018-05-07 Luis Machado <luis.machado@linaro.org>
1278 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1279 <prefetch_dynamic_strides>: New const bool field.
1280 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1281 prefetch_dynamic_strides.
1282 (exynosm1_prefetch_tune): Likewise.
1283 (thunderxt88_prefetch_tune): Likewise.
1284 (thunderx_prefetch_tune): Likewise.
1285 (thunderx2t99_prefetch_tune): Likewise.
1286 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1287 (aarch64_override_options_internal): Update to set
1288 PARAM_PREFETCH_DYNAMIC_STRIDES.
1289 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1290 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1291 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1292 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1293 prefetch-dynamic-strides setting.
1295 2018-05-07 Luis Machado <luis.machado@linaro.org>
1297 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1298 <minimum_stride>: New const int field.
1299 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1300 minimum_stride field.
1301 (exynosm1_prefetch_tune): Likewise.
1302 (thunderxt88_prefetch_tune): Likewise.
1303 (thunderx_prefetch_tune): Likewise.
1304 (thunderx2t99_prefetch_tune): Likewise.
1305 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1306 (aarch64_override_options_internal): Update to set
1307 PARAM_PREFETCH_MINIMUM_STRIDE.
1308 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1309 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1310 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1311 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1312 stride is constant and is below the minimum stride threshold.
1314 2018-05-07 Luis Machado <luis.machado@linaro.org>
1316 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1319 2018-05-07 Luis Machado <luis.machado@linaro.org>
1321 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1322 <prefetch_dynamic_strides>: New const bool field.
1323 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1324 prefetch_dynamic_strides.
1325 (exynosm1_prefetch_tune): Likewise.
1326 (thunderxt88_prefetch_tune): Likewise.
1327 (thunderx_prefetch_tune): Likewise.
1328 (thunderx2t99_prefetch_tune): Likewise.
1329 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1330 (aarch64_override_options_internal): Update to set
1331 PARAM_PREFETCH_DYNAMIC_STRIDES.
1332 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1333 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1334 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1335 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1336 prefetch-dynamic-strides setting.
1338 2018-05-07 Luis Machado <luis.machado@linaro.org>
1340 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1341 <minimum_stride>: New const int field.
1342 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1343 minimum_stride field.
1344 (exynosm1_prefetch_tune): Likewise.
1345 (thunderxt88_prefetch_tune): Likewise.
1346 (thunderx_prefetch_tune): Likewise.
1347 (thunderx2t99_prefetch_tune): Likewise.
1348 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1349 (aarch64_override_options_internal): Update to set
1350 PARAM_PREFETCH_MINIMUM_STRIDE.
1351 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1352 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1353 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1354 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1355 stride is constant and is below the minimum stride threshold.
1357 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1360 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1361 the type is addressable. Don't force op into register if it has
1364 2018-05-05 Roland McGrath <mcgrathr@google.com>
1367 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1368 any section for which we don't know a specific type it should have,
1369 regardless of name. Previously this was done only for the exact
1370 names ".init_array", ".fini_array", and ".preinit_array".
1371 (default_elf_asm_named_section): Add comment about
1372 relationship with default_section_type_flags and SECTION_NOTYPE.
1373 (get_section): Don't consider it a type conflict if one side has
1374 SECTION_NOTYPE and the other doesn't, as long as neither has the
1375 SECTION_BSS et al used in the default_section_type_flags logic.
1377 2018-05-05 Tom de Vries <tom@codesourcery.com>
1380 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1381 (workaround_barsyncs): New function.
1382 (nvptx_reorg): Use workaround_barsyncs.
1383 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1384 (define_expand "nvptx_membar_cta"): New define_expand.
1385 (define_insn "*nvptx_membar_cta"): New insn.
1387 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1389 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1390 To improve optimization opportunities.
1391 * builtin-types.def: The new needed builtin types for the above.
1393 2018-05-04 Richard Biener <rguenther@suse.de>
1395 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1396 * gimple-ssa-store-merging.c
1397 (imm_store_chain_info::output_merged_store): Remove redundant create,
1398 release split_store vector contents on failure.
1399 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1400 scalar stmt vector on cache hit.
1402 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1404 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1406 * config.gcc (powerpc-xilinx-eabi*): Remove.
1407 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1409 (fusion_addis_mem_combo_load): Ditto.
1410 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1412 (rs6000_cpu_cpp_builtins): Ditto.
1413 * config/rs6000/rs6000-linux.c
1414 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1415 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1416 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1418 (rs6000_setup_reg_addr_masks): Ditto.
1419 (rs6000_init_hard_regno_mode_ok): Ditto.
1420 (rs6000_option_override_internal): Ditto.
1421 (legitimate_lo_sum_address_p): Ditto.
1422 (rs6000_legitimize_address): Ditto.
1423 (rs6000_legitimize_reload_address): Ditto.
1424 (rs6000_legitimate_address_p): Ditto.
1425 (abi_v4_pass_in_fpr): Ditto.
1426 (setup_incoming_varargs): Ditto.
1427 (rs6000_gimplify_va_arg): Ditto.
1428 (rs6000_split_multireg_move): Ditto.
1429 (rs6000_savres_strategy): Ditto.
1430 (rs6000_emit_prologue_components): Ditto.
1431 (rs6000_emit_epilogue_components): Ditto.
1432 (rs6000_emit_prologue): Ditto.
1433 (rs6000_emit_epilogue): Ditto.
1434 (rs6000_elf_file_end): Ditto.
1435 (rs6000_function_value): Ditto.
1436 (rs6000_libcall_value): Ditto.
1437 * config/rs6000/rs6000.h: Ditto.
1438 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1439 (TARGET_MINMAX): ... this. New.
1440 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1441 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1442 (*movsi_internal1_single): Delete.
1443 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1444 mfpu=, mxilinx-fpu): Delete.
1445 * config/rs6000/singlefp.h: Delete.
1446 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1447 * config/rs6000/t-rs6000: Ditto.
1448 * config/rs6000/t-xilinx: Delete.
1449 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1450 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1451 (VStype_simple): Delete.
1452 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1453 * config/rs6000/xfpu.h: Delete.
1454 * config/rs6000/xfpu.md: Delete.
1455 * config/rs6000/xilinx.h: Delete.
1456 * config/rs6000/xilinx.opt: Delete.
1457 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1458 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1460 2018-05-04 Tom de Vries <tom@codesourcery.com>
1463 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1466 2018-05-04 Richard Biener <rguenther@suse.de>
1469 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1470 (expand_complex_div_wide): Likewise.
1471 (expand_complex_operations_1): Likewise.
1472 (expand_complex_libcall): Preserve EH info of the original stmt.
1473 (tree_lower_complex): Handle removed blocks.
1474 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1475 on complex multiplication and division libcall builtins.
1477 2018-05-04 Richard Biener <rguenther@suse.de>
1480 * fold-const.c (negate_expr_p): Restrict negation of operand
1481 zero of a division to when we know that can happen without
1483 (fold_negate_expr_1): Likewise.
1485 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1488 * real.h (real_nextafter): Declare.
1489 * real.c (real_nextafter): New function.
1490 * fold-const-call.c (fold_const_nextafter): New function.
1491 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1492 CASE_CFN_NEXTTOWARD.
1493 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1494 even when arg1_mode is different from arg0_mode.
1496 2018-05-03 Nathan Sidwell <nathan@acm.org>
1498 * doc/extend.texi (Deprecated Features): Remove
1500 (Backwards Compatibility): Likewise.
1501 * doc/invoke.texi (C++ Language Options): Likewise.
1502 (C++ Dialect Options): Likewise.
1504 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1507 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1508 _mm512_mask_mullox_epi64): New intrinsics.
1510 2018-05-03 Tom de Vries <tom@codesourcery.com>
1513 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1514 dump files): Add offload-tree.
1516 2018-05-03 Richard Biener <rguenther@suse.de>
1518 PR tree-optimization/85615
1519 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1520 to loops not nested in BBs loop father to avoid creating multi-entry
1523 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1525 PR tree-optimization/70291
1526 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1527 arguments. Change return type to tree. Emit libcall as a new
1528 statement rather than replacing existing one when inplace_p is true.
1529 (expand_complex_multiplication_components): New function.
1530 (expand_complex_multiplication): Expand floating-point complex
1531 multiplication using the above.
1532 (expand_complex_division): Rename inner_type parameter to type.
1533 Update expand_complex_libcall call-site.
1534 (expand_complex_operations_1): Update expand_complex_multiplication
1535 and expand_complex_division call-sites.
1537 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1540 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1541 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1542 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1543 the highest significant bit of the shift count mask is clear. In
1544 check whether and[sq]i3 is needed verify that all significant bits
1545 of the shift count other than the highest are set.
1547 2018-05-02 Tom de Vries <tom@codesourcery.com>
1550 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1551 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1552 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1553 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1554 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1555 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1556 __builtin_goacc_parlevel_size.
1558 2018-05-02 Richard Biener <rguenther@suse.de>
1560 PR tree-optimization/85597
1561 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1562 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1565 2018-05-02 Tom de Vries <tom@codesourcery.com>
1568 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1569 dump files): Add ltrans-tree.
1571 2018-05-02 Tom de Vries <tom@codesourcery.com>
1574 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1575 dump files): Add wpa-ipa.
1577 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1579 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1580 powerpc*-*-linux*paired* target.
1581 * config/rs6000/750cl.h: Delete.
1582 * config/rs6000/paired.h: Delete.
1583 * config/rs6000/paired.md: Delete.
1584 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1586 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1587 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1588 comment. Remove paired float support.
1589 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1590 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1592 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1593 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1595 * config/rs6000/rs6000.c: Remove paired float support.
1596 (paired_expand_vector_init, paired_expand_vector_move,
1597 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1598 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1599 paired_expand_builtin, paired_expand_predicate_builtin,
1600 paired_init_builtins): Delete.
1601 * config/rs6000/rs6000.h: Remove paired float support.
1602 * config/rs6000/rs6000.md: Remove paired float support.
1603 (move_from_CR_ov_bit): Delete.
1604 * config/rs6000/rs6000.opt (mpaired): Delete.
1605 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1606 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1608 2018-05-02 Richard Biener <rguenther@suse.de>
1611 * gimplify.c (gimplify_save_expr): When in SSA form allow
1612 SAVE_EXPRs to compute to SSA vars.
1614 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1617 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1618 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1619 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1620 clobber operands[2], instead use a new pseudo. Formatting fixes.
1622 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1624 PR tree-optimization/85586
1625 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1626 exit early for statements in the same group if the accesses are
1629 2018-05-02 Tom de Vries <tom@codesourcery.com>
1632 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1635 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1637 PR tree-optimization/85143
1638 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1640 2018-05-01 Tom de Vries <tom@codesourcery.com>
1643 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1644 not found" error message.
1646 2018-05-01 Tom de Vries <tom@codesourcery.com>
1649 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1650 * vec.c (test_ordered_remove_if): New function.
1651 (vec_c_tests): Call test_ordered_remove_if.
1652 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1653 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1654 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1655 VEC_ORDERED_REMOVE_IF.
1657 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1659 PR tree-optimization/82665
1660 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1661 pointer subtraction where arguments come from a memchr call.
1663 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1665 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1666 --push-state --as-needed and --pop-state instead of --as-needed and
1667 --no-as-needed if ld supports it.
1668 * configure: Regenerated.
1671 * doc/install.texi2html: Replace _002d with - and _002a with * in
1672 generated html files using sed.
1674 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1677 * gcc-rich-location.c (blank_line_before_p): New function.
1678 (use_new_line): New function.
1679 (gcc_rich_location::add_fixit_insert_formatted): New function.
1680 * gcc-rich-location.h
1681 (gcc_rich_location::add_fixit_insert_formatted): New function.
1683 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1685 * selftest.c (assert_streq): Rename "expected" and "actual" to
1686 "val1" and "val2". Extend NULL-handling to cover both inputs
1687 symmetrically, while still requiring both to be non-NULL for a pass.
1688 * selftest.h (assert_streq): Rename "expected" and "actual" to
1690 (ASSERT_EQ): Likewise.
1691 (ASSERT_EQ_AT): Likewise.
1692 (ASSERT_KNOWN_EQ): Likewise.
1693 (ASSERT_KNOWN_EQ_AT): Likewise.
1694 (ASSERT_NE): Likewise.
1695 (ASSERT_MAYBE_NE): Likewise.
1696 (ASSERT_MAYBE_NE_AT): Likewise.
1697 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1698 the assertion to pass.
1699 (ASSERT_STREQ_AT): Likewise.
1701 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1703 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1704 interaction with -pie.
1706 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1708 * selftest.h: Fix alphabetization of per-source-file selftest
1711 2018-04-30 Jason Merrill <jason@redhat.com>
1713 PR c++/61982 - dead stores to destroyed objects.
1714 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1717 2018-04-30 Jason Merrill <jason@redhat.com>
1719 * tree.c (build_clobber): New.
1720 * tree.h: Declare it.
1721 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1723 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1725 * diagnostic-show-locus.c (layout::layout): Update for
1726 location_get_source_line returning a char_span.
1727 (struct char_span): Move to input.h.
1728 (struct correction): Update for fields in char_span becoming
1730 (struct source_line): Update for location_get_source_line
1731 returning a char_span.
1732 (layout::print_line): Likewise.
1733 * edit-context.c (edited_file::print_content): Likewise.
1734 (edited_file::print_diff_hunk): Likewise.
1735 (edited_file::print_run_of_changed_lines): Likewise.
1736 (edited_file::get_num_lines): Likewise.
1737 (edited_line::edited_line): Likewise.
1738 * final.c (asm_show_source): Likewise.
1739 * input.c (location_get_source_line): Convert return type
1740 from const char * to char_span, losing the final "line_len"
1742 (dump_location_info): Update for the above.
1743 (get_substring_ranges_for_loc): Likewise. Use a char_span
1744 when handling the literal within the line.
1745 (test_reading_source_line): Update for location_get_source_line
1746 returning a char_span.
1747 * input.h (class char_span): Move here from
1748 diagnostic-show-locus.c, converting from a struct to a class.
1749 Make data members private.
1750 (char_span::operator bool): New.
1751 (char_span::length): New.
1752 (char_span::get_buffer): New.
1753 (char_span::operator[]): New.
1754 (char_span::subspan): Make const.
1755 (char_span::xstrdup): New.
1756 (location_get_source_line): Convert return type from const char *
1757 to char_span, losing the final "line_size" param.
1759 2018-04-30 Jan Hubicka <jh@suse.cz>
1761 * lto-wrapper.c (ltrans_priorities): New static var.
1762 (cmp_priority): New.
1763 (run_gcc): Read priorities and if doing parallel build order
1764 the Makefile by them.
1766 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1768 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1770 2018-04-30 Richard Biener <rguenther@suse.de>
1772 * tree-cfg.c (verify_address): Remove base argument, add
1773 flag whether to check TREE_ADDRESSABLE and do that.
1774 (verify_expr): Remove.
1775 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1776 (verify_gimple_assign_single): Likewise.
1777 (verify_gimple_switch): Likewise.
1778 (verify_expr_location_1): Dereference tp once. Add (disabled)
1779 piece from verify_expr.
1780 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1782 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1784 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1786 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1788 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1789 (small_data_pattern): Likewise.
1790 (arc_rewrite_small_data): Likewise.
1791 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1792 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1793 (get_symbol_alignment): New function.
1794 (legitimate_small_data_address_p): Likewise.
1795 (legitimate_scaled_address): Update, call
1796 legitimate_small_data_address_p.
1797 (output_sdata): New static variable.
1798 (arc_print_operand): Update how we handle small data operands.
1799 (arc_print_operand_address): Likewise.
1800 (arc_legitimate_address_p): Update, use
1801 legitimate_small_data_address_p.
1802 (arc_rewrite_small_data_p): Remove.
1803 (arc_rewrite_small_data_1): Likewise.
1804 (arc_rewrite_small_data): Likewise.
1805 (small_data_pattern): Likewise.
1806 (compact_sda_memory_operand): Update to use
1807 legitimate_small_data_address_p and get_symbol_alignment.
1808 (prepare_move_operands): Don't rewite sdata pattern.
1809 (prepare_extend_operands): Remove.
1810 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1812 (zero_extendqisi2): Likewise.
1813 (zero_extendhisi2): Likewise.
1814 (extendqihi2): Likewise.
1815 (extendqisi2): Likewise.
1816 (extendhisi2): Likewise.
1820 * config/arc/constraints.md (Usd): Change it to memory constraint.
1822 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1824 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1825 as source of std instructions.
1826 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1827 allow 6-bit constants as source for store instructions.
1828 (movdi_insn): Update instruction pattern to allow 6-bit constants
1829 as source for store instructions.
1831 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1833 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1835 2018-04-30 Nathan Sidwell <nathan@acm.org>
1836 Sandra Loosemore <sandra@codesourcery.com>
1838 * dumpfile.c (dump_open): Allow '-' for stdout.
1839 * doc/invoke.texi (Developer Options): Document dump filename
1840 determination early. Document stdin/stdout selection.
1842 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1844 Microblaze Target: PIC data text relative
1846 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1847 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1849 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1850 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1851 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1852 New addressing mode for data-text relative position indepenedent code.
1853 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1854 'ADDRESS_SYMBOLIC_TXT_REL'.
1855 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1856 (microblaze_legitimate_pic_operand): Exclude function calls from
1857 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1858 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1860 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1861 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1862 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1863 for 'address + offset'.
1864 (microblaze_expand_prologue): Add new function prologue call for
1866 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1867 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1868 table in case of TARGET_PIC_DATA_TEXT_REL.
1869 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1870 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1871 Add new macros 'UNSPEC_TEXT',
1872 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1873 + exclude function calls from 'UNSPEC_PLT' in case of data text
1875 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1876 new target hook for generating address diff vector tables in case of
1878 * doc/tm.texi : Regenerate.
1879 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1880 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1881 of addr diff vector generation.
1882 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1883 target hook definition.
1884 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1885 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1886 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1887 option for data text relative.
1889 2018-04-30 Richard Biener <rguenther@suse.de>
1891 * tree-chrec.h (evolution_function_is_constant_p): Remove
1893 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1895 2018-04-30 Richard Biener <rguenther@suse.de>
1898 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1900 2018-04-30 Richard Biener <rguenther@suse.de>
1902 PR tree-optimization/28364
1903 PR tree-optimization/85275
1904 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1905 copying first exit test.
1907 2018-04-28 Mark Wielaard <mark@klomp.org>
1909 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1911 (dwarf_AT): Handle DW_AT_addr_base.
1912 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1914 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1917 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1918 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1919 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1920 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1922 2018-04-28 Richard Biener <rguenther@suse.de>
1924 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1925 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1926 to reflect use. Only add interesting stmts.
1928 2018-04-27 Martin Jambor <mjambor@suse.cz>
1931 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1932 the jump function allows for passing through aggregate values.
1934 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1936 * input.h (in_system_header_at): Convert from macro to inline
1938 (from_macro_expansion_at): Likewise.
1939 (from_macro_definition_at): Likewise.
1941 2018-04-27 Jeff Law <law@redhat.com>
1943 * config.gcc: Mark tile* targets as deprecated/obsolete.
1945 2018-04-27 Richard Biener <rguenther@suse.de>
1947 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1950 2018-04-27 Richard Biener <rguenther@suse.de>
1952 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1954 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1956 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1957 with Yd constraint. Set "preferred_for_speed" attribute from
1958 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1960 (*movdi_internal): Ditto.
1961 (movti_interunit splitters): Remove
1962 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1963 (movdi_interunit splitters): Ditto.
1964 * config/i386/constraints.md (Ye): Remove.
1965 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1967 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1970 * config/aarch64/constraints.md (Usg): Limit to 31.
1973 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1975 PR tree-optimization/85529
1976 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1977 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1978 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1979 zero extension or masking of the MSB bit.
1980 (optimize_range_tests): Add FIRST_BB argument, pass it through
1981 to optimize_range_tests_var_bound.
1982 (maybe_optimize_range_tests, reassociate_bb): Adjust
1983 optimize_range_tests callers.
1985 2018-04-26 Richard Biener <rguenther@suse.de>
1986 Jakub Jelinek <jakub@redhat.com>
1988 * cgraph.h (symbol_table): Just declare debug method here.
1989 * symtab.c (symbol_table::debug): Define.
1991 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1993 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1995 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1997 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1998 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1999 (*movdi_internal): Substitute Yi and Yj constraint with x
2000 and Ym and Yn constraint with y constraint. Update "isa"
2001 attribute and set "preferred_for_speed" attribute from
2002 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2003 (*movsi_internal): Ditto.
2004 (*movdf_internal): Ditto.
2005 (*movsf_internal): Ditto.
2006 (*zero_extendsidi2): Ditto.
2007 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2008 (sse2_loadld): Ditto.
2009 (*vec_extract<ssevecmodelower>_0): Ditto.
2010 (*vec_extractv4si_0_zext_sse4): Ditto.
2011 (vec_concatv2di): Ditto.
2012 (*vec_dup<mode>): Ditto.
2013 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2014 * config/i386/constraints.md (Yi): Remove.
2019 2018-04-26 Nathan Sidwell <nathan@acm.org>
2021 * dumpfile.c (dump_open): New.
2022 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2023 (dump_finish): Detect stdio/stderr by value not name.
2025 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2027 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2029 2018-04-26 Tom de Vries <tom@codesourcery.com>
2032 * config/nvptx/nvptx.c (verify_neutering_jumps)
2033 (verify_neutering_labels): New function
2034 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2036 2018-04-26 Tom de Vries <tom@codesourcery.com>
2039 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2040 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2043 2018-04-26 Richard Biener <rguenther@suse.de>
2044 Tom de Vries <tom@codesourcery.com>
2047 * lto-streamer-out.c (output_function): Fixup loops if required to match
2048 discovery done in the reader.
2050 2018-04-26 Richard Biener <rguenther@suse.de>
2052 PR tree-optimization/85116
2053 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2054 have a loop exit from the single latch predecessor. Remove
2055 case of header with just condition.
2056 (ch_base::copy_headers): Exclude infinite loops from any
2058 (pass_ch::execute): Record exits.
2060 2018-04-26 Richard Biener <rguenther@suse.de>
2062 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2063 prologue cost vector and pass it to vect_get_load_cost.
2064 (vect_get_peeling_costs_all_drs): Likewise.
2065 (vect_peeling_hash_get_lowest_cost): Likewise.
2066 (vect_enhance_data_refs_alignment): Likewise.
2068 2018-04-26 Richard Biener <rguenther@suse.de>
2071 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2072 checking of integer<->pointer conversions.
2073 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2074 sign-/zero-extending pointer types.
2075 (expand_omp_for_static_chunk): Likewise.
2077 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2078 Jean Lee <xiaoyur347@gmail.com>
2080 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2081 (TARGET_ASAN_SHADOW_OFFSET): Define.
2082 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2083 true for -fsanitize=address.
2085 2018-04-25 Mark Wielaard <mark@klomp.org>
2087 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2090 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2092 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2093 than "alu", remove explicit "memory" and "imm_disp" attributes.
2094 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2097 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2098 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2099 gen_lowpart_no_emit.
2101 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2104 * config/i386/i386.c (ix86_expand_builtin): Change memory
2105 operand to XI, extend p0 to Pmode.
2106 * config/i386/i386.md: Change unspec volatile and operand
2107 1 mode to XI, change operand 0 mode to P.
2109 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2111 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2112 GET_MODE_MASK before any checking.
2113 (nds32_can_use_bset_p): Likewise.
2114 (nds32_can_use_btgl_p): Likewise.
2116 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2118 * config/nds32/nds32-doubleword.md: New define_split pattern for
2119 illegal register number.
2121 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2123 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2125 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2127 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2129 2018-04-25 Richard Biener <rguenther@suse.de>
2131 * lto-streamer.h (LTO_major_version): Bump to 8.
2133 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2135 * BASE-VER: Set to 9.0.0.
2137 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2139 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2140 in __abskf2 and __powikf2.
2142 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2145 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2146 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2147 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2148 Use the above on operand 2. Reindent.
2149 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2151 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2154 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2156 * config/i386/i386.opt (mcet): Removed.
2157 * doc/install.texi: Remove -mcet documentation.
2158 * doc/invoke.texi: Likewise.
2160 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2163 * doc/install.texi: Remove -mcet from bootstrap-cet.
2165 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2168 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2169 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2173 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2174 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2175 containing a CONST_VECTOR.
2177 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2179 * doc/install.texi: Update newlib dependency for nvptx.
2181 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2184 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2185 instead of INTVAL when shifting x left.
2187 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2189 PR tree-optimization/85478
2190 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2191 vect_grouped_store_supported for single element vectors.
2193 2018-04-24 Richard Biener <rguenther@suse.de>
2196 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2197 load cost increase to the case of non-constant step.
2199 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2202 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2203 destination if any_malformed_asm.
2205 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2208 * expr.c (store_field): In the bitfield case, if the value comes from
2209 a function call and is returned in registers by means of a PARALLEL,
2210 do not change the mode of the temporary unless BLKmode and VOIDmode.
2212 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2214 PR rtl-optimization/85423
2215 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2216 dependencies to debug insns when the previous insn is non-debug.
2218 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2220 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2221 enums into a single definition.
2222 (fls): Fix predicates and printing.
2225 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2227 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2228 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2229 and short u6 immediate.
2230 (check_if_valid_sleep_operand): Remove.
2231 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2233 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2235 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2236 flag_always_save_lp condition.
2237 * config/nds32/nds32.opt (malways-save-lp): New option.
2239 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2241 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2242 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2243 * config/nds32/nds32.h
2244 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2245 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2247 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2249 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2250 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2252 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2253 Chung-Ju Wu <jasonwucj@gmail.com>
2255 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2257 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2258 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2260 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2262 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2264 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2266 * config/nds32/nds32-protos.h (nds32_data_alignment,
2267 nds32_local_alignment): Declare.
2268 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2269 nds32_local_alignment): New functions.
2270 (TARGET_CONSTANT_ALIGNMENT): Define.
2271 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2273 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2275 * config/nds32/nds32.c
2276 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2277 (TARGET_MODES_TIEABLE_P): Likewise.
2279 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2281 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2284 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2285 Chung-Ju Wu <jasonwucj@gmail.com>
2287 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2288 for unaligned access.
2289 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2291 * config/nds32/nds32-intrinsic.md: Likewise.
2292 * config/nds32/nds32_intrinsic.h: Likewise.
2293 * config/nds32/nds32.h (nds32_builtins): Likewise.
2294 * config/nds32/nds32.opt (munaligned-access): New option.
2295 * config/nds32/nds32.c (nds32_asm_file_start): Display
2296 flag_unaligned_access status.
2298 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2300 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2301 -mno-relax is present.
2302 * config/riscv/linux.h (LINK_SPEC): Ditto.
2304 2018-04-20 Martin Sebor <msebor@redhat.com>
2307 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2309 (gimple_fold_builtin_stxcpy_chk): Same.
2310 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2312 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2315 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2316 __powikf2 when long double is IEEE 128-bit.
2318 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2320 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2321 step to make sure stack always aligned.
2323 2018-04-20 Carl Love <cel@us.ibm.com>
2326 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2327 size check for arg0.
2329 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2330 Tom de Vries <tom@codesourcery.com>
2333 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2334 Emit insns for calls too.
2335 (nvptx_find_par): Always look for worker-level predecessor insn.
2336 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2338 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2339 (nvptx_process_pars): Propagate frames for calls.
2341 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2344 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2346 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2347 (ix86_handle_option): Don't handle OPT_mibt.
2348 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2350 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2352 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2353 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2354 (ix86_target_macros): Define __CET__ with flag_cf_protection
2355 for -fcf-protection.
2356 * config/i386/i386.c (isa2_opts): Remove -mibt.
2357 * config/i386/i386.h (TARGET_IBT): Removed.
2358 (TARGET_IBT_P): Likewise.
2359 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2360 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2361 * config/i386/i386.opt (mcet): Update help message.
2364 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2365 -mcet as an alias for -mshstk.
2367 2018-04-20 Richard Biener <rguenther@suse.de>
2370 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2371 complexity by forcing a single use of the multiply operand.
2373 2018-04-20 Martin Jambor <mjambor@suse.cz>
2376 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2377 recursion dependency to only apply to non-clones.
2379 2018-04-20 Martin Jambor <mjambor@suse.cz>
2382 * ipa-cp.c (create_specialized_node): Check that clones of
2383 self-recursive edges exist during IPA-CP.
2385 2018-04-19 Toon Moene <toon@moene.org>
2387 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2390 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2392 PR tree-optimization/85467
2393 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2394 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2395 VECTOR_CST element to type.
2397 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2400 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2401 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2402 (builtin_longjmp): Likewise.
2403 (save_stack_nonlocal): New pattern.
2404 (restore_stack_nonlocal): Likewise.
2406 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2409 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2410 Replace ASM_OUTPUT_LABEL with fprintf.
2412 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2415 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2416 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2417 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2418 define __IBT__ and __SHSTK__ for -fcf-protection.
2419 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2421 (ix86_trampoline_init): Likewise.
2422 (x86_output_mi_thunk): Likewise.
2423 (ix86_notrack_prefixed_insn_p): Likewise.
2424 (ix86_option_override_internal): Don't disallow -fcf-protection.
2425 * config/i386/i386.md (rdssp<mode>): Also enable for
2427 (incssp<mode>): Likewise.
2428 (nop_endbr): Likewise.
2429 * config/i386/i386.opt (mcet): Change help message to built-in
2433 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2434 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2435 enable CET built-in functions.
2437 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2439 * common/config/i386/i386-common.c
2440 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2441 OPTION_MASK_ISA_MOVDIRI_UNSET,
2442 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2443 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2444 * config.gcc (movdirintrin.h): New header.
2445 * config/i386/cpuid.h (bit_MOVDIRI,
2446 bit_MOVDIR64B): New bits.
2447 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2449 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2450 (VOID, PVOID, PCVOID)): New function types.
2451 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2452 __builtin_ia32_directstoreu_u64,
2453 __builtin_ia32_movdir64b): New builtins.
2454 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2455 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2457 (ix86_valid_target_attribute_inner_p): Ditto.
2458 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2459 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2460 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2461 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2462 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2463 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2464 (movdiri<mode>, movdir64b_<mode>): New.
2465 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2466 * config/i386/immintrin.h: Include movdirintrin.h.
2467 * config/i386/movdirintrin.h: New file.
2468 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2470 2018-04-19 Richard Biener <rguenther@suse.de>
2473 * cfg.c (clear_bb_flags): When loop state says we have
2474 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2476 2018-04-19 Richard Biener <rguenther@suse.de>
2478 PR tree-optimization/84737
2479 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2480 copying restrict info.
2481 (vect_setup_realignment): Use it.
2482 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2483 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2484 the first DR to all generated stores.
2485 (vectorizable_load): Likewise for loads.
2487 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2489 PR tree-optimization/85446
2490 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2491 the integral and pointer types to have the same precision.
2493 * doc/install.texi: Document --disable-cet being the default and
2496 2018-04-18 Martin Liska <mliska@suse.cz>
2498 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2501 2018-04-18 Martin Liska <mliska@suse.cz>
2504 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2507 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2508 arguments if they are comparable.
2510 2018-04-18 Martin Liska <mliska@suse.cz>
2513 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2516 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2519 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2522 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2523 ENDBR after calling __morestack.
2525 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2528 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2529 by using gcc_base_ver to generate a gcc_driver_version, and use
2530 it when generating GCC_DRIVER_NAME.
2531 * configure: Regenerate.
2533 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2536 * config.gcc: Obsolete powerpc*-*-*spe*.
2538 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2541 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2542 (stabstr_D): Change type of unum from unsigned int to
2543 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2546 2018-04-17 Jim Wilson <jimw@sifive.com>
2549 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2550 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2551 Set arg_pointer_offset after using pretend_args_size.
2553 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2555 PR rtl-optimization/85431
2556 * dse.c (record_store): Ignore zero width stores.
2559 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2560 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2561 __builtin_stack_restore rather than after it.
2562 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2563 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2564 argument instead of virtual_dynamic_stack_rtx.
2566 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2568 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2570 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2571 Add note to error message to explain internal mapping of overloaded
2572 built-in function name to non-overloaded built-in function name.
2573 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2576 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2579 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2580 where the inputs overlap with the output.
2582 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2584 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2585 (=v, v) alternative and explicit "memory" attribute.
2586 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2587 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2589 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2590 "sselog1" type instead of "sselog".
2591 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2592 "sselog". Remove explicit "memory" attribute.
2593 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2594 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2596 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2597 "sselog1" type instead of "sselog". Remove explicit "memory"
2599 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2600 use "sselog1" type instead of "sselog". Remove explicit "memory"
2602 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2603 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2605 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2606 "sselog1" type instead of "sselog". Remove explicit "memory"
2608 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2609 use "sselog1" type instead of "sselog". Remove explicit "memory"
2613 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2616 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2619 2018-04-17 Martin Jambor <mjambor@suse.cz>
2622 * ipa-cp.c (create_specialized_node): Call
2623 expand_all_artificial_thunks if necessary.
2625 2018-04-17 Martin Liska <mliska@suse.cz>
2628 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2629 in message, remote space in between '_G' and '('.
2631 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2634 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2635 avx512f_vmcmp<mode>3<round_saeonly_name>,
2636 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2637 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2638 avx512f_rndscale<mode><round_saeonly_name>,
2639 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2640 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2641 Use %<iptr>2 instead of %2 for -masm=intel.
2642 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2643 avx512f_vcvttss2usi<round_saeonly_name>,
2644 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2646 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2647 avx512f_vcvttsd2usi<round_saeonly_name>,
2648 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2649 Use %q1 instead of %1 for -masm=intel.
2650 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2651 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2652 of %3 for -masm=intel.
2653 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2655 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2657 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2659 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2660 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2662 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2664 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2665 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2666 %g1 and one with %0 and %1.
2667 (avx512er_vmrcp28<mode><round_saeonly_name>,
2668 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2670 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2671 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2672 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2673 of %0 and %{%4%} for -masm=intel.
2674 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2675 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2676 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2677 order of %0 and %{%5%}%{z%} for -masm=intel.
2679 2018-04-17 Jan Hubicka <jh@suse.cz>
2682 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2684 2018-04-17 Martin Liska <mliska@suse.cz>
2687 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2688 for target_clone error message. Make default implementation
2689 clone to be a local declaration.
2690 (separate_attrs): Add new argument and check for an empty
2692 (expand_target_clones): Handle it.
2693 (ipa_target_clone): Make redirection just for target_clones
2696 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2697 Tom de Vries <tom@codesourcery.com>
2700 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2701 tiled basic blocks without omp continue statements.
2703 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2706 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2707 vec_extract expression as having side effects to make sure it gets
2710 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2713 * config/i386/i386.c (get_builtin_code_for_version): Check
2716 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2719 * gcc/config.gcc: Support "skylake".
2720 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2722 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2723 (processor_target_table): Add "skylake".
2724 (ix86_option_override_internal): Add "skylake".
2725 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2726 PROCESSOR_CANNONLAKE.
2727 (get_builtin_code_for_version): Fix priority for
2728 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2729 PROCESSOR_SKYLAKE-AVX512.
2730 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2731 (processor_type): Add PROCESSOR_SKYLAKE.
2733 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2734 Jason Merrill <jason@redhat.com>
2737 * convert.c (convert_to_integer_1): Use direct recursion for
2738 enumeral types and types with a precision less than the number
2739 of bits in their mode.
2741 2018-04-16 Julia Koval <julia.koval@intel.com>
2744 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2745 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2747 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2750 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2751 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2752 and -mno-direct-move.
2754 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2757 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2758 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2759 Ensure negative shifts result in {0}.
2761 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2763 PR rtl-optimization/79916
2764 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2765 regs (if any) to define how to gnerate SD moves when LRA is in
2768 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2770 PR rtl-optimization/85393
2771 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2772 * except.c (expand_dw2_landing_pad_for_region): Make static.
2773 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2774 a label and unconditional jump to old_bb, rather than
2775 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2778 PR rtl-optimization/85376
2779 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2780 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2781 instead of a specific value.
2783 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2784 Bin Cheng <bin.cheng@arm.com>
2786 PR tree-optimization/82965
2787 PR tree-optimization/83991
2788 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2789 by_profile_only parameter.
2790 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2791 information if the loop was predicted to iterate too many times.
2792 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2794 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2797 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2800 2018-04-13 Martin Liska <mliska@suse.cz>
2801 Jakub Jelinek <jakub@redhat.com>
2804 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2805 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2806 * builtins.c (expand_builtin_memory_copy_args): Use
2807 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2808 handle dest_addr == pc_rtx.
2810 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2813 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2814 asked to not generate direct moves.
2815 (fix_trunc<mode>si2_stfiwx): Similar.
2816 (fix_trunc<mode>si2_internal): Similar.
2818 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2821 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2822 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2823 lookup if dest in some wider mode is known to be const0_rtx and
2824 if so, record permanent equivalence for it to be ZERO_EXTEND of
2825 the narrower mode destination.
2827 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2829 * lto-streamer-out.c (output_function): Revert 259346.
2830 * omp-expand.c (expand_oacc_for): Likewise.
2832 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2834 PR rtl-optimization/85354
2835 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2836 * sel-sched.c (sel_global_init): ... here.
2838 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2841 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2842 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2843 mode for PE-COFF targets.
2844 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2845 (i386_pe_asm_lto_end): Likewise.
2846 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2847 (TARGET_ASM_LTO_END): Likewise.
2848 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2849 (i386_pe_asm_lto_start): New function.
2850 (i386_pe_asm_lto_end): Likewise.
2852 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2853 Richard Biener <rguenther@suse.de>
2856 * lto-streamer-out.c (output_function): Fix CFG loop state before
2858 * omp-expand.c (expand_oacc_for): Handle calls to internal
2859 functions like regular functions.
2861 2018-04-12 Richard Biener <rguenther@suse.de>
2864 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2865 for the early LTO debug to properly generate references to it
2866 during DIE emission. Do not re-use that for the skeleton for
2868 (dwarf2out_early_finish): Likewise.
2870 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2873 * config/i386/sse.md
2874 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2875 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2876 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2877 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2878 and output is a reg, avoid creating invalid lowpart subreg, but
2879 instead split into a 512-bit move. Don't split if not AVX512VL,
2880 input is xmm16+ reg and output is a mem.
2881 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2882 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2883 xmm16+ reg and output is a mem.
2885 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2887 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2888 also for flag_dwarf2_cfi_asm.
2890 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2892 PR rtl-optimization/85342
2893 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2894 a bool scalar var inside of the loop instead. Don't try to update
2895 recog_data.operand after failed apply_change_group.
2897 2018-04-12 Tom de Vries <tom@codesourcery.com>
2900 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2901 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2902 array with flexible array member as array without given dimension.
2903 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2904 argument for undefined param to true.
2906 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2909 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2910 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2911 from PowerPC section.
2912 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2913 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2914 help text that is too long.
2915 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2916 help text that is too long.
2917 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2918 help text that is too long.
2920 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2922 * config/alpha/alpha.md (stack_probe_internal): Rename
2923 from "probe_stack". Update all callers.
2925 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2927 PR rtl-optimization/84566
2928 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2929 sched_macro_fuse_insns.
2931 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2934 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2935 (compute_block_dependences): ... from here.
2937 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2939 PR tree-optimization/85331
2940 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2941 from int to HOST_WIDE_INT.
2943 2018-04-11 Martin Jambor <mjambor@suse.cz>
2946 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2947 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2948 not the same as the source val.
2949 (cgraph_edge_brings_value_p): New parameter.
2950 (gather_edges_for_value): Pass destination value to
2951 cgraph_edge_brings_value_p.
2952 (perhaps_add_new_callers): Likewise.
2953 (get_info_about_necessary_edges): Likewise and exclude values brought
2954 only by self-recursive edges.
2955 (create_specialized_node): Redirect only clones of self-calling edges.
2956 (+self_recursive_pass_through_p): New function.
2957 (find_more_scalar_values_for_callers_subset): Use it.
2958 (find_aggregate_values_for_callers_subset): Likewise.
2959 (known_aggs_to_agg_replacement_list): Removed.
2960 (decide_whether_version_node): Re-calculate known constants for all
2961 remaining context clones.
2963 2018-04-11 Richard Biener <rguenther@suse.de>
2966 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2967 from early DWARF output.
2968 (dwarf2out_early_finish): Output line info unconditionally into
2969 early DWARF and add reference to it.
2971 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2974 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2975 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2976 other than V2DFmode using iptr mode attribute.
2977 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2979 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2981 PR rtl-optimization/84659
2982 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2984 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2987 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2989 (output_loc_list): Pass address of a dummy size variable even in the
2990 locview handling loop.
2991 (index_location_lists): Add comment on why skip_loc_list_entry can't
2994 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2997 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3000 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3003 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3004 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3005 and -mstring-compare-inline-limit.
3007 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3010 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3011 for stack clash protection in a register whenever we need it to be in
3014 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3016 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3017 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3019 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3022 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3024 (mlong-double-): Ditto.
3025 * config/rs6000/sysv4.opt (msdata=): Ditto.
3026 (mtls-size=): Ditto.
3028 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3030 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3031 erroneous entries for
3032 "vector int vec_ldl (int, long int *)", and
3033 "vector unsigned int vec_ldl (int, unsigned long int *)".
3034 Add comments and entries for
3035 "vector bool char vec_ldl (int, bool char *)",
3036 "vector bool short vec_ldl (int, bool short *)",
3037 "vector bool int vec_ldl (int, bool int *)",
3038 "vector bool long long vec_ldl (int, bool long long *)",
3039 "vector pixel vec_ldl (int, pixel *)",
3040 "vector long long vec_ldl (int, long long *)",
3041 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3042 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3043 type tree bool_long_long_type_node and correct definition of
3044 bool_V2DI_type_node to make reference to this new type tree.
3045 (rs6000_mangle_type): Replace erroneous reference to
3046 bool_long_type_node with bool_long_long_type_node.
3047 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3048 comments to emphasize sign distinctions for char and int types and
3049 replace RS6000_BTI_bool_long constant with
3050 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3051 use of RS6000_BTI_pixel.
3052 (bool_long_type_node): Remove this macro definition.
3053 (bool_long_long_type_node): New macro definition
3055 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3057 PR rtl-optimization/85300
3058 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3059 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3060 simplify_unary_operation fails.
3062 2018-04-10 Martin Liska <mliska@suse.cz>
3064 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3065 cgraph_edge and ipa_ref.
3067 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3071 * config/i386/sse.md
3072 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3073 computation of the VEC_MERGE selector from mask.
3074 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3075 Fix decoding of the VEC_MERGE selector into mask.
3077 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3079 PR tree-optimization/85286
3080 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3082 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3084 * final.c (final_1): Set insn_last_address as well as
3085 insn_current_address.
3087 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3090 * explow.c (emit_stack_probe): Call validize_mem on memory location
3091 before passing it to gen_probe_stack. Create address operand and
3092 legitimize it for the probe_stack_address case.
3094 2018-04-09 Jan Hubicka <jh@suse.cz>
3097 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3098 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3099 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3101 (free_lang_data): Rebuild type inheritance graph.
3103 2018-04-09 Martin Sebor <msebor@redhat.com>
3105 * invoke.texi (-finline-small-functions): Mention other optimization
3107 (-findirect-inlining, -fpartial-inlining): Same.
3108 (-finline-functions-called-once): Same.
3109 (-freorder-blocks-and-partition): Same.
3111 2018-04-09 Jan Hubicka <jh@suse.cz>
3114 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3115 jumps; choose last target that matches the criteria (i.e.
3116 no partition changes for non-crossing jumps).
3117 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3118 support for redirecting crossing jumps to non-crossing.
3120 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3122 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3123 also for naked functions.
3125 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3127 * config/arc/arc.md (add_shift): New pattern.
3128 (add_shift2): Likewise.
3129 (sub_shift): Likewise.
3130 (sub_shift_cmp0_noout): Likewise.
3131 (compare_si_ashiftsi): Likewise.
3132 (xbfu_cmp0_noout): New combine pattern.
3133 (xbfu_cmp0"): Likewise.
3134 (movsi_set_cc_insn): Place the predicable variant first.
3135 (commutative_binary_cmp0_noout): Remove clobber.
3136 (commutative_binary_cmp0): New pattern.
3137 (noncommutative_binary_cmp0): Likewise.
3138 (noncommutative_binary_cmp0_noout): Likewise.
3139 (noncommutative_binary_comparison_result_used): Removed.
3140 (rsub_cmp0): New pattern.
3141 (rsub_cmp0_noout): Likewise.
3142 (extzvsi): Changed, keep only meaningful variants.
3143 (SQH, SEZ): New iterators.
3144 (SQH_postfix): New mode attribute.
3145 (SEZ_prefix): New code attribute.
3146 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3147 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3148 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3150 (noncommutative_operator): Check the availability of barrel
3153 2018-04-09 Richard Biener <rguenther@suse.de>
3155 PR tree-optimization/85284
3156 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3157 Only use the niter constraining form of simple_iv when the exit
3160 2018-04-09 Tom de Vries <tom@codesourcery.com>
3163 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3164 (define_expand "*memory_barrier"): New define_expand.
3165 (define_insn "memory_barrier"): New insn.
3167 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3169 PR rtl-optimization/80463
3170 PR rtl-optimization/83972
3171 PR rtl-optimization/83480
3173 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3174 correct producer for the insn.
3175 (tidy_control_flow): Fixup seqnos in case of debug insns.
3177 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3179 PR rtl-optimization/83913
3181 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3182 different sched-times when merging exprs.
3184 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3186 PR rtl-optimization/83962
3188 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3189 tidy_fallthru_edge and tidy_control_flow.
3191 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3193 PR rtl-optimization/83530
3195 * sel-sched.c (force_next_insn): New global variable.
3196 (remove_insn_for_debug): When force_next_insn is true, also leave only
3197 next insn in the ready list.
3198 (sel_sched_region): When the region wasn't scheduled, make another pass
3199 over it with force_next_insn set to 1.
3201 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3203 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3205 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3206 for interrupt control.
3207 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3208 functions for interrupt control.
3209 * config/nds32/nds32-intrinsic.md: Likewise.
3210 * config/nds32/nds32_intrinsic.h: Likewise.
3211 * config/nds32/nds32.h (nds32_builtins): Likewise.
3213 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3215 * config/nds32/nds32.c (nds32_init_machine_status,
3216 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3217 strict_aligned_p field.
3218 (nds32_expand_to_rtl_hook): New function.
3219 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3220 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3222 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3223 Chung-Ju Wu <jasonwucj@gmail.com>
3225 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3226 * config/nds32/nds32-n7.md: New file.
3227 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3228 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3230 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3231 * config/nds32/nds32.md (pipeline_model): Add n7.
3232 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3233 * config/nds32/pipelines.md: Include n7 settings.
3235 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3236 Chung-Ju Wu <jasonwucj@gmail.com>
3238 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3239 * config/nds32/nds32-e8.md: New file.
3240 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3241 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3243 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3244 * config/nds32/nds32.md (pipeline_model): Add e8.
3245 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3246 * config/nds32/pipelines.md: Include e8 settings.
3248 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3249 Chung-Ju Wu <jasonwucj@gmail.com>
3251 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3252 * config/nds32/nds32-n8.md: New file.
3253 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3254 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3256 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3257 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3258 * config/nds32/nds32.md (pipeline_model): Add n8.
3259 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3260 * config/nds32/pipelines.md: Include n8 settings.
3262 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3263 Chung-Ju Wu <jasonwucj@gmail.com>
3265 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3266 * config/nds32/nds32-n9-2r1w.md: New file.
3267 * config/nds32/nds32-n9-3r2w.md: New file.
3268 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3269 nds32_register_ports): New or modify for cpu n9.
3270 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3272 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3273 * config/nds32/nds32-utils.c: New file.
3274 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3275 TARGET_MUL_SLOW): Define.
3276 * config/nds32/nds32.md (pipeline_model): New attribute.
3277 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3278 New options that support cpu n9.
3279 * config/nds32/pipelines.md: Include n9 settings.
3280 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3282 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3284 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3285 information if necessary.
3286 (output_cond_branch_compare_zero): Likewise.
3287 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3288 (nds32_target_alignment): Refine for alignment.
3289 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3290 (FUNCTION_BOUNDARY): Modify.
3291 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3293 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3295 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3297 * config/nds32/constants.md (unspec_volatile_element): Add values for
3298 TLB operation and data prefetch.
3299 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3300 functions for TLB operation and data prefetch.
3301 * config/nds32/nds32-intrinsic.md: Likewise.
3302 * config/nds32/nds32_intrinsic.h: Likewise.
3303 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3304 (nds32_print_operand): Likewise.
3305 * config/nds32/nds32.h (nds32_builtins): Likewise.
3307 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3308 Andrew Pinski <pinsika@gcc.gnu.org>
3311 * match.pd: Use constant_boolean_node of correct type instead of
3312 boolean_true_node or boolean_false_node for simplifying
3313 pointer comparisons to zero.
3315 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3317 PR tree-optimization/80021
3318 * tree.c (verify_type_variant): Make error call in verify_variant_match
3319 translatable and remove final full stop.
3321 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3323 * config/nds32/constants.md (unspec_volatile_element): Add
3324 UNSPEC_VOLATILE_EH_RETURN.
3325 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3326 nds32_output_stack_pop): Support dwarf exception handling process.
3327 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3328 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3329 exception handling process.
3330 (nds32_compute_stack_frame): Likewise.
3331 (nds32_return_addr_rtx): Likewise.
3332 (nds32_initial_elimination_offset): Likewise.
3333 (nds32_expand_prologue): Likewise.
3334 (nds32_expand_epilogue): Likewise.
3335 (nds32_dynamic_chain_address): New function.
3336 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3338 (DYNAMIC_CHAIN_ADDRESS): Define.
3339 (EH_RETURN_DATA_REGNO): Define.
3340 (EH_RETURN_STACKADJ_RTX): Define.
3341 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3342 patterns for dwarf exception handling.
3344 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3346 * config/nds32/nds32.h: Clean up obsolete macros.
3348 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3350 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3351 Add enum values for particular instructions.
3352 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3353 particular intrinsic functions.
3354 * config/nds32/nds32-intrinsic.md: Likewise.
3355 * config/nds32/nds32_intrinsic.h: Likewise.
3356 * config/nds32/nds32.h (nds32_builtins): Likewise.
3357 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3358 (btst, ave): New patterns for particular instructions.
3360 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3362 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3363 Add enum values for atomic load/store and memory sync.
3364 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3366 * config/nds32/nds32-intrinsic.md: Likewise.
3367 * config/nds32/nds32_intrinsic.h: Likewise.
3368 * config/nds32/nds32.h (nds32_builtins): Likewise.
3370 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3372 PR tree-optimization/85257
3373 * fold-const.c (native_encode_vector): If not all elts could fit
3374 and off is -1, return 0 rather than offset.
3375 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3376 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3377 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3378 adjust buffer in native_interpret_expr call.
3380 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3382 * config/nds32/constants.md (unspec_volatile_element): Add cache
3383 control enum values.
3384 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3385 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3386 * config/nds32/nds32.c (nds32_cctl_names): New.
3387 (nds32_print_operand): Handle cache control register names.
3388 * config/nds32/nds32.h (nds32_builtins): New enum values.
3389 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3391 * config/nds32/nds32.md (type): Add mmu.
3392 * config/nds32/pipelines.md (simple_insn): Add mmu.
3394 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3396 * config/nds32/nds32.md (type): Remove call.
3397 * config/nds32/pipelines.md (simple_insn): Likewise.
3399 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3401 * config/nds32/constants.md (unspec_volatile_element): Add
3402 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3403 UNSPEC_VOLATILE_FMFCFG.
3404 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3405 description for fmfcfg and fmfcsr.
3406 (bdesc_1arg): Add fmtcsr.
3407 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3408 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3409 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3410 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3411 unspec_fmfcfg): New patterns.
3412 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3413 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3414 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3415 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3416 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3417 __nds32__fmfcfg): Define.
3419 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3421 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3422 intrinsic register names.
3423 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3424 intrinsic register enum values and macros.
3426 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3428 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3429 for load/store addressing form.
3430 (nds32_print_operand_address): Likewise.
3432 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3435 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3436 based on LABEL_REF. Remove useless assertion.
3437 (pic_address_needs_scratch): Fix formatting.
3438 (sparc_legitimize_pic_address): Minor tweaks.
3439 (sparc_delegitimize_address): Adjust assertion accordingly.
3440 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3441 into symbolic_operand.
3442 (movsi_high_pic_label_ref): Likewise.
3443 (movsi_lo_sum_pic_label_ref): Likewise.
3444 (movdi_pic_label_ref): Likewise.
3445 (movdi_high_pic_label_ref): Likewise.
3446 (movdi_lo_sum_pic_label_ref): Likewise.
3448 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3450 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3451 custom LIB_SPEC setup.
3453 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3454 Kito Cheng <kito.cheng@gmail.com>
3456 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3457 * config/riscv/freebsd.h: New.
3459 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3461 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3462 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3465 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3466 Kito Cheng <kito.cheng@gmail.com>
3468 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3469 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3470 * config/nds32/nds32-protos.h (nds32_output_call,
3471 nds32_output_return): Declare.
3472 * config/nds32/nds32.md: Refine all the call and return patterns.
3474 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3477 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3478 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3480 PR rtl-optimization/84872
3481 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3482 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3485 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3487 * expr.c (copy_blkmode_to_reg): Revert 254862.
3488 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3490 2018-04-06 Richard Biener <rguenther@suse.de>
3493 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3494 after seeing a component reference with an adjacent field. Treat
3495 refs to arrays at struct end of external decls similar to
3496 refs to unconstrained commons.
3498 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3501 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3502 look through SAVE_EXPRs with non-side-effects argument. Adjust
3504 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3507 2018-04-06 Richard Biener <rguenther@suse.de>
3510 * alias.c (find_base_term): New wrapper around find_base_term
3511 unwinding CSELIB_VAL_PTR changes.
3512 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3515 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3517 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3519 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3520 constant definitions.
3521 ("nop"): lr 0,0 -> nopr r0
3522 ("nop_lr0", "nop_lr1"): New insn definitions.
3524 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3526 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3527 NDS32_V3PUSH_AVAILABLE_P macro.
3529 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3530 Chung-Ju Wu <jasonwucj@gmail.com>
3532 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3533 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3534 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3535 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3536 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3537 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3538 * config/nds32/constraints.md: New constraints and checking for hard
3539 float configuration.
3540 * config/nds32/iterators.md: New mode iterator and attribute for hard
3541 float configuration.
3542 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3544 * config/nds32/nds32-fpu.md: New file.
3545 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3546 deal with hard float code generation.
3547 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3549 (abi_type, float_reg_number): New enum type.
3550 * config/nds32/nds32-predicates.c: New predicates for hard float.
3551 * config/nds32/nds32-protos.h: Declare functions for hard float.
3552 * config/nds32/nds32.c: Implementation for hard float configuration.
3553 * config/nds32/nds32.h: Definitions for hard float configuration.
3554 * config/nds32/nds32.md: Include hard float machine description and
3555 modify patterns for hard float configuration.
3556 * config/nds32/nds32.opt: New options for hard float configuration.
3557 * config/nds32/predicates.md: New predicates for hard float
3560 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3562 * common/config/nds32/nds32-common.c
3563 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3565 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3568 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3569 CONSTRUCTOR_ELT (ctor, ...)->value.
3571 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3574 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3576 2018-04-05 Tom de Vries <tom@codesourcery.com>
3579 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3582 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3583 Kito Cheng <kito.cheng@gmail.com>
3585 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3586 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3587 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3588 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3590 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3591 Kito Cheng <kito.cheng@gmail.com>
3593 * config/nds32/constraints.md (Ufe): New memory constraint.
3594 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3595 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3596 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3598 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3599 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3601 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3603 * config/nds32/nds32.md: Use optimize_size in the condition for
3604 alu-shift instructions.
3606 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3608 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3610 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3612 * config/nds32/nds32.md (negsi2): Refine pattern.
3614 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3615 Chung-Ju Wu <jasonwucj@gmail.com>
3617 * config/nds32/iterators.md (shift_rotate): New code iterator.
3618 (shift): New code attribute.
3619 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3620 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3621 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3622 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3623 bit-wise operations.
3624 (andsi3, *andsi3): Ditto.
3625 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3626 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3627 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3628 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3629 nds32_ior_operand, nds32_xor_operand): New predicates.
3631 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3633 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3634 (addsi3, subsi3): ... this.
3636 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3638 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3640 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3642 * config/nds32/nds32.md: Adjust indention.
3644 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3646 * config/nds32/nds32.md (feature): New attribute.
3648 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3650 * config/nds32/nds32.md (subtype): New attribute.
3652 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3655 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3656 expansion to perform a bitwise AND of the argument followed by a
3657 boolean negation of the result.
3659 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3661 PR rtl-optimization/84878
3662 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3663 the basic block. Assert the use reference is not artificial and that
3664 it has an associated insn.
3666 2018-04-04 Michael Matz <matz@suse.de>
3668 * builtins.c (compute_objsize): Pass correct operand
3669 to array_at_struct_end_p.
3671 2018-04-04 Richard Biener <rguenther@suse.de>
3674 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3675 from contexts for DINFO_LEVEL_TERSE and below.
3677 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3679 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3680 resiter_operand condition.
3681 * config/nds32/nds32.md (*move<mode>): Ditto.
3683 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3684 Monk Chiang <sh.chiang04@gmail.com>
3686 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3688 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3690 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3692 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3693 Kito Cheng <kito.cheng@gmail.com>
3695 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3696 nds32_cond_code_str, output_cond_branch,
3697 output_cond_branch_compare_zero, nds32_expand_cbranch,
3698 nds32_expand_cstore, nds32_expand_movcc,
3699 nds32_output_cbranchsi4_equality_zero,
3700 nds32_output_cbranchsi4_equality_reg,
3701 nds32_output_cbranchsi4_equality_reg_or_const_int,
3702 nds32_output_cbranchsi4_greater_less_zero: New functions.
3703 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3704 nds32_expand_cstore, nds32_expand_movcc,
3705 nds32_output_cbranchsi4_equality_zero,
3706 nds32_output_cbranchsi4_equality_reg,
3707 nds32_output_cbranchsi4_equality_reg_or_const_int,
3708 nds32_output_cbranchsi4_greater_less_zero): Declare.
3709 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3710 nds32_rimm11s_operand): New predicates.
3711 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3712 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3715 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3717 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3718 * config/nds32/nds32.md: Ditto.
3719 * config/nds32/pipelines.md: Ditto.
3721 2018-04-04 Richard Biener <rguenther@suse.de>
3723 PR tree-optimization/85168
3724 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3725 propagating abnormals.
3727 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3729 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3731 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3732 Kito Cheng <kito.cheng@gmail.com>
3734 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3735 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3736 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3737 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3738 * config/nds32/nds32.md (sibcall_internal): New.
3739 (sibcall_register): Remove.
3740 (sibcall_immediate): Remove.
3741 (sibcall_value_internal): New.
3742 (sibcall_value_register): Remove.
3743 (sibcall_value_immediate): Remove.
3744 * config/nds32/predicates.md (nds32_general_register_operand): New.
3745 (nds32_call_address_operand): New.
3747 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3749 PR rtl-optimization/85167
3750 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3751 bb_defs if *split_p, instead preinitialize it to NULL.
3753 PR tree-optimization/85156
3754 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3755 evaluating the argument multiple times.
3757 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3759 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3761 (_mm_cvtpd_ps): Likewise.
3762 (_mm_cvttpd_epi32): Likewise.
3763 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3764 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3765 vector, pixel, and bool following altivec.h include.
3767 2018-04-03 Martin Sebor <msebor@redhat.com>
3769 * doc/extend.texi (Common Function Attributes): Clarify.
3770 (const attribute): Likewise.
3771 (pure attribute): Likewise.
3773 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3776 * config/i386/i386.c (ix86_expand_vector_set): Use
3777 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3779 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3781 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3782 instructions when changing rounding bits to preserve precision bits
3783 in the x87 control word.
3785 2018-04-03 Martin Liska <mliska@suse.cz>
3787 PR tree-optimization/82491
3788 * rtl.h (strip_offset_and_add): Replace += suboffset with
3789 poly_uint64 () + suboffset.
3791 2018-03-29 Martin Liska <mliska@suse.cz>
3792 Martin Jambor <mjambor@suse.cz>
3795 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3796 param_type is not an integral or pointer type.
3798 2018-04-03 Richard Biener <rguenther@suse.de>
3800 * sese.h (recompute_all_dominators): Remove.
3802 2018-04-02 Martin Sebor <msebor@redhat.com>
3804 * doc/invoke.texi (-Wrestrict): Fix typos.
3806 2018-04-02 Jim Wilson <jimw@sifive.com>
3808 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3809 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3810 (<optab>di3, <optab>si3_extend): Likewise.
3811 (<optab>si3_mask, <optab>si3_mask_1): New.
3812 (<optab>di3_mask, <optab>di3_mask_1): New.
3813 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3814 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3815 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3817 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3819 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3822 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3824 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3825 (nds32_canonicalize_comparison): New function.
3827 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3828 Kito Cheng <kito.cheng@gmail.com>
3829 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3831 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3832 * config/nds32/constants.md (unspec_volatile_element): Add
3833 UNSPEC_VOLATILE_RELAX_GROUP.
3834 * config/nds32/nds32-relax-opt.c: New file.
3835 * config/nds32/nds32-predicates.c
3836 (nds32_symbol_load_store_p): New function.
3837 * config/nds32/nds32-protos.h
3838 (nds32_symbol_load_store_p): Declare function.
3839 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3840 * config/nds32/nds32.c
3841 (nds32_register_pass): New function to register pass.
3842 (nds32_register_passes): New function to register passes.
3843 * config/nds32/nds32.md (relax_group): New pattern.
3844 * config/nds32/nds32.opt (mrelax-hint): New option.
3845 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3847 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3849 * config/nds32/t-nds32: Modify files dependency.
3851 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3853 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3854 (PROFILE_HOOK): Define its implementation.
3856 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3858 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3859 type and 32-bit size.
3861 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3864 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3865 (V_128_256): New mode iterator.
3866 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3867 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3868 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3870 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3871 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3873 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3876 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3877 NaN inputs correctly.
3879 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3882 * config/rs6000/vsx.md (??r): New mode attribute.
3883 (*vsx_mov<mode>_64bit): Use it.
3884 (*vsx_mov<mode>_32bit): Likewise.
3886 2018-03-30 Martin Sebor <msebor@redhat.com>
3888 PR tree-optimization/84818
3889 * builtins.c (check_access): Use warning_n.
3891 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3894 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3896 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3899 2018-03-30 Julia Koval <julia.koval@intel.com>
3902 * x86-tune.def (movx, partial_reg_dependency): Enable for
3905 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3908 * lra-constraints.c (process_alt_operands): Move setting
3909 this_alternative_matches below.
3911 2018-03-29 Martin Liska <mliska@suse.cz>
3914 * doc/invoke.texi: Document how LTO works with debug info.
3915 Describe auto-load support of binutils. Mention 'x86-64'
3916 as valid option value of -march option.
3918 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3920 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3923 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3924 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3925 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3928 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3931 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3932 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3933 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3934 for RS6000_BTM_POWERPC64.
3935 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3936 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3937 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3942 2018-03-28 Carl Love <cel@us.ibm.com>
3945 2017-09-27 Carl Love <cel@us.ibm.com>
3947 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3948 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3949 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3952 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3954 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3955 instead of __vector bool.
3956 (_mm_max_pu8): Likewise.
3957 (_mm_min_pi16): Likewise.
3959 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3962 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3963 (DIVWEUO): Likewise.
3965 (DIVDEUO): Likewise.
3966 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3967 DIVWEUO and DIVDEUO.
3968 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3969 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3970 (div_extend): Likewise.
3971 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3973 (__builtin_divweuo): Likewise.
3974 (__builtin_divdeo): Likewise.
3975 (__builtin_divdeuo): Likewise.
3977 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3980 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3981 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3983 PR tree-optimization/82004
3984 * gimple-match-head.c (optimize_pow_to_exp): New function.
3985 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3986 Don't fold to exp if optimize_pow_to_exp is false.
3988 2018-03-28 Martin Liska <mliska@suse.cz>
3991 * calls.c (initialize_argument_information): Fix trailing space.
3992 * common.opt: Fix typo and provide better explanation for
3993 -fsanitize-coverage option.
3994 * config/i386/i386.opt: Fix typo.
3996 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3997 Martin Liska <mliska@suse.cz>
4000 * gimplify.c (asan_poison_variable): Don't do the check for
4001 gimplify_omp_ctxp here.
4002 (gimplify_decl_expr): Do it here.
4003 (gimplify_target_expr): Likewise.
4005 2018-03-28 Martin Liska <mliska@suse.cz>
4008 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4009 chkp_type_bounds_count if MPX is not enabled.
4011 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4013 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4015 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4018 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4019 function to create the function decl for complex long double
4020 multiply and divide for -mabi=ieeelongdouble.
4021 (init_float128_ieee): Call it.
4023 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4026 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4027 -fcf-protection=branch -mibt.
4028 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4030 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4033 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4035 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4038 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4039 extern array declarations.
4041 2018-03-27 Richard Biener <rguenther@suse.de>
4044 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4045 explicit single_use checks.
4047 2018-03-27 Richard Biener <rguenther@suse.de>
4049 PR tree-optimization/85082
4050 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4053 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4055 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4056 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4057 Turn on fasynchronous-unwind-tables and funwind-tables.
4059 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4062 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4063 (*bmi_blsr_<mode>_ccz): Ditto.
4065 2018-03-26 Tom de Vries <tom@codesourcery.com>
4067 PR tree-optimization/85063
4068 * omp-general.c (offloading_function_p): New function. Factor out
4070 * omp-offload.c (pass_omp_target_link::gate): ... here.
4071 * omp-general.h (offloading_function_p): Declare.
4072 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4073 with attribute omp declare target for offloading functions.
4075 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4077 PR tree-optimization/84005
4078 * tree-data-ref.h (get_base_for_alignment): Declare.
4079 * tree-data-ref.c (get_base_for_alignment_1): New function.
4080 (get_base_for_alignment): Likewise.
4081 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4082 get_base_for_alignment to find a suitable base object, instead
4083 of always using drb->base_address.
4085 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4088 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4089 known size by default.
4091 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4094 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4095 and non BLKmode operands.
4097 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4100 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4101 Clean up attributes.
4103 2018-03-23 Richard Biener <rguenther@suse.de>
4106 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4107 we are going to emit early debug for LTO.
4109 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4112 * function.c (match_asm_constraints_1): Don't optimize if input
4113 doesn't satisfy general_operand predicate for output's mode.
4116 * alias.c (write_dependence_p): Don't require for x_canonicalized
4117 non-VOIDmode if x has VOIDmode.
4120 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4121 just don't try to optimize it rather than assert it never happens.
4123 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4125 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4126 macro expansions for definition of ST_INTERNAL_<mode> and
4127 LD_INTERNAL_<mode> builtins.
4128 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4130 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4132 (altivec_expand_st_builtin): Likewise.
4133 (altivec_expand_builtin): Remove calls to deleted functions.
4134 (rs6000_address_for_altivec): Delete this function.
4135 * config/rs6000/vector.md: Remove expands for
4136 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4138 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4141 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4142 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4143 re-computing once computed.
4144 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4145 (arm_init_machine_status): Initialize
4146 machine->static_chain_stack_bytes.
4148 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4151 * doc/extend.texi: Add four new prototypes for vec_ld.
4152 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4153 definitions for more logical presentation.
4154 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4155 entries for V1TI variants of __builtin_altivec_ld builtin.
4156 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4157 handling of V1TI variant of LVX icode pattern.
4158 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4159 (rs6000_gimple_fold_builtin): Likewise.
4160 (altivec_init_builtins): Add code to define
4161 __builtin_altivec_lvx_v1ti function.
4163 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4166 * function.c (match_asm_constraints_1): Don't do the optimization
4167 if input isn't a REG, SUBREG, MEM or constant.
4169 2018-03-22 Tom de Vries <tom@codesourcery.com>
4171 PR tree-optimization/84956
4172 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4173 bb_has_abnormal_pred.
4175 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4178 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4179 DECL_INITIAL (decl) to decl at the end.
4180 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4183 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4185 * doc/extend.texi (__builtin_tgmath): Document when complex
4186 integer types are treated as _Complex _Float64.
4188 2018-03-21 Tom de Vries <tom@codesourcery.com>
4190 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4192 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4194 PR tree-optimization/84960
4195 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4196 if it is ENTRY block, move them into single succ of ENTRY in that case.
4198 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4200 PR tree-optimization/84811
4201 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4202 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4203 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4204 (known_subrange_p): Update accordingly. Cast each value involved
4205 in the size comparison, rather than casting the result of the
4208 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4210 PR tree-optimization/84982
4211 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4212 by flipping the least significant bit rather than all bits from
4213 bitpos to bitpos + bitsize - 1.
4215 2018-03-21 Nathan Sidwell <nathan@acm.org>
4217 * doc/extend.texi (Deprecated Features): Remove mention of
4218 long-deleted deprecations.
4220 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4223 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4224 * configure: Regenerate.
4226 2018-03-21 Tom de Vries <tom@codesourcery.com>
4228 PR tree-optimization/83126
4229 * tree-parloops.c (num_phis): New function.
4230 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4232 2018-03-21 Nathan Sidwell <nathan@acm.org>
4234 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4235 mention anon-struct/union members and trailing attributes.
4237 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4239 PR tree-optimization/84969
4240 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4241 builtin memset partitions if they set different rhs values.
4243 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4245 PR rtl-optimization/84989
4246 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4247 VEC_DUPLICATE with scalar result mode.
4249 2018-03-21 Martin Liska <mliska@suse.cz>
4252 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4253 not intended return statement.
4255 2018-03-21 Martin Liska <mliska@suse.cz>
4258 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4259 (chkp_find_bound_slots_1): Limit number of iterations.
4261 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4264 * Minor grammar fixes for x86 options.
4266 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4269 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4270 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4272 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4275 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4276 (altivec_lvx_<mode>_1op): Likewise.
4277 (altivec_stvx_<mode>_2op): Likewise.
4278 (altivec_stvx_<mode>_1op): Likewise.
4279 (altivec_lvx_<VM2:mode>): New define_expand.
4280 (altivec_stvx_<VM2:mode>): Likewise.
4281 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4282 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4283 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4284 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4285 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4286 (rs6000_gen_lvx): Likewise.
4287 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4288 (altivec_expand_stv_builtin): Likewise.
4289 (altivec_expand_builtin): Likewise.
4290 * config/rs6000/vector.md: Likewise.
4292 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4295 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4298 2018-03-20 Richard Biener <rguenther@suse.de>
4301 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4302 sign-conversions as zero, fall back to standard scalar_stmt
4305 2018-03-20 Martin Liska <mliska@suse.cz>
4308 * predict.c (rebuild_frequencies): Handle case when we have
4309 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4311 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4314 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4315 flag_section_anchors.
4316 * varasm.c (use_blocks_for_decl_p): Remove hack for
4317 dw2_force_const_mem.
4320 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4322 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4323 be created, use lowpart_subreg of operands[0] rather than operands[0]
4325 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4326 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4327 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4328 and n constraint instead of aarch64_shift_imm_di and Usd.
4329 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4330 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4332 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4335 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4336 to favor GPR over NEON registers.
4337 (<shift>di3_neon): Likewise.
4339 2018-03-20 Tom de Vries <tom@codesourcery.com>
4342 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4343 (nvptx_process_pars): Emit bar.sync asap and alap.
4345 2018-03-20 Tom de Vries <tom@codesourcery.com>
4348 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4349 seen_label if seen_label is already set.
4351 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4354 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4355 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4356 Use 1U instead of 1. Formatting fixes.
4359 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4360 instead of TREE_TYPE (s1) for the return value.
4362 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4364 PR tree-optimization/84946
4365 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4366 bitsize + bitsize in poly_uint64 rather than poly_int64.
4369 * dwarf2asm.c: Include fold-const.c.
4370 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4371 of decl rather than decl itself.
4373 PR rtl-optimization/84643
4374 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4376 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4379 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4380 calling assemble_variable.
4382 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4385 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4386 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4388 2018-03-19 Jim Wilson <jimw@sifive.com>
4391 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4392 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4393 (riscv_first_stack_step): Likewise.
4394 (riscv_option_override): Use STACK_BOUNDARY instead of
4396 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4398 (BIGGEST_ALIGNMENT): Set to 128.
4399 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4400 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4403 2018-03-19 Richard Biener <rguenther@suse.de>
4405 PR tree-optimization/84933
4406 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4407 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4409 2018-03-19 Richard Biener <rguenther@suse.de>
4411 PR tree-optimization/84859
4412 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4413 (cond_if_else_store_replacement): Perform sinking operation on
4414 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4415 Generalize what a BB with a single eligible store is.
4417 2018-03-19 Richard Biener <rguenther@suse.de>
4419 PR tree-optimization/84929
4420 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4421 chrec_is_positive against non-chrec arg.
4423 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4426 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4428 2018-03-18 Martin Liska <mliska@suse.cz>
4430 PR rtl-optimization/84635
4431 * regrename.c (build_def_use): Use matches_mode only when
4434 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4436 PR tree-optimization/84913
4437 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4438 vectorize chains of COND_EXPRs.
4440 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4442 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4444 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4446 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4448 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4450 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4452 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4453 Kito Cheng <kito.cheng@gmail.com>
4455 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4456 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4457 (nds32_adjust_reg_alloc_order): New function.
4458 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4460 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4462 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4463 nds32_print_operand, nds32_print_operand_address): Use
4464 HOST_WIDE_INT_PRINT_DEC instead.
4466 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4468 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4470 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4473 * config/i386/i386.c (initial_ix86_tune_features,
4474 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4476 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4477 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4478 rather than 1u << ix86_tune. Formatting fix.
4479 (ix86_option_override_internal): Change ix86_arch_mask from
4480 unsigned int to unsigned HOST_WIDE_INT, initialize to
4481 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4482 (ix86_function_specific_restore): Likewise.
4484 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4487 * postreload.c (reload_combine_recognize_pattern): Perform
4488 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4489 truncate_int_for_mode the result for the destination's mode.
4492 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4493 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4495 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4498 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4501 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4502 regno_allocno_class_array and sorted_pseudos.
4503 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4504 insns where regno is used.
4506 2018-03-16 Martin Liska <mliska@suse.cz>
4509 * multiple_target.c (create_dispatcher_calls): Redirect
4510 reference in the symbol table.
4512 2018-03-16 Martin Liska <mliska@suse.cz>
4515 * multiple_target.c (create_dispatcher_calls): Redirect also
4518 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4522 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4523 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4524 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4526 2018-03-16 Julia Koval <julia.koval@intel.com>
4528 * doc/invoke.texi (Skylake Server): Add CLWB.
4529 Cannonlake): Remove CLWB.
4531 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4533 PR tree-optimization/84841
4534 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4536 (FLOAT_ONE_CONST_TYPE): Define.
4537 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4538 (sort_by_operand_rank): Put entries with higher constant_type last
4539 rather than first to match comments.
4541 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4543 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4546 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4549 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4550 check and use address of target always.
4552 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4555 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4556 (indirect_thunk_bnd_needed): Likewise.
4557 (indirect_thunks_used): Likewise.
4558 (indirect_thunks_bnd_used): Likewise.
4559 (indirect_return_needed): New.
4560 (indirect_return_bnd_needed): Likewise.
4561 (output_indirect_thunk_function): Add a bool argument for
4563 (output_indirect_thunk_function): Don't generate alias for
4564 function return thunk.
4565 (ix86_code_end): Call output_indirect_thunk_function to generate
4566 function return thunks.
4567 (ix86_output_function_return): Set indirect_return_bnd_needed
4568 and indirect_return_needed instead of indirect_thunk_bnd_needed
4569 and indirect_thunk_needed.
4571 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4573 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4574 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4575 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4577 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4578 Paul Hua <paul.hua.gm@gmail.com>
4581 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4583 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4585 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4586 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4589 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4592 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4593 instead of GET_MODE_SIZE when comparing Units.
4595 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4598 * varasm.c (hash_section): Return an unchangeble hash value
4599 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4600 Return !aarch64_can_use_per_function_literal_pools_p ().
4602 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4605 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4606 rather than address of cmode as last argument to prepare_cmp_insn.
4608 2018-03-15 Julia Koval <julia.koval@intel.com>
4610 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4611 F_AVX512VNNI, F_AVX512BITALG): New.
4613 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4616 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4617 insn for floating-point loads and stores.
4619 2018-03-14 Carl Love <cel@us.ibm.com>
4621 * config/rs6000/rs6000-c.c: Add macro definitions for
4622 ALTIVEC_BUILTIN_VEC_PERMXOR.
4623 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4624 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4625 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4626 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4628 * config/doc/extend.texi: Add prototypes for vec_permxor.
4630 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4633 * diagnostic-show-locus.c (class layout_point): Convert m_line
4634 from int to linenum_type.
4635 (line_span::comparator): Use linenum "compare" function when
4636 comparing line numbers.
4637 (test_line_span): New function.
4638 (layout_range::contains_point): Convert param "row" from int to
4640 (layout_range::intersects_line_p): Likewise.
4641 (layout::will_show_line_p): Likewise.
4642 (layout::print_source_line): Likewise.
4643 (layout::should_print_annotation_line_p): Likewise.
4644 (layout::print_annotation_line): Likewise.
4645 (layout::print_leading_fixits): Likewise.
4646 (layout::annotation_line_showed_range_p): Likewise.
4647 (struct line_corrections): Likewise for field m_row.
4648 (line_corrections::line_corrections): Likewise for param "row".
4649 (layout::print_trailing_fixits): Likewise.
4650 (layout::get_state_at_point): Likewise.
4651 (layout::get_x_bound_for_row): Likewise.
4652 (layout::print_line): Likewise.
4653 (diagnostic_show_locus): Likewise for locals "last_line" and
4655 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4656 * input.c (selftest::test_linenum_comparisons): New function.
4657 (selftest::input_c_tests): Call it.
4658 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4659 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4660 * selftest.h (ASSERT_GT): New macro.
4661 (ASSERT_GT_AT): New macro.
4662 (ASSERT_LT): New macro.
4663 (ASSERT_LT_AT): New macro.
4665 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4667 PR rtl-optimization/84780
4668 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4670 2018-03-14 Martin Liska <mliska@suse.cz>
4672 * tree.c (record_node_allocation_statistics): Use
4673 get_stats_node_kind.
4674 (get_stats_node_kind): New function extracted from
4675 record_node_allocation_statistics.
4676 (free_node): Use get_stats_node_kind.
4678 2018-03-14 Richard Biener <rguenther@suse.de>
4680 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4681 that the value-set of ANTIC_IN doesn't grow.
4684 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4686 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4687 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4689 2018-03-14 Julia Koval <julia.koval@intel.com>
4691 * config.gcc (icelake-client, icelake-server): New.
4693 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4694 (initial_ix86_arch_features): Ditto.
4695 (PTA_SKYLAKE): Add SGX.
4696 (PTA_ICELAKE): Remove.
4697 (PTA_ICELAKE_CLIENT): New.
4698 (PTA_ICELAKE_SERVER): New.
4699 (ix86_option_override_internal): Split up icelake on icelake client and
4701 (get_builtin_code_for_version): Ditto.
4702 (fold_builtin_cpu): Ditto.
4703 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4704 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4705 * config/i386/i386.h (processor_type): Ditto.
4706 * doc/invoke.texi: Ditto.
4708 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4711 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4712 INTEGER_CST offset, add it together with bitpos / 8 and
4713 sign extend based on POINTER_SIZE.
4717 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4720 * config/i386/constraints.md (Yc): New register constraint.
4721 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4722 Use Yc constraint for alternative 2 of operand 0. Remove
4723 preferred_for_speed attribute.
4725 2018-03-14 Richard Biener <rguenther@suse.de>
4727 PR tree-optimization/84830
4728 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4729 with the old one to avoid oscillations.
4731 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4734 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4736 (assign_by_spills): Return a flag of reload assignment failure.
4737 Do not process the reload assignment failures. Do not spill other
4738 reload pseudos if they has the same reg class. Update n if
4740 (lra_assign): Add a return arg. Set up from the result of
4741 assign_by_spills call.
4742 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4743 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4744 usage_insns if it is not NULL.
4745 (spill_hard_reg_in_range): New function.
4746 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4747 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4748 function prototypes.
4749 (lra_assign): Change prototype.
4750 * lra.c (lra): Add code to deal with fails by splitting hard reg
4753 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4755 * config/riscv/riscv.opt (mrelax): New option.
4756 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4757 "norelax" when riscv_mrelax is disabled.
4758 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4760 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4763 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4764 reassociation for int modes.
4766 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4768 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4769 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4771 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4772 * config/aarch64/aarch64-sve.md
4773 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4774 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4775 (*extend<mode><Vwide>2): Rename to...
4776 (aarch64_sve_extend<mode><Vwide>2): ...this.
4777 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4778 renaming the old pattern to...
4779 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4781 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4782 define_expand, renaming the old pattern to...
4783 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4784 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4785 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4786 account when deciding which SVE instruction the optab should use.
4787 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4789 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4791 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4792 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4793 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4794 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4795 (tlsdesc_small_<mode>): Turn a define_expand and use
4796 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4797 (tlsdesc_small_advsimd_<mode>): ...this.
4798 (tlsdesc_small_sve_<mode>): New pattern.
4800 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4802 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4803 (UNSPEC_UMUL_HIGHPART): New constants.
4804 (MUL_HIGHPART): New int iteraor.
4805 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4806 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4808 (*<su>mul<mode>3_highpart): New define_insn.
4810 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4813 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4816 2018-03-13 Martin Liska <mliska@suse.cz>
4819 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4821 (sem_item_optimizer::~sem_item_optimizer): Release it.
4822 (sem_item_optimizer::merge_classes): Register variable aliases.
4823 (sem_item_optimizer::fixup_pt_set): New function.
4824 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4825 * ipa-icf.h: Declare new variables and functions.
4827 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4830 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4831 integer_pow2p@2 and test integer_pow2p in condition.
4832 (A < 0 ? C : 0): Similarly for @1.
4835 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4836 characters starting at p contain '\0' character, don't look beyond
4840 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4841 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4844 * reg-stack.c (change_stack): Change update_end var from int to
4845 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4846 also call set_block_for_insn on the newly added insns and rescan.
4849 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4850 on the last operand.
4853 * tree.c (stabilize_reference_1): Return save_expr (e) for
4854 STATEMENT_LIST even if it doesn't have side-effects.
4856 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4858 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4860 2018-03-12 Renlin Li <renlin.li@arm.com>
4862 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4863 aarch64_output_scalar_simd_mov_immediate.
4865 2018-03-12 Martin Sebor <msebor@redhat.com>
4867 PR tree-optimization/83456
4868 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4869 for perfectly overlapping calls to memcpy.
4870 (gimple_fold_builtin_memory_chk): Same.
4871 (gimple_fold_builtin_strcpy): Handle no-warning.
4872 (gimple_fold_builtin_stxcpy_chk): Same.
4873 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4875 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4877 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4878 parameter. Use it for SFmode.
4879 (rs6000_function_arg_advance_1): Adjust.
4880 (rs6000_function_arg): Adjust.
4881 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4883 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4885 PR rtl-optimization/84169
4886 PR rtl-optimization/84780
4887 * combine.c (can_combine_p): Check for a 2-insn combination whether
4888 the destination register is used between the two insns, too.
4890 2018-03-12 Richard Biener <rguenther@suse.de>
4892 PR tree-optimization/84803
4893 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4894 for refs DR analysis didn't process.
4896 2018-03-12 Richard Biener <rguenther@suse.de>
4898 PR tree-optimization/84777
4899 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4900 force-vectorize loops ignore whether we are optimizing for size.
4902 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4904 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4905 (TARGET_MD_ASM_ADJUST): Define.
4907 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4908 Kito Cheng <kito.cheng@gmail.com>
4909 Chung-Ju Wu <jasonwucj@gmail.com>
4911 * config/nds32/nds32.c (nds32_compute_stack_frame,
4912 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4913 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4914 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4915 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4916 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4917 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4918 * config/nds32/nds32.md (prologue, epilogue): Use macro
4919 NDS32_V3PUSH_AVAILABLE_P to do checking.
4921 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4924 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4925 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4926 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4927 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4928 addition of most attributes on !orig_type_die or the attribute not
4929 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4931 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4932 Chung-Ju Wu <jasonwucj@gmail.com>
4934 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4936 * config/nds32/nds32.opt (mvh): New option.
4938 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4939 Chung-Ju Wu <jasonwucj@gmail.com>
4941 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4943 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4944 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4947 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4948 Chung-Ju Wu <jasonwucj@gmail.com>
4950 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4952 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4953 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4955 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4956 Kito Cheng <kito.cheng@gmail.com>
4957 Chung-Ju Wu <jasonwucj@gmail.com>
4959 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4960 UNSPEC_FFMISM and UNSPEC_FLMISM.
4961 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4962 for ffb, ffmism and flmism.
4963 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4964 (unspec_ffmism): Ditto.
4965 (unspec_flmism): Ditto.
4966 (nds32_expand_builtin_impl): Check if string extension is available.
4967 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4968 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4970 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4973 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4976 * lra-assigns.c (assign_by_spills): Return a flag of reload
4977 assignment failure. Do not process the reload assignment
4978 failures. Do not spill other reload pseudos if they has the same
4980 (lra_assign): Add a return arg. Set up from the result of
4981 assign_by_spills call.
4982 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4983 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4984 usage_insns if it is not NULL.
4985 (spill_hard_reg_in_range): New function.
4986 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4987 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4988 function prototypes.
4989 (lra_assign): Change prototype.
4990 * lra.c (lra): Add code to deal with fails by splitting hard reg
4993 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4996 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4998 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5001 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5002 (dw_val_node): Add val_symbolic_view.
5003 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5004 (symview_upper_bound): New.
5005 (new_line_info_table): Initialize symviews_since_reset.
5006 (dwarf2out_source_line): Count symviews_since_reset and set
5007 symview_upper_bound.
5008 (dw_val_equal_p): Handle symview.
5009 (add_AT_symview): New.
5010 (print_dw_val): Handle symview.
5011 (attr_checksum, attr_checksum_ordered): Likewise.
5012 (same_dw_val_p, size_of_die): Likewise.
5013 (value_format, output_die): Likewise.
5014 (add_high_low_attributes): Use add_AT_symview for entry_view.
5015 (dwarf2out_finish): Reset symview_upper_bound, clear
5018 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5021 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5022 Add strict argument and use it.
5023 (rs6000_split_multireg_move): Update for new strict argument.
5024 (mem_operand_gpr): Disallow all non-offsettable addresses.
5025 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5027 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5030 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5031 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5032 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5035 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5036 decl, use remap_type if we want to use the type.
5038 2018-03-09 Martin Sebor <msebor@redhat.com>
5040 PR tree-optimization/84526
5041 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5043 (builtin_access::generic_overlap): Be prepared to handle non-array
5046 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5048 PR rtl-optimization/84682
5049 * lra-constraints.c (process_address_1): Check is_address flag
5050 for address constraints.
5051 (process_alt_operands): Likewise.
5052 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5053 preprocess_constraints.
5054 * recog.h (preprocess_constraints): Add oploc parameter.
5056 * recog.c (preprocess_constraints): Test address_operand for
5057 CT_ADDRESS constraints.
5059 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5062 * lra-assigns.c (assign_by_spills): Return a flag of reload
5063 assignment failure. Do not process the reload assignment
5064 failures. Do not spill other reload pseudos if they has the same
5066 (lra_assign): Add a return arg. Set up from the result of
5067 assign_by_spills call.
5068 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5069 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5070 usage_insns if it is not NULL.
5071 (spill_hard_reg_in_range): New function.
5072 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5073 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5074 function prototypes.
5075 (lra_assign): Change prototype.
5076 * lra.c (lra): Add code to deal with fails by splitting hard reg
5079 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5082 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5083 Accept complain bool parameter. Only emit errors if it is true.
5084 (arm_parse_cpu_option_name): Likewise.
5085 (arm_target_thumb_only): Adjust callers of the above.
5086 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5087 prototype to take a default true bool parameter.
5088 (arm_parse_arch_option_name): Likewise.
5090 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5091 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5095 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5096 * configure: Regenerate.
5097 * configure.ac ("linker --version-script option"): New.
5098 ("linker soname option"): New.
5100 2018-03-09 Richard Biener <rguenther@suse.de>
5102 PR tree-optimization/84775
5103 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5104 immediate uses of predicate stmts and mark them modified.
5107 PR tree-optimization/84178
5108 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5110 (version_loop_for_if_conversion): Delay update_ssa call.
5111 (tree_if_conversion): Delay update_ssa until after predicate
5114 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5117 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5118 when the function accesses prior frames.
5120 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5123 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5124 gen_llsym, otherwise call maybe_gen_llsym.
5127 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5128 has ',' character inside of it.
5130 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5133 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5134 as clobbering CC_REGNUM.
5136 2018-03-08 Richard Biener <rguenther@suse.de>
5139 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5140 (follow_copies_to_constant): Do not follow SSA names registered
5143 2018-03-08 Richard Biener <rguenther@suse.de>
5145 PR tree-optimization/84178
5146 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5148 (version_loop_for_if_conversion): Delay update_ssa call.
5149 (tree_if_conversion): Delay update_ssa until after predicate
5152 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5154 PR tree-optimization/84178
5155 * tree-if-conv.c (release_bb_predicate): Remove the
5156 the assertion that the stmts have NULL use_ops.
5157 Discard the statements, asserting that they haven't
5158 yet been added to a BB.
5160 2018-03-08 Richard Biener <rguenther@suse.de>
5162 PR tree-optimization/84746
5163 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5164 (phi_translate): Pass in destination ANTIC_OUT set.
5165 (phi_translate_1): Likewise. For a simplified result lookup
5166 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5167 (phi_translate_set): Adjust.
5168 (do_pre_regular_insertion): Likewise.
5169 (do_pre_partial_partial_insertion): Likewise.
5171 2018-03-08 Martin Liska <mliska@suse.cz>
5173 PR gcov-profile/84735
5174 * doc/gcov.texi: Document usage of profile files.
5175 * gcov-io.h: Document changes in the format.
5177 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5181 * dwarf2out.c (struct dw_line_info_table): Update comments for
5183 (FORCE_RESET_NEXT_VIEW): New.
5184 (FORCE_RESETTING_VIEW_P): New.
5185 (RESETTING_VIEW_P): Check for -1 too.
5186 (ZERO_VIEW_P): Likewise.
5187 (new_line_info_table): Force-reset next view.
5188 (dwarf2out_begin_function): Likewise.
5189 (dwarf2out_source_line): Simplify zero_view_p initialization.
5190 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5191 view directly. Omit view when omitting .loc at line 0.
5193 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5195 PR tree-optimization/84740
5196 * tree-switch-conversion.c (process_switch): Call build_constructors
5197 only if info.phi_count is non-zero.
5199 PR tree-optimization/84739
5200 * tree-tailcall.c (find_tail_calls): Check call arguments against
5201 DECL_ARGUMENTS (current_function_decl) rather than
5202 DECL_ARGUMENTS (func) when checking for tail recursion.
5204 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5206 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5207 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5208 Volker Reichelt's entry and add entries for people that perform
5209 GCC fuzzy testing and report numerous bugs.
5211 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5214 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5215 readonly data in sdata, if that is disabled.
5216 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5217 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5218 -mreadonly-in-sdata option.
5220 2018-03-07 Martin Sebor <msebor@redhat.com>
5222 PR tree-optimization/84468
5223 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5224 basic block when looking for nul assignment.
5226 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5229 * except.h (output_function_exception_table): Adjust prototype.
5230 * except.c (output_function_exception_table): Remove FNNAME parameter
5231 and add SECTION parameter. Ouput one part of the table at a time.
5232 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5233 the first part of the exception table and emit unwind directives.
5234 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5235 (i386_pe_seh_cold_init): Likewise.
5236 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5237 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5238 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5239 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5240 * config/i386/winnt.c: Include except.h.
5241 (struct seh_frame_state): Add reg_offset, after_prologue and
5242 in_cold_section fields.
5243 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5244 (i386_pe_seh_cold_init): New function.
5245 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5246 to seh->in_cold_section.
5247 (seh_emit_push): Record the offset of the push.
5248 (seh_emit_save): Record the offet of the save.
5249 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5250 Test seh->after_prologue to disregard the epilogue.
5251 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5252 (i386_pe_end_cold_function): New function.
5254 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5257 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5258 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5261 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5263 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5264 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5267 * multiple_target.c: Include tree-inline.h and intl.h.
5268 (expand_target_clones): Diagnose and fail if node->definition and
5269 !tree_versionable_function_p (node->decl).
5271 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5273 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5275 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5276 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5277 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5279 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5282 * combine.c (try_combine): Use reg_or_subregno instead of handling
5283 just paradoxical SUBREGs and REGs.
5285 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5287 * config/arc/arc.c (arc_finalize_pic): Remove function.
5288 (arc_must_save_register): We use single base PIC register, remove
5289 checks to save/restore the PIC register.
5290 (arc_expand_prologue): Likewise.
5291 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5293 (arc_verify_short): Likewise.
5294 (arc_attr_type): Likewise.
5295 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5296 (walk_stores): Likewise.
5297 (arc_address_cost): Make it static.
5298 (arc_verify_short): Likewise.
5299 (branch_dest): Likewise.
5300 (arc_attr_type): Likewise.
5301 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5302 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5303 (arc_final_prescan_insn): Remove inserting the nops due to
5304 hardware hazards. It is done in reorg step.
5305 (insn_length_variant_t): Remove.
5306 (insn_length_parameters_t): Likewise.
5307 (arc_insn_length_parameters): Likewise.
5308 (arc_get_insn_variants): Likewise.
5309 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5311 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5314 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5317 PR tree-optimization/84687
5318 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5320 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5322 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5324 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5325 Rename to ppc_speculation_barrier.
5326 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5327 __builtin_ppc_speculation_barrier.
5329 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5332 * combine.c (combine_simplify_rtx): Don't try to simplify if
5333 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5336 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5338 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5339 to 32 bytes when compiling for POWER9.
5341 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5344 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5345 regparm >= 3 with no arg reg available also for calls with
5346 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5349 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5351 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5353 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5356 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5358 2018-03-05 Richard Biener <rguenther@suse.de>
5360 PR tree-optimization/84486
5361 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5362 When inserting a __builtin_assume_aligned call set the LHS
5363 SSA name alignment info accordingly.
5365 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5367 PR tree-optimization/84114
5368 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5369 Avoid reassociation of FLOAT_MODE addition.
5371 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5373 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5374 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5375 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5376 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5377 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5378 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5379 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5381 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5382 __builtin_ia32_wbinvd): New builtins.
5383 (SPECIAL_ARGS2): New.
5384 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5385 (SPECIAL_ARGS2): New.
5386 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5387 (ix86_valid_target_attribute_inner_p): Ditto.
5388 (ix86_init_mmx_sse_builtins): Add special_args2.
5389 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5390 TARGET_WBNOINVD_P): New.
5391 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5392 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5393 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5394 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5395 * config/i386/pconfigintrin.h: New file.
5396 * config/i386/wbnoinvdintrin.h: Ditto.
5397 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5398 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5400 2018-03-05 Richard Biener <rguenther@suse.de>
5402 PR tree-optimization/84670
5403 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5405 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5406 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5407 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5408 doesn't grow if all successors (recursively) were visited at least
5411 2018-03-05 Richard Biener <rguenther@suse.de>
5413 PR tree-optimization/84650
5414 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5415 if executed in the loop pipeline.
5417 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5419 * doc/configfiles.texi (Configuration Files): Move info about
5420 conditionalizing $target-protos.h to...
5421 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5422 differs from $target-protos.h.
5424 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5425 Chung-Ju Wu <jasonwucj@gmail.com>
5427 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5428 * config/nds32/nds32-multiple.md (setmemsi): Define.
5429 * config/nds32/nds32-memory-manipulation.c
5430 (nds32_gen_dup_4_byte_to_word_value): New.
5431 (emit_setmem_word_loop): New.
5432 (emit_setmem_byte_loop): New.
5433 (nds32_expand_setmem_loop): New.
5434 (nds32_expand_setmem_loop_v3m): New.
5435 (nds32_expand_setmem_unroll): New.
5436 (nds32_expand_setmem): New.
5438 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5439 Chung-Ju Wu <jasonwucj@gmail.com>
5441 * config/nds32/nds32-memory-manipulation.c
5442 (nds32_emit_load_store): New.
5443 (nds32_emit_post_inc_load_store): New.
5444 (nds32_emit_mem_move): New.
5445 (nds32_emit_mem_move_block): New.
5446 (nds32_expand_movmemsi_loop_unknown_size): New.
5447 (nds32_expand_movmemsi_loop_known_size): New.
5448 (nds32_expand_movmemsi_loop): New.
5449 (nds32_expand_movmemsi_unroll): New.
5450 (nds32_expand_movmemqi): Rename ...
5451 (nds32_expand_movmemsi): ... to this.
5452 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5453 (movmemsi): ... to this.
5454 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5455 (nds32_expand_movmemsi): ... to this.
5457 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5458 Monk Chiang <sh.chiang04@gmail.com>
5459 Chung-Ju Wu <jasonwucj@gmail.com>
5461 * config/nds32/nds32-protos.h
5462 (nds32_expand_load_multiple): New arguments.
5463 (nds32_expand_store_multiple): Ditto.
5464 (nds32_valid_multiple_load_store): Rename ...
5465 (nds32_valid_multiple_load_store_p): ... to this.
5466 * config/nds32/nds32-memory-manipulation.c
5467 (nds32_expand_load_multiple): Refine implementation.
5468 (nds32_expand_store_multiple): Ditto.
5469 * config/nds32/nds32-multiple.md
5470 (load_multiple): Update nds32_expand_load_multiple interface.
5471 (store_multiple): Update nds32_expand_store_multiple interface.
5472 * config/nds32/nds32-predicates.c
5473 (nds32_valid_multiple_load_store): Rename ...
5474 (nds32_valid_multiple_load_store_p): ... to this and refine
5476 * config/nds32/predicates.md
5477 (nds32_load_multiple_and_update_address_operation): New predicate.
5478 (nds32_store_multiple_and_update_address_operation): New predicate.
5480 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5481 Chung-Ju Wu <jasonwucj@gmail.com>
5483 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5484 (combo): New attribute.
5485 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5487 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5489 * config/nds32/nds32.opt: Change -mcmodel= default value.
5491 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5492 Monk Chiang <sh.chiang04@gmail.com>
5493 Chung-Ju Wu <jasonwucj@gmail.com>
5495 * config/nds32/constants.md (unspec_element): New enum.
5496 * config/nds32/constraints.md (Umw): New constraint.
5497 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5498 * config/nds32/nds32-intrinsic.md: Likewise.
5499 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5500 (nds32_valid_smw_lwm_base_p): New.
5501 (nds32_output_smw_single_word): New.
5502 (nds32_output_lmw_single_word): New.
5503 (nds32_expand_unaligned_load): New.
5504 (nds32_expand_unaligned_store): New.
5505 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5506 (nds32_output_smw_single_word): Declare.
5507 (nds32_output_lmw_single_word): Declare.
5508 (nds32_expand_unaligned_load): Declare.
5509 (nds32_expand_unaligned_store): Declare.
5510 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5511 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5512 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5513 NDS32_BUILTIN_UASTORE_DW.
5514 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5517 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5518 Kito Cheng <kito.cheng@gmail.com>
5519 Chung-Ju Wu <jasonwucj@gmail.com>
5521 * config/nds32/nds32-intrinsic.c
5522 (nds32_expand_builtin_null_ftype_reg): Delete.
5523 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5524 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5525 (nds32_read_argument): New.
5526 (nds32_legitimize_target): Ditto.
5527 (nds32_legitimize_argument): Ditto.
5528 (nds32_check_constant_argument): Ditto.
5529 (nds32_expand_unop_builtin): Ditto.
5530 (nds32_expand_unopimm_builtin): Ditto.
5531 (nds32_expand_binop_builtin): Ditto.
5532 (nds32_builtin_decl_impl): Ditto.
5533 (builtin_description): Ditto.
5534 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5535 (nds32_init_builtins_impl): Ditto.
5536 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5537 (nds32_builtin_decl): New.
5538 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5539 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5541 2018-03-02 Jeff Law <law@redhat.com>
5543 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5544 (redundant_insn, fill_simple_delay_slots): Likewise.
5545 (fill_slots_from_thread): Likewise.
5546 * resource.c (mark_referenced_resources): Likewise.
5547 (mark_set_resources, find_dead_or_set_registers): Likewise.
5549 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5551 * substring-locations.h (format_warning_va): Formatting fix for
5553 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5555 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5556 * substring-locations.c: Include intl.h.
5557 (format_warning_va): Turned into small wrapper around
5558 format_warning_n_va, renamed to ...
5559 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5560 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5562 (format_warning_at_substring_n): New function.
5563 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5564 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5565 format_warning_at_substring with just a shorter name instead of
5566 const function pointer.
5567 (fmtwarn_n): New function.
5568 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5569 appropriate, get rid of all the fmtstr temporaries, move conditionals
5570 with G_() wrapped string literals directly into fmtwarn arguments,
5571 cast dir.len to (int), formatting fixes.
5573 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5575 * doc/invoke.texi: Remove "Cilk Plus" references.
5577 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5578 Richard Biener <rguenther@suse.de>
5581 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5582 for error or warning attributes if CALL_FROM_THUNK_P is set.
5585 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5588 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5589 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5592 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5593 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5595 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5596 instead of -1U in last predictors element's probability member.
5598 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5601 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5602 arguments if they are comparable.
5604 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5606 PR tree-optimization/84634
5607 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5608 masks and masked_loop_p with a single loop_masks, making sure it's
5609 null for bb vectorization.
5611 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5613 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5614 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5615 loop->force_vectorize to check whether there is no alias.
5617 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5620 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5622 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5624 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5625 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5626 instead of a loop around prev_real_insn.
5627 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5631 * config/i386/i386.c (ix86_print_operand): Use conditional
5632 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5635 2018-03-02 Richard Biener <rguenther@suse.de>
5637 PR tree-optimization/84427
5638 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5639 (bitmap_set_subtract_values): Rewrite to handle multiple
5642 (prune_clobbered_mems): Likewise.
5643 (phi_translate): Take edge instead of pred/phiblock.
5644 (phi_translate_1): Likewise.
5645 (phi_translate_set): Likewise. Insert all translated
5646 exprs for a value into the set, keeping possibly multiple
5647 expressions per value.
5648 (compute_antic_aux): Adjust for phi_translate changes.
5649 When intersecting union the expressions and prune those
5650 not in the final value set, keeping possibly multiple
5651 expressions per value. Do not use value-insertion
5652 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5653 all expressions. Add verification that the value-sets
5654 only shrink during iteration.
5655 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5656 (do_pre_regular_insertion): Likewise.
5657 (do_pre_partial_partial_insertion): Likewise.
5659 2018-03-02 Richard Biener <rguenther@suse.de>
5662 * config/darwin.c (saved_debug_info_level): New static global.
5663 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5664 (darwin_asm_lto_end): Restore debug info generation settings.
5666 2018-03-01 Martin Liska <mliska@suse.cz>
5669 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5672 2018-03-01 Richard Biener <rguenther@suse.de>
5675 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5676 type annotation with LTO when debug was disabled at compile-time.
5678 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5680 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5682 (mips_final_postscan_insn): Likewise.
5684 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5686 PR rtl-optimization/84528
5687 * alias.c (init_alias_target): Add commentary.
5688 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5689 a unique base value if the frame pointer is not eliminated
5690 to the stack pointer.
5692 2018-03-01 Tom de Vries <tom@codesourcery.com>
5694 PR rtl-optimization/83327
5695 * lra-int.h (hard_regs_spilled_into): Declare.
5696 * lra.c (hard_regs_spilled_into): Define.
5697 (init_reg_info): Init hard_regs_spilled_into.
5698 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5699 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5700 (process_bb_lives): Handle hard_regs_spilled_into.
5701 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5702 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5704 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5706 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5707 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5708 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5709 * config/rs6000/aix72.h: New file.
5711 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5713 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5714 instead of warning_at with conditional singular and plural messages
5718 * stor-layout.c (update_alignment_for_field): For
5719 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5720 && !DECL_PACKED (field), do the alignment update, just use
5721 only desired_align instead of MAX (type_align, desired_align)
5723 (place_field): Don't do known_align < desired_align handling
5724 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5725 is non-NULL, instead do it after rli->prev_field handling and
5726 only if not within a bitfield word. For DECL_PACKED (field)
5727 use type_align of BITS_PER_UNIT.
5729 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5731 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5732 superfluous parentheses and trailing spaces.
5734 2018-02-28 Richard Biener <rguenther@suse.de>
5736 PR tree-optimization/84584
5737 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5738 SCoPs with fake exit edge.
5740 2018-02-28 Martin Liska <mliska@suse.cz>
5743 * timevar.c (timer::print): Fix format to properly print 100%
5746 2018-02-28 Richard Biener <rguenther@suse.de>
5749 * genmatch.c (capture_info::walk_match): Do not mark
5750 captured expressions without operands as expr_p given
5751 they act more like predicates and should be subject to
5752 "lost tail" side-effect preserving.
5754 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5756 PR rtl-optimization/81611
5757 * auto-inc-dec.c (attempt_change): Move dead note from
5758 mem_insn if it's the next use of regno
5759 (find_address): Take address use of reg holding
5760 non-incremented value. Add parm to limit search to the named
5762 (merge_in_block): Attempt to use a mem insn that is the next
5763 use of the original regno.
5765 2018-02-27 Martin Sebor <msebor@redhat.com>
5768 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5769 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5771 2018-02-27 Martin Sebor <msebor@redhat.com>
5773 PR translation/84207
5774 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5775 n argument to unsigned HOST_WIDE_INT.
5776 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5777 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5778 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5779 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5781 2018-02-27 Richard Biener <rguenther@suse.de>
5783 PR tree-optimization/84512
5784 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5785 Do not use the estimate returned from record_stmt_cost for
5786 the scalar iteration cost but sum properly using add_stmt_cost.
5788 2018-02-27 Richard Biener <rguenther@suse.de>
5790 PR tree-optimization/84466
5791 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5792 Adjust last change to less strictly validate use operands.
5794 2018-02-27 Martin Liska <mliska@suse.cz>
5796 PR gcov-profile/84548
5797 * gcov.c (process_file): Allow partial overlap and consider it
5798 also as group functions.
5799 (output_lines): Properly calculate range of lines for a group.
5801 2018-02-27 Martin Liska <mliska@suse.cz>
5803 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5804 'ggc' suffixes. Change first column width.
5805 (timer::print): Fix formatting of the column.
5807 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5809 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5810 preserve inline entry blocks for the sake of debug inline
5811 entry point markers alone.
5812 (remove_unused_locals): Suggest in comments a better place to
5813 force the preservation of inline entry blocks that are
5814 otherwise unused, but do not preserve them.
5816 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5818 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5820 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5823 * config/i386/constraints.md (Bs): Replace
5824 ix86_indirect_branch_register with
5825 TARGET_INDIRECT_BRANCH_REGISTER.
5827 * config/i386/i386.md (indirect_jump): Likewise.
5828 (tablejump): Likewise.
5829 (*sibcall_memory): Likewise.
5830 (*sibcall_value_memory): Likewise.
5831 Peepholes of indirect call and jump via memory: Likewise.
5832 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5833 (*sibcall_value_GOT_32): Likewise.
5834 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5835 (GOT_memory_operand): Likewise.
5836 (call_insn_operand): Likewise.
5837 (sibcall_insn_operand): Likewise.
5838 (GOT32_symbol_operand): Likewise.
5839 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5841 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5843 PR rtl-optimization/83496
5844 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5845 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5846 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5847 redundant insn, if any.
5848 (relax_delay_slots): Likewise.
5849 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5851 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5853 PR tree-optimization/83965
5854 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5855 that grouped statements are part of a reduction chain. Return
5856 true if the statement is not marked as a reduction itself but
5858 (vect_recog_dot_prod_pattern): Don't check whether the statement
5859 is part of a group here.
5860 (vect_recog_sad_pattern): Likewise.
5861 (vect_recog_widen_sum_pattern): Likewise.
5863 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5866 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5868 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5871 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5873 (ix86_output_indirect_function_return): New prototype.
5874 (ix86_split_simple_return_pop_internal): Likewise.
5875 * config/i386/i386.c (indirect_return_via_cx): New.
5876 (indirect_return_via_cx_bnd): Likewise.
5877 (indirect_thunk_name): Handle return va CX_REG.
5878 (output_indirect_thunk_function): Create alias for
5879 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5880 (ix86_output_indirect_jmp): Remove the bool argument.
5881 (ix86_output_indirect_function_return): New function.
5882 (ix86_split_simple_return_pop_internal): Likewise.
5883 * config/i386/i386.md (*indirect_jump): Don't pass false
5884 to ix86_output_indirect_jmp.
5885 (*tablejump_1): Likewise.
5886 (simple_return_pop_internal): Change it to define_insn_and_split.
5887 Call ix86_split_simple_return_pop_internal to split it for
5889 (simple_return_indirect_internal): Call
5890 ix86_output_indirect_function_return instead of
5891 ix86_output_indirect_jmp.
5893 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5896 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5897 memset and value initialization afterwards.
5899 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5901 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5903 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5906 * common/config/aarch64/aarch64-common.c
5907 (aarch_option_optimization_table[]): Switch
5908 off fomit-frame-pointer
5910 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5911 Chung-Ju Wu <jasonwucj@gmail.com>
5913 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5915 (store_multiple): Ditto.
5917 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5919 * config.gcc: Add --with-cpu support for nds32 target.
5920 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5921 * config/nds32/nds32.opt: Add -mcpu= option.
5923 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5925 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5926 isel=yes): Warn for these deprecated options.
5928 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5930 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5931 ISA_2_5_MASKS_EMBEDDED.
5933 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5935 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5936 p->max as pointers rather than using iterative_hash_expr.
5938 2018-02-23 Carl Love <cel@us.ibm.com>
5940 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5941 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5943 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5944 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5945 P8V_BUILTIN_VEC_VUNSIGNED2.
5947 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5950 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5951 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5953 * lra-constraints.c (curr_insn_transform): Set up
5954 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5956 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5957 * lra-eliminations.c (spill_pseudos): Ditto.
5958 (process_insn_for_elimination): Ditto.
5959 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5960 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5961 LRA_NON_CLOBBERED_ALT.
5963 2018-02-22 Martin Sebor <msebor@redhat.com>
5965 PR tree-optimization/84480
5966 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5967 to maybe_diag_stxncpy_trunc. Call it.
5968 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5969 from gimple_fold_builtin_strcpy. Print inlining stack.
5970 (handle_builtin_stxncpy): Print inlining stack.
5971 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5973 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5976 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5977 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5978 and -fcheck-pointer-bounds are used together.
5979 (indirect_thunk_prefix): New enum.
5980 (indirect_thunk_need_prefix): New function.
5981 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5982 "_nt" instead of "_bnd" for NOTRACK prefix.
5983 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5984 (output_indirect_thunk_function): Likewise.
5986 (ix86_code_end): Update output_indirect_thunk_function calls.
5987 (ix86_output_indirect_branch_via_reg): Replace
5988 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5989 (ix86_output_indirect_branch_via_push): Likewise.
5990 (ix86_output_function_return): Likewise.
5991 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5992 incompatible with -fcf-protection=branch and
5993 -fcheck-pointer-bounds.
5995 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5998 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5999 Change gcc_assert call to output_operand_lossage.
6001 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6003 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6005 2018-02-22 DJ Delorie <dj@redhat.com>
6006 Sebastian Perta <sebastian.perta@renesas.com>
6007 Oleg Endo <olegendo@gcc.gnu.org>
6009 * config/rx/rx.c (rx_rtx_costs): New function.
6010 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6012 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6014 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6016 2018-02-22 Martin Liska <mliska@suse.cz>
6019 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6020 Add "native" as a possible value.
6022 2018-02-22 Martin Liska <mliska@suse.cz>
6025 * config/i386/i386.c (ix86_option_override_internal):
6026 Add "native" as a possible value for -march and -mtune.
6028 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6031 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6032 to all type variants.
6034 PR tree-optimization/84503
6035 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6036 width as info->bitpos + info->bitsize - start.
6037 (merged_store_group::merge_overlapping): Simplify width computation.
6038 (check_no_overlap): New function.
6039 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6040 start + width and last_order of the group, fail if check_no_overlap
6042 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6043 to group if check_no_overlap fails.
6045 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6047 * config/rs6000/altivec.md: Delete contraint arguments to
6048 define_expand, define_split, and define_peephole2, and in
6049 define_insn_and_split if always unused.
6050 * config/rs6000/darwin.md: Ditto.
6051 * config/rs6000/dfp.md: Ditto.
6052 * config/rs6000/rs6000.md: Ditto.
6053 * config/rs6000/sync.md: Ditto.
6054 * config/rs6000/vector.md: Ditto.
6055 * config/rs6000/vsx.md: Ditto.
6057 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6059 * config/rs6000/altivec.md: Write output control strings as braced
6060 blocks instead of double-quoted strings.
6061 * config/rs6000/darwin.md: Ditto.
6062 * config/rs6000/rs6000.md: Ditto.
6063 * config/rs6000/vector.md: Ditto.
6064 * config/rs6000/vsx.md: Ditto.
6066 2018-02-21 Jason Merrill <jason@redhat.com>
6068 PR c++/84314 - ICE with templates and fastcall attribute.
6069 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6071 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6073 * ipa-cp.c (determine_versionability): Fix comment typos.
6075 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6078 * ipa-cp.c (determine_versionability): Do not version functions caling
6081 2018-02-21 Martin Liska <mliska@suse.cz>
6084 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6085 Add "native" as a possible value.
6086 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6087 the macro when native cpu detection is available.
6089 2018-02-21 Martin Liska <mliska@suse.cz>
6092 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6093 Add "native" as a possible value.
6094 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6095 when native cpu detection is available.
6097 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6098 Martin Sebor <msebor@redhat.com>
6100 PR tree-optimization/84478
6101 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6103 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6104 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6105 support which is conservatively correct, for 2 only stay conservative
6106 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6107 argument to the 2 argument get_range_strlen, adjust 6 arg
6108 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6110 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6111 (gimple_fold_builtin_strlen): Pass true as last argument to
6114 2018-02-20 Martin Sebor <msebor@redhat.com>
6117 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6118 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6119 (builtin_memref::builtin_memref): Factor out parts into
6120 set_base_and_offset and call it.
6122 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6125 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6126 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6127 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6128 search at the associated MODE_INT.
6130 2018-02-20 Jeff Law <law@redhat.com>
6133 PR tree-optimization/81592
6135 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6136 for range data rather than using global data.
6137 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6138 range data rather than using global data.
6139 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6140 pass it to children as needed.
6141 (struct directive::fmtresult): Similarly.
6142 (struct directive::set_width): Similarly.
6143 (struct directive::set_precision): Similarly.
6144 (format_integer, format_directive, parse_directive): Similarly.
6145 (format_none): Accept unnamed vr_values parameter.
6146 (format_percent, format_floating, format_character): Similarly.
6147 (format_string, format_plain): Similarly.
6148 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6149 the EVRP range analyzer for range data rather than using global data.
6150 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6151 gimple-ssa-evrp-analyze.h
6152 (class sprintf_dom_walker): Add after_dom_children member function.
6153 Add evrp_range_analyzer member.
6154 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6155 range analyzer as needed.
6156 (sprintf_dom_walker::after_dom_children): New member function.
6157 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6159 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6160 (evrp_range_analyzer::pop_to_marker): Likewise.
6162 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6164 PR tree-optimization/84419
6165 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6166 with the required type if its current type is compatible but
6169 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6172 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6173 after vectorization.
6175 2018-02-20 Martin Liska <mliska@suse.cz>
6178 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6179 possible values if we don't have a hint.
6181 2018-02-20 Martin Liska <mliska@suse.cz>
6185 * final.c (shorten_branches): Build align_tab array with one
6187 * opts.c (finish_options): Add alignment option limit check.
6188 (MAX_CODE_ALIGN): Likewise.
6189 (MAX_CODE_ALIGN_VALUE): Likewise.
6190 * doc/invoke.texi: Document maximum allowed option value for
6191 all -falign-* options.
6193 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6196 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6197 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6198 * var-tracking.c (emit_note_insn_var_location): Remove all references
6199 to NOTE_INSN_CALL_ARG_LOCATION.
6200 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6201 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6202 Use copy_rtx_if_shared.
6203 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6204 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6205 (dwarf2out_var_location): Remove handling of
6206 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6208 * final.c (final_scan_insn): Remove all references to
6209 NOTE_INSN_CALL_ARG_LOCATION.
6210 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6211 before dumping final insns.
6212 * except.c (emit_note_eh_region_end): Remove all references to
6213 NOTE_INSN_CALL_ARG_LOCATION.
6214 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6215 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6216 * config/arc/arc.c (hwloop_optimize): Likewise.
6217 * config/arm/arm.c (create_fix_barrier): Likewise.
6218 * config/s390/s390.c (s390_chunkify_start): Likewise.
6219 * config/sh/sh.c (find_barrier): Likewise.
6220 * config/i386/i386.c (rest_of_insert_endbranch,
6221 ix86_seh_fixup_eh_fallthru): Likewise.
6222 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6223 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6224 * config/frv/frv.c (frv_function_prologue): Likewise.
6225 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6227 (note_outside_basic_block_p): Remove all references to
6228 NOTE_INSN_CALL_ARG_LOCATION.
6229 * gengtype.c (adjust_field_rtx_def): Likewise.
6230 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6232 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6233 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6236 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6239 PR tree-optimization/84452
6240 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6241 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6244 2018-02-19 Martin Liska <mliska@suse.cz>
6247 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6249 2018-02-19 Martin Liska <mliska@suse.cz>
6250 Richard Sandiford <richard.sandiford@linaro.org>
6252 PR tree-optimization/82491
6253 * gimple-fold.c (get_base_constructor): Make earlier bail out
6256 2018-02-19 Carl Love <cel@us.ibm.com>
6258 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6259 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6261 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6262 P8V_BUILTIN_VEC_NEG.
6264 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6266 * config/rl78/rl78.md (movdf): New define expand.
6268 2018-02-19 Martin Liska <mliska@suse.cz>
6271 * doc/invoke.texi: Fix typo.
6272 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6274 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6276 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6277 handle rs6000_single_float and rs6000_double_float specially for
6280 2018-02-16 Jeff Law <law@redhat.com>
6282 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6285 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6288 * ipa-inline.c (inline_small_functions): Fix a typo.
6290 2018-02-16 Nathan Sidwell <nathan@acm.org>
6292 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6294 2018-02-16 Carl Love <cel@us.ibm.com>
6296 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6297 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6298 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6299 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6300 expansion to P8V_BUILTIN_VEC_FLOAT2.
6302 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6304 PR rtl-optimization/70023
6305 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6306 src_regno into account.
6308 2018-02-16 Carl Love <cel@us.ibm.com>
6310 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6311 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6312 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6313 * config/rs6000/rs6000.c: Remove case statements for
6314 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6315 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6316 and P9V_BUILTIN_VEC_VINSERT4B.
6317 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6318 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6319 * config/rs6000/vsx.md:
6320 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6323 2018-02-16 Carl Love <cel@us.ibm.com>
6325 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6327 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6329 * config/rs6000/rs6000-c.c: Add the definitions for
6330 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6331 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6332 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6333 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6334 definition for insert4b and define insn *insert3b_internal.
6335 * doc/extend.texi: Add documentation for vec_extract4b.
6337 2018-02-16 Nathan Sidwell <nathan@acm.org>
6339 * doc/extend.texi (Backwards Compatibility): Mention friend
6340 injection. Note for-scope is deprecated.
6341 * doc/invoke.texi (-ffriend-injection): Deprecate.
6343 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6345 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6346 that moved to I2, also allow destinations that are a paradoxical
6347 subreg (instead of a normal reg).
6349 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6352 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6355 2018-02-16 Richard Biener <rguenther@suse.de>
6357 PR tree-optimization/84037
6358 PR tree-optimization/84016
6360 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6361 Adjust vec_construct for the fact we need additional higher latency
6362 128bit inserts for AVX256 and AVX512 vector builds.
6363 (ix86_add_stmt_cost): Scale vector construction cost for
6366 2018-02-16 Richard Biener <rguenther@suse.de>
6368 PR tree-optimization/84417
6369 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6370 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6371 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6373 2018-02-16 Martin Liska <mliska@suse.cz>
6376 * internal-fn.def (ASAN_CHECK): Set proper flags.
6377 (ASAN_MARK): Likewise.
6379 2018-02-16 Julia Koval <julia.koval@intel.com>
6381 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6382 from PTA_CANNONLAKE.
6384 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6387 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6388 Use ++iter rather than iter++ for std::list iterators.
6389 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6390 defer deleting them until all nodes in the forest are processed. Do
6391 free even leaf nodes. Change to_process into auto_vec.
6394 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6395 * vec.h (vec_default_construct): Use memset instead of placement new
6396 if BROKEN_VALUE_INITIALIZATION is defined.
6397 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6398 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6401 PR rtl-optimization/83723
6402 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6403 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6404 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6406 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6408 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6410 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6412 PR rtl-optimization/81443
6413 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6414 from inner REGs to paradoxical SUBREGs.
6416 2018-02-16 Richard Biener <rguenther@suse.de>
6418 PR tree-optimization/84399
6419 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6420 For operands we can analyze at their definition make sure we can
6421 analyze them at each use as well.
6423 2018-02-16 Richard Biener <rguenther@suse.de>
6425 PR tree-optimization/84190
6426 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6427 volatile accesses if the decl isn't volatile.
6429 2018-02-15 Jason Merrill <jason@redhat.com>
6431 PR c++/84314 - ICE with templates and fastcall attribute.
6432 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6433 TYPE_CANONICAL on an existing type.
6435 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6437 PR tree-optimization/84383
6438 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6439 dstoff nor call operand_equal_p if dstbase is NULL.
6441 PR tree-optimization/84334
6442 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6443 also a CONSTANT_CLASS_P, punt.
6445 2018-02-14 Jim Wilson <jimw@sifive.com>
6447 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6448 first SMALL_OPERAND check. New local min_second_step. Move assert
6449 to where locals are set. Add TARGET_RVC support.
6450 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6452 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6454 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6456 2018-02-14 Martin Sebor <msebor@redhat.com>
6458 PR tree-optimization/83698
6459 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6460 arrays constrain the offset range to their bounds.
6461 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6462 (builtin_access::overlap): Avoid setting the size of overlap if it's
6464 (maybe_diag_overlap): Also consider arrays when deciding what values
6465 of offsets to include in diagnostics.
6467 2018-02-14 Martin Sebor <msebor@redhat.com>
6470 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6471 that correspond to the kind of a declaration.
6473 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6476 * config/pa/pa.md: Load address of PIC label using the linkage table
6477 if the label is nonlocal.
6479 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6481 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6482 warning message if user requests -maltivec=be.
6483 * doc/invoke.texi: Document deprecation of -maltivec=be.
6485 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6488 * config/rs6000/rs6000-c.c: Update definitions for
6489 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6490 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6492 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6495 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6496 add _get_ssp intrinsics. Remove argument from
6497 __builtin_ia32_rdssp[d|q].
6498 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6499 * config/i386/i386-builtin.def: Remove argument from
6500 __builtin_ia32_rdssp[d|q].
6501 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6502 ix86_expand_special_args_builtin for _rdssp[d|q].
6503 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6504 Clear register before usage.
6505 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6506 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6508 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6510 PR tree-optimization/84357
6511 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6512 operand 1 of an ARRAY_REF too.
6514 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6517 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6518 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6520 (set_of_reg): New struct.
6521 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6522 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6523 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6525 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6526 Split into bitclr, bitset, bitinvert patterns if appropriate.
6527 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6528 use rx_fuse_in_memory_bitop.
6529 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6530 to named insn, correct maximum insn length.
6532 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6535 * machmode.def: Define a complex mode for PARTIAL_INT.
6536 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6538 * doc/rtl.texi: Document CSPImode.
6539 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6541 (msp430_hard_regno_nregs_with_padding): Likewise.
6543 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6546 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6548 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6550 PR rtl-optimization/84169
6551 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6552 we generated a parallel as new i3 and we split that to new i2 and i3
6553 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6554 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6555 those to i2, not i1. Partially rewrite this scan code.
6557 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6560 * stor-layout.c (place_field): For variable length fields, adjust
6561 offset_align afterwards not just based on the field's alignment,
6562 but also on the size.
6565 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6566 of exps and logs in the use_exp2 case.
6568 2018-02-13 Jeff Law <law@redhat.com>
6570 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6573 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6576 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6580 * common.opt (gas-loc-support, gas-locview-support): New.
6581 (ginline-points, ginternal-reset-location-views): New.
6582 * doc/invoke.texi: Document them. Use @itemx where intended.
6583 (gvariable-location-views): Adjust.
6584 * target.def (reset_location_view): New.
6585 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6586 (TARGET_RESET_LOCATION_VIEW): New.
6587 * doc/tm.texi: Rebuilt.
6588 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6589 (dwarf2out_default_as_locview_support): New.
6590 (output_asm_line_debug_info): Use option variables.
6591 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6592 (output_loc_list): Likewise.
6593 (add_high_low_attributes): Check option variables.
6594 Don't output entry view attribute in strict mode.
6595 (gen_inlined_subroutine_die): Check option variables.
6596 (dwarf2out_inline_entry): Likewise.
6597 (init_sections_and_labels): Likewise.
6598 (dwarf2out_early_finish): Likewise.
6599 (maybe_reset_location_view): New, from...
6600 (dwarf2out_var_location): ... here. Call it.
6601 * debug.h (dwarf2out_default_as_loc_support): Declare.
6602 (dwarf2out_default_as_locview_support): Declare.
6603 * hooks.c (hook_int_rtx_insn_0): New.
6604 * hooks.h (hook_int_rtx_insn_0): Declare.
6605 * toplev.c (process_options): Take -gas-loc-support and
6606 -gas-locview-support from dwarf2out. Enable
6607 -gvariable-location-views by default only with locview
6608 assembler support. Enable -ginternal-reset-location-views by
6609 default only if the target defines the corresponding hook.
6610 Enable -ginline-points by default if location views are
6611 enabled; force it disabled if statement frontiers are
6613 * tree-inline.c (expand_call_inline): Check option variables.
6614 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6616 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6618 PR tree-optimization/84321
6619 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6620 handling. Also check whether the anti-range contains any values
6621 that satisfy the mask; switch to a VR_RANGE if not.
6623 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6626 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6628 2018-02-13 Martin Jambor <mjambor@suse.cz>
6631 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6632 of call statements, also set location of a load to a temporary.
6634 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6636 * config/rl78/rl78.c (add_vector_labels): New function.
6637 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6638 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6639 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6640 which checks that no arguments are passed.
6641 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6642 * doc/extend.texi: Documentation for the new attribute.
6644 2018-02-13 Andreas Schwab <schwab@suse.de>
6646 * config/riscv/linux.h (CPP_SPEC): Define.
6648 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6651 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6652 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6653 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6654 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6655 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6656 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6657 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6658 temporarily for AES and PCLMUL builtins.
6660 PR tree-optimization/84339
6661 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6662 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6666 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6667 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6668 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6670 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6672 * omp-simd-clone.h: New file.
6673 * omp-simd-clone.c: Include omp-simd-clone.h.
6674 (expand_simd_clones): No longer static.
6675 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6676 cgraph.h and omp-simd-clone.h.
6677 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6678 (vect_recog_widen_shift_pattern): Formatting fix.
6679 (vect_pattern_recog_1): Don't check optab for calls.
6682 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6683 operands[2] into a REG before using gen_lowpart on it.
6685 2018-02-12 Jeff Law <law@redhat.com>
6688 * config/sh/sh.c (find_barrier): Consider a sibling call
6691 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6692 successfully back substituting a reg.
6694 2018-02-12 Richard Biener <rguenther@suse.de>
6696 PR tree-optimization/84037
6697 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6698 parameter, move visited init to caller.
6699 (vect_slp_analyze_operations): Separate cost from validity
6700 check, initialize visited once for all instances.
6701 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6703 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6705 (vect_model_promotion_demotion_cost): Likewise.
6706 (vectorizable_bswap): Guard cost modeling with !slp_node
6707 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6709 (vectorizable_call): Likewise.
6710 (vectorizable_conversion): Likewise.
6711 (vectorizable_assignment): Likewise.
6712 (vectorizable_shift): Likewise.
6713 (vectorizable_operation): Likewise.
6714 (vectorizable_store): Likewise.
6715 (vectorizable_load): Likewise.
6716 (vectorizable_condition): Likewise.
6717 (vectorizable_comparison): Likewise.
6719 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6722 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6723 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6724 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6726 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6729 * params.def (inline-min-speedup): Increase from 8 to 15.
6730 (max-inline-insns-auto): Decrease from 40 to 30.
6731 * ipa-split.c (consider_split): Add some buffer for function to
6732 be considered inlining candidate.
6733 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6736 2018-02-12 Richard Biener <rguenther@suse.de>
6738 PR tree-optimization/84037
6739 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6740 matched stmts if we cannot swap the non-matched ones.
6742 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6744 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6745 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6746 _mm_maskz_scalef_round_ss): New intrinsics.
6747 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6748 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6749 __builtin_ia32_scalefss_round): Remove.
6750 (__builtin_ia32_scalefsd_mask_round,
6751 __builtin_ia32_scalefss_mask_round): New intrinsics.
6752 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6753 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6754 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6755 "<round_constraint>")): Changed to ...
6756 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6757 "<round_scalar_constraint>")): ... this.
6758 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6759 %0, %1, %2<round_op3>}"): Changed to ...
6760 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6761 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6762 %2<round_scalar_mask_op3>}"): ... this.
6763 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6765 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6767 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6768 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6769 (_mm_maskz_sqrt_round_ss): New intrinsics.
6770 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6771 (__builtin_ia32_sqrtsd_mask_round)
6772 (__builtin_ia32_sqrtss_mask_round): New builtins.
6773 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6774 (__builtin_ia32_sqrtss_round): Remove.
6775 (__builtin_ia32_sqrtsd_mask_round)
6776 (__builtin_ia32_sqrtss_mask_round): New builtins.
6777 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6778 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6779 ((match_operand:VF_128 1 "vector_operand"
6780 "xBm,<round_constraint>")): Changed to ...
6781 ((match_operand:VF_128 1 "vector_operand"
6782 "xBm,<round_scalar_constraint>")): ... this.
6783 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6784 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6785 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6786 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6787 %<iptr>1<round_scalar_mask_op3>}): ... this.
6788 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6789 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6791 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6794 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6795 Cast vec_cmpeq result to correct type.
6796 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6797 Cast vec_cmpgt result to correct type.
6799 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6801 * final.c (final_scan_insn_1): Renamed from...
6802 (final_scan_insn): ... this. New wrapper, to recover
6803 seen from the outermost call in recursive ones.
6804 * config/sparc/sparc.c (output_return): Drop seen from call.
6805 (output_sibcall): Likewise.
6806 * config/visium/visium.c (output_branch): Likewise.
6808 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6810 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6813 2018-02-10 Alan Modra <amodra@gmail.com>
6816 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6817 Specify LR as an input.
6819 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6822 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6823 remove_member_access_dummy_vars): New functions.
6824 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6825 lower_omp_1, execute_lower_omp): Use them.
6827 PR rtl-optimization/84308
6828 * shrink-wrap.c (spread_components): Release todo vector.
6830 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6832 PR rtl-optimization/57193
6833 * ira-color.c (struct allocno_color_data): Add member
6834 conflict_allocno_hard_prefs.
6835 (update_conflict_allocno_hard_prefs): New.
6836 (bucket_allocno_compare_func): Add a preference based on
6837 conflict_allocno_hard_prefs.
6838 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6839 (color_allocnos): Remove a dead code. Initiate
6840 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6842 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6845 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6846 constraint from =wa to wa. Avoid a subreg on the output operand,
6847 instead use a pseudo and subreg it in a move.
6848 (p9_xxbrd_<mode>): Changed to ...
6849 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6850 (p9_xxbrd_v2df): New expander.
6851 (p9_xxbrw_<mode>): Changed to ...
6852 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6853 (p9_xxbrw_v4sf): New expander.
6855 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6857 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6859 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6862 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6863 multiply in 32-bit mode.
6864 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6865 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6868 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6870 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6871 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6872 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6873 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6875 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6878 * dwarf2out.c (is_trivial_indirect_ref): New function.
6879 (dwarf2out_late_global_decl): Do not generate a location
6880 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6881 and that are not defined in the current unit.
6883 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6885 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6886 instead of a libcall for UNORDERED.
6888 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6891 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6892 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6894 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6897 * config/s390/s390.c (s390_set_current_function): Invoke
6898 s390_indirect_branch_settings also if fndecl didn't change.
6900 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6902 * config/rs6000/rs6000.md (blockage): Set length to zero.
6904 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6906 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6908 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6911 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6912 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6916 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6917 PARALLEL incoming that failed vt_get_decl_and_offset check.
6920 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6921 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6923 (get_variable_section): For decls in named .bss* sections pass true as
6924 second argument to bss_initializer_p.
6926 2018-02-09 Marek Polacek <polacek@redhat.com>
6927 Jakub Jelinek <jakub@redhat.com>
6930 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6931 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6932 Sync some changes from cxx_fold_indirect_ref.
6934 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6936 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6938 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6939 (BLOCK_INLINE_ENTRY_LABEL): New.
6940 (dwarf2out_var_location): Disregard inline entry markers.
6941 (inline_entry_data): New struct.
6942 (inline_entry_data_hasher): New hashtable type.
6943 (inline_entry_data_hasher::hash): New.
6944 (inline_entry_data_hasher::equal): New.
6945 (inline_entry_data_table): New variable.
6946 (add_high_low_attributes): Add DW_AT_entry_pc and
6947 DW_AT_GNU_entry_view attributes if a pending entry is found
6948 in inline_entry_data_table. Add old entry_pc attribute only
6949 if debug nonbinding markers are disabled.
6950 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6951 markers are enabled.
6952 (block_within_block_p, dwarf2out_inline_entry): New.
6953 (dwarf2out_finish): Check that no entries remained in
6954 inline_entry_data_table.
6955 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6956 (final_scan_insn, notice_source_line): Likewise.
6957 (rest_of_clean_state): Skip inline entry markers.
6958 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6960 * gimple.c (gimple_build_debug_inline_entry): New.
6961 * gimple.h (enum gimple_debug_subcode): Add
6962 GIMPLE_DEBUG_INLINE_ENTRY.
6963 (gimple_build_debug_inline_entry): Declare.
6964 (gimple_debug_inline_entry_p): New.
6965 (gimple_debug_nonbind_marker_p): Adjust.
6966 * insn-notes.def (INLINE_ENTRY): New.
6967 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6968 inline entry marker notes.
6969 (print_insn): Likewise.
6970 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6971 (INSN_DEBUG_MARKER_KIND): Likewise.
6972 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6973 * tree-inline.c (expand_call_inline): Build and insert
6974 debug_inline_entry stmt.
6975 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6976 inline entry blocks early, if nonbind markers are enabled.
6977 (dump_scope_block): Dump fragment info.
6978 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6979 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6980 (gimple_build_debug_inline_entry): New.
6981 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6982 Enable/disable inline entry points too.
6983 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6984 (DEBUG_INSN): Describe inline entry markers.
6986 * common.opt (gvariable-location-views): New.
6987 (gvariable-location-views=incompat5): New.
6988 * config.in: Rebuilt.
6989 * configure: Rebuilt.
6990 * configure.ac: Test assembler for view support.
6991 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6992 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6993 * dwarf2out.c (var_loc_view): New typedef.
6994 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6995 (dwarf2out_locviews_in_attribute): New.
6996 (dwarf2out_locviews_in_loclist): New.
6997 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6998 (enum dw_line_info_opcode): Add LI_adv_address.
6999 (struct dw_line_info_table): Add view.
7000 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7001 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7002 (zero_view_p): New variable.
7003 (ZERO_VIEW_P): New macro.
7004 (output_asm_line_debug_info): New.
7005 (struct var_loc_node): Add view.
7006 (add_AT_view_list, AT_loc_list): New.
7007 (add_var_loc_to_decl): Add view param. Test it against last.
7008 (new_loc_list): Add view params. Record them.
7009 (AT_loc_list_ptr): Handle loc and view lists.
7010 (view_list_to_loc_list_val_node): New.
7011 (print_dw_val): Handle dw_val_class_view_list.
7012 (size_of_die): Likewise.
7013 (value_format): Likewise.
7014 (loc_list_has_views): New.
7015 (gen_llsym): Set vl_symbol too.
7016 (maybe_gen_llsym, skip_loc_list_entry): New.
7017 (dwarf2out_maybe_output_loclist_view_pair): New.
7018 (output_loc_list): Output view list or entries too.
7019 (output_view_list_offset): New.
7020 (output_die): Handle dw_val_class_view_list.
7021 (output_dwarf_version): New.
7022 (output_compilation_unit_header): Use it.
7023 (output_skeleton_debug_sections): Likewise.
7024 (output_rnglists, output_line_info): Likewise.
7025 (output_pubnames, output_aranges): Update version comments.
7026 (output_one_line_info_table): Output view numbers in asm comments.
7027 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7028 Call maybe_gen_llsym.
7029 (loc_list_from_tree_1): Adjust.
7030 (add_AT_location_description): Create view list attribute if
7031 needed, check it's absent otherwise.
7032 (convert_cfa_to_fb_loc_list): Adjust.
7033 (maybe_emit_file): Call output_asm_line_debug_info for test.
7034 (dwarf2out_var_location): Reset views as needed. Precompute
7035 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7036 attribute. Set view.
7037 (new_line_info_table): Reset next view.
7038 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7039 (dwarf2out_source_line): Likewise. Output view resets and labels to
7040 the assembler, or select appropriate line info opcodes.
7041 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7042 (optimize_string_length): Catch it. Adjust.
7043 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7044 dw_val_class_view_list, and remove it if no longer needed.
7045 (hash_loc_list): Hash view numbers.
7046 (loc_list_hasher::equal): Compare them.
7047 (optimize_location_lists): Check whether a view list symbol is
7048 needed, and whether the locview attribute is present, and
7049 whether they match. Remove the locview attribute if no longer
7051 (index_location_lists): Call skip_loc_list_entry for test.
7052 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7053 Use output_dwarf_version.
7054 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7055 (struct dw_val_node): Add val_view_list.
7056 * final.c (SEEN_NEXT_VIEW): New.
7057 (set_next_view_needed): New.
7058 (clear_next_view_needed): New.
7059 (maybe_output_next_view): New.
7060 (final_start_function): Rename to...
7061 (final_start_function_1): ... this. Take pointer to FIRST,
7062 add SEEN parameter. Emit param bindings in the initial view.
7063 (final_start_function): Reintroduce SEEN-less interface.
7064 (final): Rename to...
7065 (final_1): ... this. Take SEEN parameter. Output final pending
7066 next view at the end.
7067 (final): Reintroduce seen-less interface.
7068 (final_scan_insn): Output pending next view before switching
7069 sections or ending a block. Mark the next view as needed when
7070 outputting variable locations. Notify debug backend of section
7071 changes, and of location view changes.
7072 (rest_of_handle_final): Adjust.
7073 * toplev.c (process_options): Autodetect value for debug variable
7074 location views option. Warn on incompat5 without -gdwarf-5.
7075 * doc/invoke.texi (gvariable-location-views): New.
7076 (gvariable-location-views=incompat5): New.
7077 (gno-variable-location-views): New.
7079 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7081 PR tree-optimization/84136
7082 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7083 that the result of find_edge is non-NULL.
7085 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7088 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7089 storing integer register in SImode. Fix cost of 256 and 512
7090 byte aligned SSE register store.
7092 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7094 * config/i386/i386.c (ix86_multiplication_cost): Fix
7095 multiplication cost for TARGET_AVX512DQ.
7097 2018-02-08 Marek Polacek <polacek@redhat.com>
7099 PR tree-optimization/84238
7100 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7103 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7105 PR tree-optimization/84265
7106 * tree-vect-stmts.c (vectorizable_store): Don't treat
7107 VMAT_CONTIGUOUS accesses as grouped.
7108 (vectorizable_load): Likewise.
7110 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7112 PR tree-optimization/81635
7113 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7114 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7115 (test_round_for_mask): New functions.
7116 (wide_int_cc_tests): Call test_round_for_mask.
7117 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7118 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7119 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7120 range returned by get_range_info.
7122 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7125 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7126 * symtab.c: Include builtins.h
7127 (symtab_node::output_to_lto_symbol_table_p): Move here
7128 from lto-streamer-out.c:output_symbol_p.
7129 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7130 (output_symbol_p): Move all logic to symtab.c
7131 (produce_symtab): Update.
7133 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7135 * config/s390/s390-opts.h (enum indirect_branch): Define.
7136 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7137 (s390_indirect_branch_via_thunk)
7138 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7139 (enum s390_indirect_branch_type): Define.
7140 * config/s390/s390.c (struct s390_frame_layout, struct
7141 machine_function): Remove.
7142 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7143 (indirect_branch_table_label_no, indirect_branch_table_name):
7145 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7146 (enum s390_indirect_branch_option): Define.
7147 (s390_return_addr_from_memory): New function.
7148 (s390_handle_string_attribute): New function.
7149 (s390_attribute_table): Add new attribute handler.
7150 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7151 (s390_indirect_branch_via_thunk): New function.
7152 (s390_indirect_branch_via_inline_thunk): New function.
7153 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7154 sibling call optimization for non z10 compiles.
7155 (s390_emit_call): Force indirect branch target to be a single
7156 register. Add r1 clobber for non-z10 compiles.
7157 (s390_emit_epilogue): Emit return jump via return_use expander.
7158 (s390_reorg): Handle JUMP_INSNs as execute targets.
7159 (s390_option_override_internal): Perform validity checks for the
7160 new command line options.
7161 (s390_indirect_branch_attrvalue): New function.
7162 (s390_indirect_branch_settings): New function.
7163 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7164 (s390_output_indirect_thunk_function): New function.
7165 (s390_code_end): Implement target hook.
7166 (s390_case_values_threshold): Implement target hook.
7167 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7169 * config/s390/s390.h (struct s390_frame_layout)
7170 (struct machine_function): Move here from s390.c.
7171 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7172 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7173 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7174 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7175 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7176 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7177 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7178 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7179 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7180 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7181 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7182 (mnemonic attribute): Add values which aren't recognized
7184 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7185 pattern for branch conversion. Fix mnemonic attribute.
7186 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7187 indirect branch via thunk if requested.
7188 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7189 ("*indirect_jump"): Disable for branch conversion using out of
7191 ("indirect_jump_via_thunk<mode>_z10")
7192 ("indirect_jump_via_thunk<mode>")
7193 ("indirect_jump_via_inlinethunk<mode>_z10")
7194 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7195 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7196 ("casesi_jump_via_inlinethunk<mode>_z10")
7197 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7198 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7199 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7200 ("*indirect2_jump"): Disable for branch conversion.
7201 ("casesi_jump"): Turn into expander and expand patterns for branch
7203 ("return_use"): New expander.
7204 ("*return"): Emit return via thunk and rename it to ...
7205 ("*return<mode>"): ... this one.
7206 * config/s390/s390.opt: Add new options and and enum for the
7209 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7211 * lra-constraints.c (match_reload): Unconditionally use
7212 gen_lowpart_SUBREG, rather than selecting between that
7213 and equivalent gen_rtx_SUBREG code.
7215 2018-02-08 Richard Biener <rguenther@suse.de>
7217 PR tree-optimization/84233
7218 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7219 changed flag instead of boguously re-using phi_inserted.
7221 2018-02-08 Martin Jambor <mjambor@suse.cz>
7223 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7224 static local variables.
7226 2018-02-08 Richard Biener <rguenther@suse.de>
7228 PR tree-optimization/84278
7229 * tree-vect-stmts.c (vectorizable_store): When looking for
7230 smaller vector types to perform grouped strided loads/stores
7231 make sure the mode is supported by the target.
7232 (vectorizable_load): Likewise.
7234 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7236 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7237 Increase LDP/STP opportunities by adding adjacent callee-saves.
7239 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7241 PR rtl-optimization/84068
7242 PR rtl-optimization/83459
7243 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7245 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7247 PR tree-optimization/84224
7248 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7249 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7252 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7255 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7256 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7257 count, remove one USE.
7259 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7261 * doc/install.texi (Configuration): Document the
7262 --with-long-double-format={ibm,ieee} PowerPC configuration
7266 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7267 Convert from define_expand to be define_insn_and_split. Rework
7268 float/double/_Float128 conversions to QI/HI/SImode to work with
7269 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7270 conversions to QI/HImode types did a store and then a load to
7271 truncate the value. For conversions to VSX registers, don't split
7272 the insn, instead emit the code directly. Use the code iterator
7273 any_fix to combine signed and unsigned conversions.
7274 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7275 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7276 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7277 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7278 (fix_<mode>di2_hw): Likewise.
7279 (fixuns_<mode>di2_hw): Likewise.
7280 (fix_<mode>si2_hw): Likewise.
7281 (fixuns_<mode>si2_hw): Likewise.
7282 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7283 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7284 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7285 fix<uns>_trunc<SFDF:mode>si2_p8.
7286 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7288 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7289 (fix<uns>_<mode>_mem): Likewise.
7290 (fctiw<u>z_<mode>_mem): Likewise.
7291 (fix<uns>_<mode>_mem): Likewise.
7292 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7293 the register allocator from doing a direct move to the GPRs to do
7294 a store, and instead use the ISA 3.0 store byte/half-word from
7295 vector register instruction. For IEEE 128-bit floating point,
7296 also optimize stores of 32-bit ints.
7297 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7299 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7301 * genextract.c (push_pathstr_operand): New function to support
7303 (walk_rtx): Call push_pathstr_operand.
7304 (print_path): Support [a-zA-Z].
7306 2018-02-07 Richard Biener <rguenther@suse.de>
7308 PR tree-optimization/84037
7309 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7310 (cse_and_gimplify_to_preheader): Declare.
7311 (vect_get_place_in_interleaving_chain): Likewise.
7312 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7314 (_loop_vec_info::~_loop_vec_info): Delete it.
7315 (cse_and_gimplify_to_preheader): New function.
7316 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7317 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7318 (vectorizable_load): Likewise. For grouped stores always base
7319 the IV on the first element.
7320 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7321 condition before gimplifying.
7323 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7325 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7326 *DIV_EXPR and *MOD_EXPR.
7328 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7331 * config/i386/i386.c (ix86_option_override_internal): Mask out
7332 the CF_SET bit when checking -fcf-protection.
7334 2018-02-07 Tom de Vries <tom@codesourcery.com>
7337 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7340 2018-02-07 Richard Biener <rguenther@suse.de>
7342 PR tree-optimization/84204
7343 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7346 PR tree-optimization/84205
7347 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7348 special-case isl_ast_op_zdiv_r.
7350 PR tree-optimization/84223
7351 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7352 Only add conditions from within the region.
7353 (gather_bbs::after_dom_children): Adjust.
7355 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7358 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7359 * config/avr/avr.md: Only post-reload split REG-REG moves if
7360 either register is GENERAL_REG_P.
7362 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7364 PR tree-optimization/84235
7365 * tree-ssa-scopedtables.c
7366 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7367 if the subtraction is performed in floating point type where NaNs are
7368 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7369 build 1. Formatting fix.
7371 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7374 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7375 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7376 and skip it regardless of bb boundaries. Use CALL_P macro,
7377 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7378 unnecessarily, formatting fix.
7380 2018-02-06 Michael Collison <michael.collison@arm.com>
7382 * config/arm/thumb2.md:
7383 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7384 (*thumb_mov_notscc): Ditto.
7386 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7389 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7390 unsigned_fix, not "s".
7392 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7394 * configure.ac (gcc_fn_eh_frame_ro): New function.
7395 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7396 correct .eh_frame permissions.
7397 * configure: Regenerate.
7399 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7401 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7404 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7406 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7407 Display warning message for -mno-speculate-indirect-jumps.
7409 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7411 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7413 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7415 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7417 PR tree-optimization/84225
7418 * tree-eh.c (find_trapping_overflow): Only call
7419 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7421 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7424 * config/i386/i386.c: Reimplement the check of possible options
7425 -mibt/-mshstk conbination. Change error messages.
7426 * doc/invoke.texi: Fix a typo: remove extra '='.
7428 2018-02-06 Marek Polacek <polacek@redhat.com>
7430 PR tree-optimization/84228
7431 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7433 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7436 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7437 emitted arch directives.
7438 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7439 __ARM_FEATURE_COPROC before changing architectures.
7441 2018-02-06 Richard Biener <rguenther@suse.de>
7443 * config/i386/i386.c (print_reg): Fix typo.
7444 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7446 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7448 * configure: Regenerate.
7450 2018-02-05 Martin Sebor <msebor@redhat.com>
7452 PR tree-optimization/83369
7453 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7456 2018-02-05 Martin Liska <mliska@suse.cz>
7458 * doc/invoke.texi: Cherry-pick upstream r323995.
7460 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7462 * ira.c (ira_init_register_move_cost): Adjust comment.
7464 2018-02-05 Martin Liska <mliska@suse.cz>
7466 PR gcov-profile/84137
7467 * doc/gcov.texi: Fix typo in documentation.
7469 2018-02-05 Martin Liska <mliska@suse.cz>
7471 PR gcov-profile/83879
7472 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7473 using dlopen functionality.
7475 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7477 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7478 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7479 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7480 _mm_maskz_range_round_ss): New intrinsics.
7481 (__builtin_ia32_rangesd128_round)
7482 (__builtin_ia32_rangess128_round): Remove.
7483 (__builtin_ia32_rangesd128_mask_round,
7484 __builtin_ia32_rangess128_mask_round): New builtins.
7485 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7486 __builtin_ia32_rangess128_round): Remove.
7487 (__builtin_ia32_rangesd128_mask_round,
7488 __builtin_ia32_rangess128_mask_round): New builtins.
7489 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7490 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7491 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7492 "<round_saeonly_constraint>")): Changed to ...
7493 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7494 "<round_saeonly_scalar_constraint>")): ... this.
7495 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7496 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7497 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7498 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7499 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7501 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7503 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7505 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7506 Remove all values except native, 8540 and 8548.
7508 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7510 * config/i386/i386.c (ix86_output_function_return): Pass
7511 INVALID_REGNUM, instead of -1, as invalid register number to
7512 indirect_thunk_name and output_indirect_thunk.
7514 2018-02-02 Julia Koval <julia.koval@intel.com>
7516 * config.gcc: Add -march=icelake.
7517 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7518 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7519 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7520 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7521 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7522 (processor_target_table): Add icelake.
7523 (ix86_option_override_internal): Handle new PTAs.
7524 (get_builtin_code_for_version): Handle icelake.
7525 (M_INTEL_COREI7_ICELAKE): New.
7526 (fold_builtin_cpu): Handle icelake.
7527 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7528 * doc/invoke.texi: Add -march=icelake.
7530 2018-02-02 Julia Koval <julia.koval@intel.com>
7532 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7533 to wide_int_bitmask.
7534 * wide-int-bitmask.h: New.
7536 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7539 * config/i386/i386.md: Replace Pmode with word_mode in
7540 builtin_setjmp_setup and builtin_longjmp to support x32.
7542 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7546 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7548 (rs6000_supported_cpu_names): New static variable.
7549 (linux_cpu_translation_table): Likewise.
7550 (elf_platform) <cpu>: Define new static variable and use it.
7551 Translate kernel AT_PLATFORM name to canonical name if needed.
7552 Error if platform name is unknown.
7554 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7557 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7559 2018-02-01 Jeff Law <law@redhat.com>
7562 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7563 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7564 the scratch if RELEASE_VIA_POP is false.
7565 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7566 If we have to save a temporary register, decrement SIZE appropriately.
7567 Pass new arguments to release_scratch_register_on_entry.
7568 (ix86_adjust_stack_and_probe): Likewise.
7569 (ix86_emit_probe_stack_range): Pass new arguments to
7570 release_scratch_register_on_entry.
7572 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7574 PR rtl-optimization/84157
7575 * combine.c (change_zero_ext): Use REG_P predicate in
7576 front of HARD_REGISTER_P predicate.
7578 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7580 * config/avr/avr.c (avr_option_override): Move disabling of
7581 -fdelete-null-pointer-checks to...
7582 * common/config/avr/avr-common.c (avr_option_optimization_table):
7585 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7587 PR tree-optimization/81635
7588 * tree-data-ref.c (split_constant_offset_1): For types that
7589 wrap on overflow, try to use range info to prove that wrapping
7592 2018-02-01 Renlin Li <renlin.li@arm.com>
7595 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7597 (aarch64_register_move_cost): Likewise.
7598 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7600 (REG_CLASS_NAMES): Likewise.
7601 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7602 TAILCALL_ADDR_REGS. Remove IP registers.
7603 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7605 2018-02-01 Richard Biener <rguenther@suse.de>
7607 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7608 for specifying RPO order and allow NULL for that.
7609 * domwalk.c (dom_walker::dom_walker): Likewise.
7610 (dom_walker::walk): Handle NULL RPO order.
7611 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7613 (rewrite_update_dom_walker): Likewise.
7614 (mark_def_dom_walker): Likewise.
7616 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7618 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7619 (aarch64_maybe_expand_sve_subreg_move): Declare.
7620 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7621 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7623 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7624 that are semantically a reverse operation.
7625 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7626 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7627 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7629 (aarch64_can_change_mode_class): For big-endian, forbid changes
7630 between two SVE modes if they have different element sizes.
7632 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7634 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7635 the TImode handling for big-endian targets.
7637 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7639 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7640 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7642 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7643 Remove BSWAP handing for big-endian targets and use the form of
7644 LD1RQ appropariate for the mode.
7646 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7648 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7649 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7652 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7655 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7656 check for operands that need to go through aarch64_sve_reload_be.
7658 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7660 PR tree-optimization/81661
7661 PR tree-optimization/84117
7662 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7663 * tree-eh.c: Include gimplify.h.
7664 (find_trapping_overflow, replace_trapping_overflow,
7665 rewrite_to_non_trapping_overflow): New functions.
7666 * tree-vect-loop.c: Include tree-eh.h.
7667 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7668 * tree-data-ref.c: Include tree-eh.h.
7669 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7671 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7673 PR rtl-optimization/84123
7674 * combine.c (change_zero_ext): Check if hard register satisfies
7675 can_change_dest_mode before calling gen_lowpart_SUBREG.
7677 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7680 * ira.c (ira_init_register_move_cost): Remove assert.
7682 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7684 PR rtl-optimization/84071
7685 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7686 * doc/tm.texi: Regenerate.
7688 2018-01-31 Richard Biener <rguenther@suse.de>
7690 PR tree-optimization/84132
7691 * tree-data-ref.c (analyze_miv_subscript): Properly
7692 check whether evolution_function_is_affine_multivariate_p
7693 before calling gcd_of_steps_may_divide_p.
7695 2018-01-31 Julia Koval <julia.koval@intel.com>
7698 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7699 * config/i386/i386.md (rdpid_rex64) New.
7700 (rdpid): Make 32bit only.
7702 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7705 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7706 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7708 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7711 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7713 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7715 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7717 PR rtl-optimization/84071
7718 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7719 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7721 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7723 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7724 (arc_attribute_table): Add 'aux' attribute.
7725 (arc_in_small_data_p): Consider aux like variables.
7726 (arc_is_aux_reg_p): New function.
7727 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7728 (arc_get_aux_arg): New function.
7729 (prepare_move_operands): Handle aux-register access.
7730 (arc_handle_aux_attribute): New function.
7731 * doc/extend.texi (ARC Variable attributes): Add subsection.
7733 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7735 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7736 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7737 (arc_attribute_table): Add 'uncached' attribute.
7738 (arc_print_operand): Print '.di' flag for uncached memory
7740 (arc_in_small_data_p): Do not consider for small data the uncached
7742 (arc_is_uncached_mem_p): New function.
7743 * config/arc/predicates.md (compact_store_memory_operand): Check
7744 for uncached memory accesses.
7745 (nonvol_nonimm_operand): Likewise.
7746 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7748 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7751 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7752 falign-loops=): Add Optimization flag.
7754 2018-01-30 Jeff Law <law@redhat.com>
7757 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7758 INT_REGISTERS_SAVED. Check it prior to calling
7759 get_scratch_register_on_entry.
7760 (ix86_adjust_stack_and_probe): Similarly.
7761 (ix86_emit_probe_stack_range): Similarly.
7762 (ix86_expand_prologue): Corresponding changes.
7764 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7767 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7768 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7770 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7773 * lra-constraints.c (curr_insn_transform): Process AND in the
7776 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7778 PR rtl-optimization/83986
7779 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7780 dependence against last_pending_memory_flush in addition to
7783 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7785 PR tree-optimization/81611
7786 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7789 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7792 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7795 2018-01-30 Richard Biener <rguenther@suse.de>
7796 Jakub Jelinek <jakub@redhat.com>
7798 PR tree-optimization/84111
7799 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7800 inner loops added during recursion, as they don't have up-to-date
7803 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7806 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7807 (can_inline_edge_by_limits_p): ... here.
7808 (can_early_inline_edge_p, check_callers,
7809 update_caller_keys, update_callee_keys, recursive_inlining,
7810 add_new_edges_to_heap, speculation_useful_p,
7811 inline_small_functions,
7812 inline_small_functions, flatten_function,
7813 inline_to_all_callers_1): Update.
7815 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7817 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7820 2018-01-30 Richard Biener <rguenther@suse.de>
7822 PR tree-optimization/83008
7823 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7824 invariant and constant vector uses in stmts when they need
7827 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7830 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7831 * configure: Regenerate.
7833 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7835 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7837 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7838 Use gen_rtx_REG rather than gen_lowpart.
7840 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7842 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7843 rather than 0 when creating partial subregs.
7845 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7847 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7850 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7853 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7854 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7855 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7856 flags. This restores the settings used before the 2017-07-24.
7857 Turning off pre increment/decrement/modify allows IVOPTS to
7858 optimize DF/SF loops where the index is an int.
7860 2018-01-29 Richard Biener <rguenther@suse.de>
7861 Kelvin Nilsen <kelvin@gcc.gnu.org>
7864 * tree-vect-stmts.c (vectorizable_call): Don't call
7865 targetm.vectorize_builtin_md_vectorized_function if callee is
7868 2018-01-22 Carl Love <cel@us.ibm.com>
7870 * doc/extend.tex: Fix typo in second arg in
7871 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7873 2018-01-29 Richard Biener <rguenther@suse.de>
7875 PR tree-optimization/84086
7876 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7877 (flush_ssaname_freelist): When SSA names were released reset
7878 the SCEV hash table.
7880 2018-01-29 Richard Biener <rguenther@suse.de>
7882 PR tree-optimization/84057
7883 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7884 removed paths when removing edges.
7886 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7888 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7889 -mfunction-return=@var{choice}.
7891 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7894 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7896 (layout::print_source_line): Likewise.
7897 (test_get_line_width_without_trailing_whitespace): Add test cases.
7899 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7902 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7905 2018-01-26 Jim Wilson <jimw@sifive.com>
7907 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7909 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7912 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7914 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7915 and CMP + SUB-immediate -> SUBS.
7917 2018-01-26 Martin Sebor <msebor@redhat.com>
7919 PR tree-optimization/83896
7920 * tree-ssa-strlen.c (get_string_len): Rename...
7921 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7922 Avoid assuming length is constant.
7923 (handle_char_store): Use HOST_WIDE_INT for string length.
7925 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7928 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7929 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7931 2018-01-26 Richard Biener <rguenther@suse.de>
7933 PR rtl-optimization/84003
7934 * dse.c (record_store): Only record redundant stores when
7935 the earlier store aliases at least all accesses the later one does.
7937 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7939 PR rtl-optimization/83985
7940 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7941 REG_CFA_RESTORE insns.
7942 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7943 REG_CFA_RESTORE insns here.
7946 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7947 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7949 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7951 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7952 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7953 (arc_init): Likewise.
7954 (arc_override_options): Likewise.
7955 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7957 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7959 * config/arc/arc.h (TARGET_DBNZ): Define.
7960 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7961 properly set the tune attribute.
7962 (dbnz): Use TARGET_DBNZ guard.
7963 * config/arc/arc.opt (mtune): Add core3 option.
7965 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7967 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7968 recognize new pic like addresses.
7969 (arc_delegitimize_address): Clean up.
7971 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7973 * config/arc/arc-arches.def: Option mrf16 valid for all
7975 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7976 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7977 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7978 * config/arc/arc-tables.opt: Regenerate.
7979 * config/arc/arc.c (arc_conditional_register_usage): Handle
7980 reduced register file case.
7981 (arc_file_start): Set must have build attributes.
7982 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7984 * config/arc/arc.opt (mrf16): Add new option.
7985 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7986 * config/arc/genmultilib.awk: Handle new mrf16 option.
7987 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7988 * config/arc/t-multilib: Regenerate.
7989 * doc/invoke.texi (ARC Options): Document mrf16 option.
7991 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7993 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7994 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7995 (arc_attribute_table): Add 'secure_call' attribute.
7996 (arc_print_operand): Print secure call operand.
7997 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7999 (arc_is_secure_call_p): New function. * config/arc/arc.md
8000 (call_i): Add support for sjli instruction.
8001 (call_value_i): Likewise.
8002 * config/arc/constraints.md (Csc): New constraint.
8004 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8005 John Eric Martin <John.Martin@emmicro-us.com>
8007 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8008 * config/arc/arc.c (_arc_jli_section): New struct.
8009 (arc_jli_section): New type.
8010 (rc_jli_sections): New static variable.
8011 (arc_handle_jli_attribute): New function.
8012 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8013 (arc_file_end): New function.
8014 (TARGET_ASM_FILE_END): Define.
8015 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8016 (arc_add_jli_section): New function.
8017 (jli_call_scan): Likewise.
8018 (arc_reorg): Call jli_call_scan.
8019 (arc_output_addsi): Remove 'S' from printing asm operand.
8020 (arc_is_jli_call_p): New function.
8021 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8023 (movhi_insn): Likewise.
8024 (movsi_insn): Likewise.
8025 (movsi_set_cc_insn): Likewise.
8026 (loadqi_update): Likewise.
8027 (load_zeroextendqisi_update): Likewise.
8028 (load_signextendqisi_update): Likewise.
8029 (loadhi_update): Likewise.
8030 (load_zeroextendhisi_update): Likewise.
8031 (load_signextendhisi_update): Likewise.
8032 (loadsi_update): Likewise.
8033 (loadsf_update): Likewise.
8034 (movsicc_insn): Likewise.
8035 (bset_insn): Likewise.
8036 (bxor_insn): Likewise.
8037 (bclr_insn): Likewise.
8038 (bmsk_insn): Likewise.
8039 (bicsi3_insn): Likewise.
8040 (cmpsi_cc_c_insn): Likewise.
8041 (movsi_ne): Likewise.
8042 (movsi_cond_exec): Likewise.
8043 (clrsbsi2): Likewise.
8054 (call_i): Remove 'S' asm letter, add jli instruction.
8055 (call_value_i): Likewise.
8056 * config/arc/arc.op (mjli-always): New option.
8057 * config/arc/constraints.md (Cji): New constraint.
8058 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8060 (subsf3_fpx): Likewise.
8061 (mulsf3_fpx): Likewise.
8062 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8064 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8066 * doc/invoke.texi (ARC): Document mjli-always option.
8068 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8070 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8071 avoid addition with 0 and use incw and decw where possible.
8073 2018-01-26 Richard Biener <rguenther@suse.de>
8075 PR tree-optimization/81082
8076 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8077 association if it requires casting to unsigned.
8078 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8079 from fold_plusminus_mult_expr to catch important cases late when
8080 range info is available.
8082 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8084 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8085 * configure.ac (hidden_linkonce): New test.
8086 * configure: Regenerate.
8087 * config.in: Regenerate.
8089 2018-01-26 Julia Koval <julia.koval@intel.com>
8091 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8092 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8093 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8094 _mm_mask_bitshuffle_epi64_mask): Fix type.
8095 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8096 USI_FTYPE_V4DI_V4DI_USI): Remove.
8097 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8098 __builtin_ia32_vpshufbitqmb256_mask,
8099 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8100 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8101 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8103 2018-01-26 Alan Modra <amodra@gmail.com>
8106 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8107 UNSPEC_VBPERMQ. Sort other unspecs.
8109 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8111 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8113 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8116 * predict.c (drop_profile): Do not push/pop cfun; update also
8118 (handle_missing_profiles): Fix logic looking for zero profiles.
8120 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8123 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8124 on functions with #pragma omp declare simd or functions with simd
8126 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8127 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8128 Remove trailing \n from warning_at calls.
8130 2018-01-25 Tom de Vries <tom@codesourcery.com>
8133 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8134 for neutered workers.
8136 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8139 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8140 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8142 2018-01-24 Jeff Law <law@redhat.com>
8145 * i386.c (get_probe_interval): Move to earlier point.
8146 (ix86_compute_frame_layout): If -fstack-clash-protection and
8147 the frame is larger than the probe interval, then use pushes
8148 to save registers rather than reg->mem moves.
8149 (ix86_expand_prologue): Remove conditional for int_registers_saved
8152 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8155 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8156 min/max for never referenced object.
8158 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8161 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8163 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8164 attributes from DECL_ATTRIBUTES (decl) without affecting
8165 DECL_ATTRIBUTES (current_function_decl).
8166 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8167 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8169 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8171 PR tree-optimization/83979
8172 * fold-const.c (fold_comparison): Use constant_boolean_node
8173 instead of boolean_{true,false}_node.
8175 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8177 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8180 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8182 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8183 Simplify the clause that sets the length attribute.
8184 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8185 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8186 clause that sets the length attribute.
8187 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8189 2018-01-24 Tom de Vries <tom@codesourcery.com>
8192 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8193 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8194 Add strict parameter.
8195 (prevent_branch_around_nothing): Insert dummy insn between branch to
8196 label and label with no ptx insn inbetween.
8197 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8199 2018-01-24 Tom de Vries <tom@codesourcery.com>
8202 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8203 for neutered threads in warp.
8204 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8206 2018-01-24 Richard Biener <rguenther@suse.de>
8208 PR tree-optimization/83176
8209 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8212 2018-01-24 Richard Biener <rguenther@suse.de>
8214 PR tree-optimization/82819
8215 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8216 code generating pluses that are no-ops in the target precision.
8218 2018-01-24 Richard Biener <rguenther@suse.de>
8221 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8223 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8225 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8226 to merge probabilities.
8227 * predict.c (probably_never_executed): Also mark as cold functions
8228 with global 0 profile and guessed local profile.
8229 * profile-count.c (profile_probability::combine_with_count): New
8231 * profile-count.h (profile_probability::operator*,
8232 profile_probability::operator*=, profile_probability::operator/,
8233 profile_probability::operator/=): Reduce precision to adjusted
8234 and set value to guessed on contradictory divisions.
8235 (profile_probability::combine_with_freq): Remove.
8236 (profile_probability::combine_wiht_count): Declare.
8237 (profile_count::force_nonzero):: Set to adjusted.
8238 (profile_count::probability_in):: Set quality to adjusted.
8239 * tree-ssa-tail-merge.c (replace_block_by): Use
8242 2018-01-23 Andrew Waterman <andrew@sifive.com>
8243 Jim Wilson <jimw@sifive.com>
8245 * config/riscv/riscv.c (riscv_stack_boundary): New.
8246 (riscv_option_override): Set riscv_stack_boundary. Handle
8247 riscv_preferred_stack_boundary_arg.
8248 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8249 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8250 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8251 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8252 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8253 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8255 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8258 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8259 of struct ix86_frame.
8260 (ix86_expand_epilogue): Likewise. Add a local variable for
8261 the reg_save_offset field in struct ix86_frame.
8263 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8265 PR tree-optimization/82604
8266 * tree-loop-distribution.c (enum partition_kind): New enum item
8267 PKIND_PARTIAL_MEMSET.
8268 (partition_builtin_p): Support above new enum item.
8269 (generate_code_for_partition): Ditto.
8270 (compute_access_range): Differentiate cases that equality can be
8271 proven at all loops, the innermost loops or no loops.
8272 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8273 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8274 (finalize_partitions, distribute_loop): Don't fuse partition of
8275 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8276 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8279 2018-01-23 Martin Liska <mliska@suse.cz>
8281 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8282 order to ignore the predictor.
8283 (PRED_POLYMORPHIC_CALL): Likewise.
8284 (PRED_RECURSIVE_CALL): Likewise.
8286 2018-01-23 Martin Liska <mliska@suse.cz>
8288 * tree-profile.c (tree_profiling): Print function header to
8289 aware reader which function we are working on.
8290 * value-prof.c (gimple_find_values_to_profile): Do not print
8291 not interesting value histograms.
8293 2018-01-23 Martin Liska <mliska@suse.cz>
8295 * profile-count.h (enum profile_quality): Add
8296 profile_uninitialized as the first value. Do not number values
8297 as they are zero based.
8298 (profile_count::verify): Update sanity check.
8299 (profile_probability::verify): Likewise.
8301 2018-01-23 Nathan Sidwell <nathan@acm.org>
8303 * doc/invoke.texi (ffor-scope): Deprecate.
8305 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8307 PR tree-optimization/83510
8308 * domwalk.c (set_all_edges_as_executable): New function.
8309 (dom_walker::dom_walker): Convert bool param
8310 "skip_unreachable_blocks" to enum reachability. Move setup of
8311 edge flags to set_all_edges_as_executable and only do it when
8312 reachability is REACHABLE_BLOCKS.
8313 * domwalk.h (enum dom_walker::reachability): New enum.
8314 (dom_walker::dom_walker): Convert bool param
8315 "skip_unreachable_blocks" to enum reachability.
8316 (set_all_edges_as_executable): New decl.
8317 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8318 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8320 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8321 but converting true to REACHABLE_BLOCKS.
8322 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8324 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8325 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8326 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8328 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8329 if check_all_array_refs will be called.
8331 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8333 * tree.c (selftest::test_location_wrappers): Add more test
8336 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8338 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8339 (selftest::test_bit_in_range): Likewise.
8341 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8344 * doc/sourcebuild.texi (vect_float): Say that the selector
8345 only describes the situation when -funsafe-math-optimizations is on.
8346 (vect_float_strict): Document.
8348 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8350 PR tree-optimization/83965
8351 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8352 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8353 instead of checking only for a reduction.
8354 (vect_recog_widen_sum_pattern): Likewise.
8356 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8358 * predict.c (probably_never_executed): Only use precise profile info.
8359 (compute_function_frequency): Skip after inlining hack since we now
8360 have quality checking.
8362 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8364 * profile-count.h (profile_probability::very_unlikely,
8365 profile_probability::unlikely, profile_probability::even): Set
8366 precision to guessed.
8368 2018-01-23 Richard Biener <rguenther@suse.de>
8370 PR tree-optimization/83963
8371 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8372 Properly terminate dominator walk when crossing the exit edge not
8373 when visiting its source block.
8375 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8378 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8379 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8381 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8383 PR tree-optimization/83957
8384 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8385 semicolon after for body surrounded by braces.
8387 PR tree-optimization/83081
8388 * profile-count.h (profile_probability::split): New method.
8389 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8390 Use profile_probability::split.
8391 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8392 when splitting a single conditional jump into 2.
8394 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8396 PR tree-optimization/69452
8397 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8400 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8402 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8403 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8404 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8406 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8408 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8409 * config/rl78/rl78.md (movdi): New define_expand.
8410 * config/rl78/rl78.c (rl78_split_movdi): New function.
8412 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8415 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8417 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8418 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8419 128-bit to produce an UNSPEC move to get the double word with the
8420 signbit and then a shift directly to do signbit.
8421 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8422 implementation with a new version that just does either a direct
8423 move or a regular move. Move memory interface to separate insns.
8424 Move insns so they are next to the expander.
8425 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8426 with signbit move. Split big and little endian case.
8427 (signbit<mode>2_dm_mem_le): Likewise.
8428 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8429 (signbit<mode>2_dm2): Likewise.
8431 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8433 * config/rl78/rl78.md (anddi3): New define_expand.
8435 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8437 * config/rl78/rl78.md (umindi3): New define_expand.
8439 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8441 * config/rl78/rl78.md (smindi3): New define_expand.
8443 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8445 * config/rl78/rl78.md (smaxdi3): New define_expand.
8447 2018-01-22 Carl Love <cel@us.ibm.com>
8449 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8450 LVX_V1TI): Add macro expansion.
8451 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8452 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8453 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8454 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8455 Change check to determine if the instruction is a byte reversing
8456 entry. Fix typo in comment.
8457 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8458 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8459 Add def_builtin calls for new builtins.
8460 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8461 Add define_insn expansion.
8463 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8465 * config/rl78/rl78.md (umaxdi3): New define_expand.
8467 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8469 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8470 for non-QImode registers.
8472 2018-01-22 Richard Biener <rguenther@suse.de>
8474 PR tree-optimization/83963
8475 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8476 including the loop exit block.
8477 (scop_detection::merge_sese): Likewise.
8478 (scop_detection::add_scop): Do it here instead.
8480 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8482 * doc/sourcebuild.texi (arm_softfloat): Document.
8484 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8487 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8488 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8489 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8491 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8492 David Edelsohn <dje.gcc@gmail.com>
8495 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8496 Change "crset eq" to "crset 2".
8497 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8498 (*call_indirect_aix<mode>_nospec): Likewise.
8499 (*call_value_indirect_aix<mode>_nospec): Likewise.
8500 (*call_indirect_elfv2<mode>_nospec): Likewise.
8501 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8502 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8503 change assembly output from . to $.
8504 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8505 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8506 (*tablejump<mode>_internal1_nospec): Likewise.
8508 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8511 * config/sh/sh_optimize_sett_clrt.cc:
8512 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8514 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8516 PR tree-optimization/83940
8517 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8518 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8519 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8520 use it to pass back the definition type.
8521 (vect_check_store_rhs): Likewise rhs_dt_out.
8522 (vect_build_gather_load_calls): Add a mask_dt argument and use
8523 it instead of a call to vect_is_simple_use.
8524 (vectorizable_store): Update calls to vect_check_load_store_mask
8525 and vect_check_store_rhs. Use the dt returned by the latter instead
8526 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8527 instead of calls to vect_is_simple_use. Pass the scalar rather
8528 than the vector operand to vect_is_simple_use when handling
8529 second and subsequent copies of an rhs value.
8530 (vectorizable_load): Update calls to vect_check_load_store_mask
8531 and vect_build_gather_load_calls. Use the cached mask_dt and
8532 gs_info.offset_dt instead of calls to vect_is_simple_use.
8534 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8537 * tree-emutls.c: Include gimplify.h.
8538 (lower_emutls_2): New function.
8539 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8540 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8541 it before further processing.
8544 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8545 UINTVAL (trueop1) instead of INTVAL (op1).
8547 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8551 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8552 INCOMING_FRAME_SP_OFFSET if not defined.
8553 (scan_trace): Add ENTRY argument. If true and
8554 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8555 emit a note to adjust the CFA offset.
8556 (create_cfi_notes): Adjust scan_trace callers.
8557 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8558 INCOMING_FRAME_SP_OFFSET in the CIE.
8559 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8560 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8562 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8563 * doc/tm.texi: Regenerated.
8565 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8567 PR rtl-optimization/83147
8568 * lra-constraints.c (remove_inheritance_pseudos): Use
8569 lra_substitute_pseudo_within_insn.
8571 2018-01-19 Tom de Vries <tom@codesourcery.com>
8572 Cesar Philippidis <cesar@codesourcery.com>
8575 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8577 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8580 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8581 spaces for function labels.
8583 2018-01-19 Martin Liska <mliska@suse.cz>
8585 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8586 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8587 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8588 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8589 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8590 (PRED_CONST_RETURN): Change from 69 to 65.
8591 (PRED_NULL_RETURN): Change from 91 to 71.
8592 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8593 (PRED_LOOP_GUARD): Change from 66 to 73.
8595 2018-01-19 Martin Liska <mliska@suse.cz>
8597 * predict.c (predict_insn_def): Add new assert.
8598 (struct branch_predictor): Change type to signed integer.
8599 (test_prediction_value_range): Amend test to cover
8601 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8602 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8603 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8604 (PRED_LOOP_IV_COMPARE): Likewise.
8605 * predict.h (PROB_UNINITIALIZED): Define new constant.
8607 2018-01-19 Martin Liska <mliska@suse.cz>
8609 * predict.c (dump_prediction): Add new format for
8610 analyze_brprob.py script which is enabled with -details
8612 * profile-count.h (precise_p): New function.
8614 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8616 PR tree-optimization/83922
8617 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8618 there are no statements that need masking.
8619 (vect_active_double_reduction_p): New function.
8620 (vect_analyze_loop_operations): Use it when handling phis that
8621 are not in the loop header.
8623 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8625 PR tree-optimization/83914
8626 * tree-vect-loop.c (vectorizable_induction): Don't convert
8627 init_expr or apply the peeling adjustment for inductions
8628 that are nested within the vectorized loop.
8630 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8632 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8635 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8639 * function.h (gimplify_parameters): Add gimple_seq * argument.
8640 * function.c: Include gimple.h and options.h.
8641 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8642 for the added local temporaries if needed.
8643 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8644 if there are any parameter cleanups, wrap whole body into a
8645 try/finally with the cleanups.
8647 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8650 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8651 Use GET_MODE_CLASS for scalar floating point.
8653 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8657 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8658 Fix call of call_cgraph_insertion_hooks.
8660 2018-01-18 Martin Sebor <msebor@redhat.com>
8662 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8664 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8667 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8670 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8673 * common.opt: (-ffile-prefix-map): New option.
8674 * opts.c (common_handle_option): Defer it.
8675 * opts-global.c (handle_common_deferred_options): Handle it.
8676 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8677 * file-prefix-map.h: New file.
8678 (remap_debug_filename, add_debug_prefix_map): ...here.
8679 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8680 * final.c (debug_prefix_map, add_debug_prefix_map
8681 remap_debug_filename): Move to...
8682 * file-prefix-map.c: New file.
8683 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8684 generalize, get rid of alloca(), use strrchr() instead of strchr().
8685 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8686 Implement in terms of add_prefix_map().
8687 (remap_macro_filename, remap_debug_filename): Implement in term of
8689 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8690 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8691 * dbxout.c: Include file-prefix-map.h.
8692 * varasm.c: Likewise.
8693 * vmsdbgout.c: Likewise.
8694 * xcoffout.c: Likewise.
8695 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8696 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8697 * doc/invoke.texi (-ffile-prefix-map): Document.
8698 (-fdebug-prefix-map): Update description.
8700 2018-01-18 Martin Liska <mliska@suse.cz>
8702 * config/i386/i386.c (indirect_thunk_name): Document that also
8704 (output_indirect_thunk): Document why both instructions
8705 (pause and lfence) are generated.
8707 2018-01-18 Richard Biener <rguenther@suse.de>
8709 PR tree-optimization/83887
8710 * graphite-scop-detection.c
8711 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8712 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8713 (scop_detection::merge_sese): Re-implement with a flood-fill
8714 algorithm that properly finds a SESE region if it exists.
8716 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8719 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8720 pointer_diff optimizations use view_convert instead of convert.
8722 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8724 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8725 Generate different code for -mno-speculate-indirect-jumps.
8726 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8727 (*call_indirect_aix<mode>): Disable for
8728 -mno-speculate-indirect-jumps.
8729 (*call_indirect_aix<mode>_nospec): New define_insn.
8730 (*call_value_indirect_aix<mode>): Disable for
8731 -mno-speculate-indirect-jumps.
8732 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8733 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8734 -mno-speculate-indirect-jumps.
8735 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8737 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8739 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8740 long double type, set the flags for noting the default long double
8741 type, even if we don't pass or return a long double type.
8743 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8746 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8749 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8751 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8752 support for merge[hl].
8753 (fold_mergehl_helper): New helper function.
8754 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8755 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8756 (altivec_vmrglw_direct): Add xxmrglw insn.
8758 2018-01-17 Andrew Waterman <andrew@sifive.com>
8760 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8761 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8763 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8766 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8767 call the lto_location_cache before reading the
8768 DECL_SOURCE_LOCATION of the types.
8770 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8771 Richard Sandiford <richard.sandiford@linaro.org>
8773 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8774 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8775 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8776 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8777 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8779 * config/aarch64/constraints.md (aarch64_movti_operand):
8781 * config/aarch64/predicates.md (Uti): Add new constraint.
8783 2018-01-17 Carl Love <cel@us.ibm.com>
8785 * config/rs6000/vsx.md (define_expand xl_len_r,
8786 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8787 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8789 (define_expand, define_insn): Move the shift left from the
8790 define_insn to the define_expand for lxvl and stxvl instructions.
8791 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8792 and XL_LEN_R definitions to PURE.
8794 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8796 * config/i386/i386.c (indirect_thunk_name): Declare regno
8797 as unsigned int. Compare regno with INVALID_REGNUM.
8798 (output_indirect_thunk): Ditto.
8799 (output_indirect_thunk_function): Ditto.
8800 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8801 in the call to output_indirect_thunk_function.
8803 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8806 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8807 rather than the size of inner_type to determine the stack slot size
8808 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8810 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8813 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8816 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8818 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8819 endian Linux systems to optionally enable multilibs for selecting
8820 the long double type if the user configured an explicit type.
8821 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8822 have no long double multilibs if not defined.
8823 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8824 warn if the user used -mabi={ieee,ibm}longdouble and we built
8825 multilibs for long double.
8826 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8827 appropriate multilib option.
8828 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8830 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8831 for building long double multilibs.
8832 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8834 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8836 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8839 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8841 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8844 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8847 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8850 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8852 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8853 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8855 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8857 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8858 different rtl trees depending on TARGET_64BIT.
8859 (rs6000_gen_lvx): Likewise.
8861 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8863 * config/visium/visium.md (nop): Tweak comment.
8864 (hazard_nop): Likewise.
8866 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8868 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8869 -mspeculate-indirect-jumps.
8870 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8871 for -mno-speculate-indirect-jumps.
8872 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8873 (*call_value_indirect_elfv2<mode>): Disable for
8874 -mno-speculate-indirect-jumps.
8875 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8876 (indirect_jump): Emit different RTL for
8877 -mno-speculate-indirect-jumps.
8878 (*indirect_jump<mode>): Disable for
8879 -mno-speculate-indirect-jumps.
8880 (*indirect_jump<mode>_nospec): New define_insn.
8881 (tablejump): Emit different RTL for
8882 -mno-speculate-indirect-jumps.
8883 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8884 (tablejumpsi_nospec): New define_expand.
8885 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8886 (tablejumpdi_nospec): New define_expand.
8887 (*tablejump<mode>_internal1): Disable for
8888 -mno-speculate-indirect-jumps.
8889 (*tablejump<mode>_internal1_nospec): New define_insn.
8890 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8893 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8895 * caller-save.c (insert_save): Drop unnecessary parameter. All
8898 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8899 Richard Biener <rguenth@suse.de>
8902 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8903 return early, inline manually is_gimple_sizepos. Make sure if we
8904 call gimplify_expr we don't end up with a gimple constant.
8905 * tree.c (variably_modified_type_p): Don't return true for
8906 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8907 * gimplify.h (is_gimple_sizepos): Remove.
8909 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8911 PR tree-optimization/83857
8912 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8913 vectorizable_live_operation for pure SLP statements.
8914 (vectorizable_live_operation): Handle PHIs.
8916 2018-01-16 Richard Biener <rguenther@suse.de>
8918 PR tree-optimization/83867
8919 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8920 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8922 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8925 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8926 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8927 If off is not INTEGER_CST, issue a may not be aligned warning
8928 rather than isn't aligned. Use isn%'t rather than isn't.
8929 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8931 <case MULT_EXPR>: Improve the case when bottom and one of the
8932 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8933 operand, in that case check if the other operand is multiple of
8934 bottom divided by the INTEGER_CST operand.
8936 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8939 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8940 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8941 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8942 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8943 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8944 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8945 (pa_function_arg_size): New function.
8947 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8949 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8950 in a separate statement.
8952 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8954 PR tree-optimization/83847
8955 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8956 group gathers and scatters.
8958 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8960 PR rtl-optimization/86620
8961 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8963 PR rtl-optimization/83213
8964 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8965 to last if both are JUMP_INSNs.
8967 PR tree-optimization/83843
8968 * gimple-ssa-store-merging.c
8969 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8970 store_immediate_info for bswap/nop orig_stores.
8972 2018-01-15 Andrew Waterman <andrew@sifive.com>
8974 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8976 <UDIV>: Increase cost if !TARGET_DIV.
8978 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8980 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8981 (define_attr "cr_logical_3op"): New.
8982 (cceq_ior_compare): Adjust.
8983 (cceq_ior_compare_complement): Adjust.
8984 (*cceq_rev_compare): Adjust.
8985 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8986 (is_cracked_insn): Adjust.
8987 (insn_must_be_first_in_group): Adjust.
8988 * config/rs6000/40x.md: Adjust.
8989 * config/rs6000/440.md: Adjust.
8990 * config/rs6000/476.md: Adjust.
8991 * config/rs6000/601.md: Adjust.
8992 * config/rs6000/603.md: Adjust.
8993 * config/rs6000/6xx.md: Adjust.
8994 * config/rs6000/7450.md: Adjust.
8995 * config/rs6000/7xx.md: Adjust.
8996 * config/rs6000/8540.md: Adjust.
8997 * config/rs6000/cell.md: Adjust.
8998 * config/rs6000/e300c2c3.md: Adjust.
8999 * config/rs6000/e500mc.md: Adjust.
9000 * config/rs6000/e500mc64.md: Adjust.
9001 * config/rs6000/e5500.md: Adjust.
9002 * config/rs6000/e6500.md: Adjust.
9003 * config/rs6000/mpc.md: Adjust.
9004 * config/rs6000/power4.md: Adjust.
9005 * config/rs6000/power5.md: Adjust.
9006 * config/rs6000/power6.md: Adjust.
9007 * config/rs6000/power7.md: Adjust.
9008 * config/rs6000/power8.md: Adjust.
9009 * config/rs6000/power9.md: Adjust.
9010 * config/rs6000/rs64.md: Adjust.
9011 * config/rs6000/titan.md: Adjust.
9013 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9015 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9016 ix86_indirect_branch_register logic.
9018 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9020 * config/i386/constraints.md (Bs): Update
9021 ix86_indirect_branch_register check. Don't check
9022 ix86_indirect_branch_register with GOT_memory_operand.
9024 * config/i386/predicates.md (GOT_memory_operand): Don't check
9025 ix86_indirect_branch_register here.
9026 (GOT32_symbol_operand): Likewise.
9028 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9030 * config/i386/predicates.md (constant_call_address_operand):
9031 Rewrite ix86_indirect_branch_register logic.
9032 (sibcall_insn_operand): Likewise.
9034 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9036 * config/i386/constraints.md (Bs): Replace
9037 ix86_indirect_branch_thunk_register with
9038 ix86_indirect_branch_register.
9040 * config/i386/i386.md (indirect_jump): Likewise.
9041 (tablejump): Likewise.
9042 (*sibcall_memory): Likewise.
9043 (*sibcall_value_memory): Likewise.
9044 Peepholes of indirect call and jump via memory: Likewise.
9045 * config/i386/i386.opt: Likewise.
9046 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9047 (GOT_memory_operand): Likewise.
9048 (call_insn_operand): Likewise.
9049 (sibcall_insn_operand): Likewise.
9050 (GOT32_symbol_operand): Likewise.
9052 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9055 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9056 type rather than type addr's type points to.
9057 (expand_omp_atomic_mutex): Likewise.
9058 (expand_omp_atomic): Likewise.
9060 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9063 * config/i386/i386.c (output_indirect_thunk_function): Use
9064 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9065 for __x86_return_thunk.
9067 2018-01-15 Richard Biener <rguenther@suse.de>
9070 * expmed.c (extract_bit_field_1): Fix typo.
9072 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9075 * config/arm/iterators.md (VF): New mode iterator.
9076 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9077 Remove integer-related logic from pattern.
9078 (neon_vabd<mode>_3): Likewise.
9080 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9083 * common.opt (fstrict-overflow): No longer an alias.
9084 (fwrapv-pointer): New option.
9085 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9086 also for pointer types based on flag_wrapv_pointer.
9087 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9088 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9089 opts->x_flag_wrapv got set.
9090 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9091 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9092 POINTER_TYPE_OVERFLOW_UNDEFINED.
9093 * match.pd: Likewise in address comparison pattern.
9094 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9096 2018-01-15 Richard Biener <rguenther@suse.de>
9099 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9100 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9101 Reset type names to their identifier if their TYPE_DECL doesn't
9102 have linkage (and thus is used for ODR and devirt).
9103 (save_debug_info_for_decl): Remove.
9104 (save_debug_info_for_type): Likewise.
9105 (add_tree_to_fld_list): Adjust.
9106 * tree-pretty-print.c (dump_generic_node): Make dumping of
9107 type names more robust.
9109 2018-01-15 Richard Biener <rguenther@suse.de>
9111 * BASE-VER: Bump to 8.0.1.
9113 2018-01-14 Martin Sebor <msebor@redhat.com>
9116 * builtins.c (check_access): Avoid warning when the no-warning bit
9119 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9121 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9122 * ira-color (allocno_hard_regs_compare): Likewise.
9124 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9127 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9128 Use .pushsection/.popsection.
9130 2018-01-14 Martin Sebor <msebor@redhat.com>
9133 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9135 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9137 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9138 entry from extra_headers.
9139 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9140 extra_headers, make the list bitwise identical to the i?86-*-* one.
9142 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9144 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9145 -mcmodel=large with -mindirect-branch=thunk,
9146 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9147 -mfunction-return=thunk-extern.
9148 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9149 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9150 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9152 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9154 * config/i386/i386.c (print_reg): Print the name of the full
9155 integer register without '%'.
9156 (ix86_print_operand): Handle 'V'.
9157 * doc/extend.texi: Document 'V' modifier.
9159 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9161 * config/i386/constraints.md (Bs): Disallow memory operand for
9162 -mindirect-branch-register.
9164 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9165 (GOT_memory_operand): Likewise.
9166 (call_insn_operand): Likewise.
9167 (sibcall_insn_operand): Likewise.
9168 (GOT32_symbol_operand): Likewise.
9169 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9170 for -mindirect-branch-register.
9171 (tablejump): Likewise.
9172 (*sibcall_memory): Likewise.
9173 (*sibcall_value_memory): Likewise.
9174 Disallow peepholes of indirect call and jump via memory for
9175 -mindirect-branch-register.
9176 (*call_pop): Replace m with Bw.
9177 (*call_value_pop): Likewise.
9178 (*sibcall_pop_memory): Replace m with Bs.
9179 * config/i386/i386.opt (mindirect-branch-register): New option.
9180 * doc/invoke.texi: Document -mindirect-branch-register option.
9182 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9184 * config/i386/i386-protos.h (ix86_output_function_return): New.
9185 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9186 set function_return_type.
9187 (indirect_thunk_name): Add ret_p to indicate thunk for function
9189 (output_indirect_thunk_function): Pass false to
9190 indirect_thunk_name.
9191 (ix86_output_indirect_branch_via_reg): Likewise.
9192 (ix86_output_indirect_branch_via_push): Likewise.
9193 (output_indirect_thunk_function): Create alias for function
9194 return thunk if regno < 0.
9195 (ix86_output_function_return): New function.
9196 (ix86_handle_fndecl_attribute): Handle function_return.
9197 (ix86_attribute_table): Add function_return.
9198 * config/i386/i386.h (machine_function): Add
9199 function_return_type.
9200 * config/i386/i386.md (simple_return_internal): Use
9201 ix86_output_function_return.
9202 (simple_return_internal_long): Likewise.
9203 * config/i386/i386.opt (mfunction-return=): New option.
9204 (indirect_branch): Mention -mfunction-return=.
9205 * doc/extend.texi: Document function_return function attribute.
9206 * doc/invoke.texi: Document -mfunction-return= option.
9208 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9210 * config/i386/i386-opts.h (indirect_branch): New.
9211 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9212 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9213 with local indirect jump when converting indirect call and jump.
9214 (ix86_set_indirect_branch_type): New.
9215 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9216 (indirectlabelno): New.
9217 (indirect_thunk_needed): Likewise.
9218 (indirect_thunk_bnd_needed): Likewise.
9219 (indirect_thunks_used): Likewise.
9220 (indirect_thunks_bnd_used): Likewise.
9221 (INDIRECT_LABEL): Likewise.
9222 (indirect_thunk_name): Likewise.
9223 (output_indirect_thunk): Likewise.
9224 (output_indirect_thunk_function): Likewise.
9225 (ix86_output_indirect_branch_via_reg): Likewise.
9226 (ix86_output_indirect_branch_via_push): Likewise.
9227 (ix86_output_indirect_branch): Likewise.
9228 (ix86_output_indirect_jmp): Likewise.
9229 (ix86_code_end): Call output_indirect_thunk_function if needed.
9230 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9232 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9233 (ix86_attribute_table): Add indirect_branch.
9234 * config/i386/i386.h (machine_function): Add indirect_branch_type
9235 and has_local_indirect_jump.
9236 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9238 (tablejump): Likewise.
9239 (*indirect_jump): Use ix86_output_indirect_jmp.
9240 (*tablejump_1): Likewise.
9241 (simple_return_indirect_internal): Likewise.
9242 * config/i386/i386.opt (mindirect-branch=): New option.
9243 (indirect_branch): New.
9246 (thunk-inline): Likewise.
9247 (thunk-extern): Likewise.
9248 * doc/extend.texi: Document indirect_branch function attribute.
9249 * doc/invoke.texi: Document -mindirect-branch= option.
9251 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9254 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9256 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9258 * ipa-inline.c (want_inline_small_function_p): Return false if
9259 inlining has already failed with CIF_FINAL_ERROR.
9260 (update_caller_keys): Call want_inline_small_function_p before
9262 (update_callee_keys): Likewise.
9264 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9266 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9268 (rs6000_quadword_masked_address_p): Likewise.
9269 (quad_aligned_load_p): Likewise.
9270 (quad_aligned_store_p): Likewise.
9271 (const_load_sequence_p): Add comment to describe the outer-most loop.
9272 (mimic_memory_attributes_and_flags): New function.
9273 (rs6000_gen_stvx): Likewise.
9274 (replace_swapped_aligned_store): Likewise.
9275 (rs6000_gen_lvx): Likewise.
9276 (replace_swapped_aligned_load): Likewise.
9277 (replace_swapped_load_constant): Capitalize argument name in
9278 comment describing this function.
9279 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9280 and stores that access quad-word aligned addresses and replace
9281 with stvx or lvx instructions when appropriate.
9282 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9283 New function prototype.
9284 (rs6000_quadword_masked_address_p): Likewise.
9285 (rs6000_gen_lvx): Likewise.
9286 (rs6000_gen_stvx): Likewise.
9287 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9288 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9289 when memory address is aligned.
9290 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9291 this split to select lvx instruction when memory address is aligned.
9292 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9293 instruction when memory address is aligned.
9294 (*vsx_le_perm_load_v16qi): Likewise.
9295 (four unnamed splitters): Modify to select the stvx instruction
9296 when memory is aligned.
9298 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9300 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9301 which appears in the queue multiple times.
9303 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9304 Alan Hayward <alan.hayward@arm.com>
9305 David Sherwood <david.sherwood@arm.com>
9307 * tree-vectorizer.h (vec_lower_bound): New structure.
9308 (_loop_vec_info): Add check_nonzero and lower_bounds.
9309 (LOOP_VINFO_CHECK_NONZERO): New macro.
9310 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9311 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9312 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9313 fields. Make seg_len the distance travelled, not including the
9315 (dr_direction_indicator): Declare.
9316 (dr_zero_step_indicator): Likewise.
9317 (dr_known_forward_stride_p): Likewise.
9318 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9320 (runtime_alias_check_p): Allow runtime alias checks with
9322 (operator ==): Compare access_size and align.
9323 (prune_runtime_alias_test_list): Rework for new distinction between
9324 the access_size and seg_len.
9325 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9327 (get_segment_min_max): New function.
9328 (create_intersect_range_checks): Use it.
9329 (dr_step_indicator): New function.
9330 (dr_direction_indicator): Likewise.
9331 (dr_zero_step_indicator): Likewise.
9332 (dr_known_forward_stride_p): Likewise.
9333 * tree-loop-distribution.c (data_ref_segment_size): Return
9334 DR_STEP * (niters - 1).
9335 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9337 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9338 (vect_preserves_scalar_order_p): New function, split out from...
9339 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9340 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9341 (vect_vfa_access_size): New function.
9342 (vect_vfa_align): Likewise.
9343 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9344 (dump_lower_bound): New function.
9345 (vect_check_lower_bound): Likewise.
9346 (vect_small_gap_p): Likewise.
9347 (vectorizable_with_step_bound_p): Likewise.
9348 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9349 depencies if the vectorization factor is 1. Convert the checks
9350 for nonzero steps into checks on the bounds of DR_STEP. Try using
9351 a bunds check for variable steps if the minimum required step is
9352 relatively small. Update calls to the dr_with_seg_len
9353 constructor and to vect_compile_time_alias.
9354 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9356 (vect_loop_versioning): Call it.
9357 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9359 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9361 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9362 Alan Hayward <alan.hayward@arm.com>
9363 David Sherwood <david.sherwood@arm.com>
9365 * doc/sourcebuild.texi (vect_scatter_store): Document.
9366 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9368 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9370 * genopinit.c (main): Add supports_vec_scatter_store and
9371 supports_vec_scatter_store_cached to target_optabs.
9372 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9373 IFN_MASK_SCATTER_STORE.
9374 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9376 * internal-fn.h (internal_store_fn_p): Declare.
9377 (internal_fn_stored_value_index): Likewise.
9378 * internal-fn.c (scatter_store_direct): New macro.
9379 (expand_scatter_store_optab_fn): New function.
9380 (direct_scatter_store_optab_supported_p): New macro.
9381 (internal_store_fn_p): New function.
9382 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9383 IFN_MASK_SCATTER_STORE.
9384 (internal_fn_mask_index): Likewise.
9385 (internal_fn_stored_value_index): New function.
9386 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9388 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9389 * optabs-query.c (supports_vec_scatter_store_p): New function.
9390 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9391 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9392 true for scatter stores.
9393 (vect_gather_scatter_fn_p): Handle scatter stores too.
9394 (vect_check_gather_scatter): Consider using scatter stores if
9395 supports_vec_scatter_store_p.
9396 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9398 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9399 internal_fn_stored_value_index.
9400 (check_load_store_masking): Handle scatter stores too.
9401 (vect_get_store_rhs): Make public.
9402 (vectorizable_call): Use internal_store_fn_p.
9403 (vectorizable_store): Handle scatter store internal functions.
9404 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9405 when deciding whether the end of the group has been reached.
9406 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9407 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9408 (mask_scatter_store<mode>): New insns.
9410 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9411 Alan Hayward <alan.hayward@arm.com>
9412 David Sherwood <david.sherwood@arm.com>
9414 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9415 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9416 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9418 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9419 Use vect_truncate_gather_scatter_offset if we can't treat the
9420 operation as a normal gather load or scatter store.
9421 (get_group_load_store_type): Take the gather_scatter_info
9422 as argument. Try using a gather load or scatter store for
9423 single-element groups.
9424 (get_load_store_type): Update calls to get_group_load_store_type
9425 and vect_use_strided_gather_scatters_p.
9427 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9428 Alan Hayward <alan.hayward@arm.com>
9429 David Sherwood <david.sherwood@arm.com>
9431 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9432 optional tree argument.
9433 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9435 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9436 but continue to use the current value as a fallback.
9437 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9438 to compare the updates.
9439 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9440 (get_load_store_type): Use it when handling a strided access.
9441 (vect_get_strided_load_store_ops): New function.
9442 (vect_get_data_ptr_increment): Likewise.
9443 (vectorizable_load): Handle strided gather loads. Always pass
9444 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9446 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9447 Alan Hayward <alan.hayward@arm.com>
9448 David Sherwood <david.sherwood@arm.com>
9450 * doc/md.texi (gather_load@var{m}): Document.
9451 (mask_gather_load@var{m}): Likewise.
9452 * genopinit.c (main): Add supports_vec_gather_load and
9453 supports_vec_gather_load_cached to target_optabs.
9454 * optabs-tree.c (init_tree_optimization_optabs): Use
9455 ggc_cleared_alloc to allocate target_optabs.
9456 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9457 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9459 * internal-fn.h (internal_load_fn_p): Declare.
9460 (internal_gather_scatter_fn_p): Likewise.
9461 (internal_fn_mask_index): Likewise.
9462 (internal_gather_scatter_fn_supported_p): Likewise.
9463 * internal-fn.c (gather_load_direct): New macro.
9464 (expand_gather_load_optab_fn): New function.
9465 (direct_gather_load_optab_supported_p): New macro.
9466 (direct_internal_fn_optab): New function.
9467 (internal_load_fn_p): Likewise.
9468 (internal_gather_scatter_fn_p): Likewise.
9469 (internal_fn_mask_index): Likewise.
9470 (internal_gather_scatter_fn_supported_p): Likewise.
9471 * optabs-query.c (supports_at_least_one_mode_p): New function.
9472 (supports_vec_gather_load_p): Likewise.
9473 * optabs-query.h (supports_vec_gather_load_p): Declare.
9474 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9475 and memory_type field.
9476 (NUM_PATTERNS): Bump to 15.
9477 * tree-vect-data-refs.c: Include internal-fn.h.
9478 (vect_gather_scatter_fn_p): New function.
9479 (vect_describe_gather_scatter_call): Likewise.
9480 (vect_check_gather_scatter): Try using internal functions for
9481 gather loads. Recognize existing calls to a gather load function.
9482 (vect_analyze_data_refs): Consider using gather loads if
9483 supports_vec_gather_load_p.
9484 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9485 (vect_get_gather_scatter_offset_type): Likewise.
9486 (vect_convert_mask_for_vectype): Likewise.
9487 (vect_add_conversion_to_patterm): Likewise.
9488 (vect_try_gather_scatter_pattern): Likewise.
9489 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9490 (vect_vect_recog_func_ptrs): Add it.
9491 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9492 internal_fn_mask_index and internal_gather_scatter_fn_p.
9493 (check_load_store_masking): Take the gather_scatter_info as an
9494 argument and handle gather loads.
9495 (vect_get_gather_scatter_ops): New function.
9496 (vectorizable_call): Check internal_load_fn_p.
9497 (vectorizable_load): Likewise. Handle gather load internal
9499 (vectorizable_store): Update call to check_load_store_masking.
9500 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9501 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9502 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9503 (aarch64_gather_scale_operand_d): New predicates.
9504 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9505 (mask_gather_load<mode>): New insns.
9507 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9508 Alan Hayward <alan.hayward@arm.com>
9509 David Sherwood <david.sherwood@arm.com>
9511 * optabs.def (fold_left_plus_optab): New optab.
9512 * doc/md.texi (fold_left_plus_@var{m}): Document.
9513 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9514 * internal-fn.c (fold_left_direct): Define.
9515 (expand_fold_left_optab_fn): Likewise.
9516 (direct_fold_left_optab_supported_p): Likewise.
9517 * fold-const-call.c (fold_const_fold_left): New function.
9518 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9519 * tree-parloops.c (valid_reduction_p): New function.
9520 (gather_scalar_reductions): Use it.
9521 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9522 (vect_finish_replace_stmt): Declare.
9523 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9524 (needs_fold_left_reduction_p): New function, split out from...
9525 (vect_is_simple_reduction): ...here. Accept reductions that
9526 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9527 (vect_force_simple_reduction): Also store the reduction type in
9528 the assignment's STMT_VINFO_REDUC_TYPE.
9529 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9530 (merge_with_identity): New function.
9531 (vect_expand_fold_left): Likewise.
9532 (vectorize_fold_left_reduction): Likewise.
9533 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9534 scalar phi in place for it. Check for target support and reject
9535 cases that would reassociate the operation. Defer the transform
9536 phase to vectorize_fold_left_reduction.
9537 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9538 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9539 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9541 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9543 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9544 call to ifc_temp_var.
9546 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9547 Alan Hayward <alan.hayward@arm.com>
9548 David Sherwood <david.sherwood@arm.com>
9550 * target.def (legitimize_address_displacement): Take the original
9551 offset as a poly_int.
9552 * targhooks.h (default_legitimize_address_displacement): Update
9554 * targhooks.c (default_legitimize_address_displacement): Likewise.
9555 * doc/tm.texi: Regenerate.
9556 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9557 as an argument, moving assert of ad->disp == ad->disp_term to...
9558 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9559 Try calling targetm.legitimize_address_displacement before expanding
9560 the address rather than afterwards, and adjust for the new interface.
9561 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9562 Match the new hook interface. Handle SVE addresses.
9563 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9566 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9568 * Makefile.in (OBJS): Add early-remat.o.
9569 * target.def (select_early_remat_modes): New hook.
9570 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9571 * doc/tm.texi: Regenerate.
9572 * targhooks.h (default_select_early_remat_modes): Declare.
9573 * targhooks.c (default_select_early_remat_modes): New function.
9574 * timevar.def (TV_EARLY_REMAT): New timevar.
9575 * passes.def (pass_early_remat): New pass.
9576 * tree-pass.h (make_pass_early_remat): Declare.
9577 * early-remat.c: New file.
9578 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9580 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9582 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9583 Alan Hayward <alan.hayward@arm.com>
9584 David Sherwood <david.sherwood@arm.com>
9586 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9587 vfm1 with a bound_epilog parameter.
9588 (vect_do_peeling): Update calls accordingly, and move the prologue
9589 call earlier in the function. Treat the base bound_epilog as 0 for
9590 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9591 this base when peeling for gaps.
9592 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9593 with fully-masked loops.
9594 (vect_estimate_min_profitable_iters): Handle the single peeled
9595 iteration in that case.
9597 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9598 Alan Hayward <alan.hayward@arm.com>
9599 David Sherwood <david.sherwood@arm.com>
9601 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9602 single-element interleaving even if the size is not a power of 2.
9603 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9604 accesses for single-element interleaving if the group size is
9607 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9608 Alan Hayward <alan.hayward@arm.com>
9609 David Sherwood <david.sherwood@arm.com>
9611 * doc/md.texi (fold_extract_last_@var{m}): Document.
9612 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9613 * optabs.def (fold_extract_last_optab): New optab.
9614 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9615 * internal-fn.c (fold_extract_direct): New macro.
9616 (expand_fold_extract_optab_fn): Likewise.
9617 (direct_fold_extract_optab_supported_p): Likewise.
9618 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9619 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9620 EXTRACT_LAST_REDUCTION.
9621 (get_initial_def_for_reduction): Do not create an initial vector
9622 for EXTRACT_LAST_REDUCTION reductions.
9623 (vectorizable_reduction): Leave the scalar phi in place for
9624 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9625 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9626 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9627 transform phase to vectorizable_condition.
9628 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9630 (vect_finish_stmt_generation): ...here.
9631 (vect_finish_replace_stmt): New function.
9632 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9633 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9635 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9637 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9638 Alan Hayward <alan.hayward@arm.com>
9639 David Sherwood <david.sherwood@arm.com>
9641 * doc/md.texi (extract_last_@var{m}): Document.
9642 * optabs.def (extract_last_optab): New optab.
9643 * internal-fn.def (EXTRACT_LAST): New internal function.
9644 * internal-fn.c (cond_unary_direct): New macro.
9645 (expand_cond_unary_optab_fn): Likewise.
9646 (direct_cond_unary_optab_supported_p): Likewise.
9647 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9648 loops using EXTRACT_LAST.
9649 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9650 (extract_last_<mode>): ...this optab.
9651 (vec_extract<mode><Vel>): Update accordingly.
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 * target.def (empty_mask_is_expensive): New hook.
9658 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9659 * doc/tm.texi: Regenerate.
9660 * targhooks.h (default_empty_mask_is_expensive): Declare.
9661 * targhooks.c (default_empty_mask_is_expensive): New function.
9662 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9663 if the target says that empty masks are expensive.
9664 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9666 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9668 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9669 Alan Hayward <alan.hayward@arm.com>
9670 David Sherwood <david.sherwood@arm.com>
9672 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9673 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9674 (vect_use_loop_mask_for_alignment_p): New function.
9675 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9676 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9677 niters_skip argument. Make sure that the first niters_skip elements
9678 of the first iteration are inactive.
9679 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9680 Update call to vect_set_loop_masks_directly.
9681 (get_misalign_in_elems): New function, split out from...
9682 (vect_gen_prolog_loop_niters): ...here.
9683 (vect_update_init_of_dr): Take a code argument that specifies whether
9684 the adjustment should be added or subtracted.
9685 (vect_update_init_of_drs): Likewise.
9686 (vect_prepare_for_masked_peels): New function.
9687 (vect_do_peeling): Skip prologue peeling if we're using a mask
9688 instead. Update call to vect_update_inits_of_drs.
9689 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9691 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9692 alignment. Do not include the number of peeled iterations in
9693 the minimum threshold in that case.
9694 (vectorizable_induction): Adjust the start value down by
9695 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9696 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9697 Take the number of skipped iterations into account when calculating
9699 * tree-vect-stmts.c (vect_gen_while_not): New function.
9701 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9702 Alan Hayward <alan.hayward@arm.com>
9703 David Sherwood <david.sherwood@arm.com>
9705 * doc/sourcebuild.texi (vect_fully_masked): Document.
9706 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9708 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9710 (vect_analyze_loop_2): ...here. Don't check the vectorization
9711 factor against the number of loop iterations if the loop is
9714 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9715 Alan Hayward <alan.hayward@arm.com>
9716 David Sherwood <david.sherwood@arm.com>
9718 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9719 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9720 (dump_groups): Update accordingly.
9721 (iv_use::mem_type): New member variable.
9722 (address_p): New function.
9723 (record_use): Add a mem_type argument and initialize the new
9725 (record_group_use): Add a mem_type argument. Use address_p.
9726 Remove obsolete null checks of base_object. Update call to record_use.
9727 (find_interesting_uses_op): Update call to record_group_use.
9728 (find_interesting_uses_cond): Likewise.
9729 (find_interesting_uses_address): Likewise.
9730 (get_mem_type_for_internal_fn): New function.
9731 (find_address_like_use): Likewise.
9732 (find_interesting_uses_stmt): Try find_address_like_use before
9733 calling find_interesting_uses_op.
9734 (addr_offset_valid_p): Use the iv mem_type field as the type
9735 of the addressed memory.
9736 (add_autoinc_candidates): Likewise.
9737 (get_address_cost): Likewise.
9738 (split_small_address_groups_p): Use address_p.
9739 (split_address_groups): Likewise.
9740 (add_iv_candidate_for_use): Likewise.
9741 (autoinc_possible_for_pair): Likewise.
9742 (rewrite_groups): Likewise.
9743 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9744 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9745 (get_alias_ptr_type_for_ptr_address): New function.
9746 (rewrite_use_address): Rewrite address uses in calls that were
9747 identified by find_address_like_use.
9749 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9750 Alan Hayward <alan.hayward@arm.com>
9751 David Sherwood <david.sherwood@arm.com>
9753 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9755 * gimple-expr.h (is_gimple_addressable: Likewise.
9756 * gimple-expr.c (is_gimple_address): Likewise.
9757 * internal-fn.c (expand_call_mem_ref): New function.
9758 (expand_mask_load_optab_fn): Use it.
9759 (expand_mask_store_optab_fn): Likewise.
9761 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9762 Alan Hayward <alan.hayward@arm.com>
9763 David Sherwood <david.sherwood@arm.com>
9765 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9766 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9767 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9768 (cond_umax@var{mode}): Document.
9769 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9770 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9771 (cond_umin_optab, cond_umax_optab): New optabs.
9772 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9773 (COND_IOR, COND_XOR): New internal functions.
9774 * internal-fn.h (get_conditional_internal_fn): Declare.
9775 * internal-fn.c (cond_binary_direct): New macro.
9776 (expand_cond_binary_optab_fn): Likewise.
9777 (direct_cond_binary_optab_supported_p): Likewise.
9778 (get_conditional_internal_fn): New function.
9779 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9780 Cope with reduction statements that are vectorized as calls rather
9782 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9783 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9784 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9785 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9786 (UNSPEC_COND_EOR): New unspecs.
9787 (optab): Add mappings for them.
9788 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9789 (sve_int_op, sve_fp_op): New int attributes.
9791 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9792 Alan Hayward <alan.hayward@arm.com>
9793 David Sherwood <david.sherwood@arm.com>
9795 * optabs.def (while_ult_optab): New optab.
9796 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9797 * internal-fn.def (WHILE_ULT): New internal function.
9798 * internal-fn.h (direct_internal_fn_supported_p): New override
9799 that takes two types as argument.
9800 * internal-fn.c (while_direct): New macro.
9801 (expand_while_optab_fn): New function.
9802 (convert_optab_supported_p): Likewise.
9803 (direct_while_optab_supported_p): New macro.
9804 * wide-int.h (wi::udiv_ceil): New function.
9805 * tree-vectorizer.h (rgroup_masks): New structure.
9806 (vec_loop_masks): New typedef.
9807 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9809 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9810 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9811 (vect_max_vf): New function.
9812 (slpeel_make_loop_iterate_ntimes): Delete.
9813 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9814 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9815 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9816 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9817 internal-fn.h, stor-layout.h and optabs-query.h.
9818 (vect_set_loop_mask): New function.
9819 (add_preheader_seq): Likewise.
9820 (add_header_seq): Likewise.
9821 (interleave_supported_p): Likewise.
9822 (vect_maybe_permute_loop_masks): Likewise.
9823 (vect_set_loop_masks_directly): Likewise.
9824 (vect_set_loop_condition_masked): Likewise.
9825 (vect_set_loop_condition_unmasked): New function, split out from
9826 slpeel_make_loop_iterate_ntimes.
9827 (slpeel_make_loop_iterate_ntimes): Rename to..
9828 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9829 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9830 (vect_do_peeling): Update call accordingly.
9831 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9833 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9834 mask_compare_type, can_fully_mask_p and fully_masked_p.
9835 (release_vec_loop_masks): New function.
9836 (_loop_vec_info): Use it to free the loop masks.
9837 (can_produce_all_loop_masks_p): New function.
9838 (vect_get_max_nscalars_per_iter): Likewise.
9839 (vect_verify_full_masking): Likewise.
9840 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9841 retries, and free the mask rgroups before retrying. Check loop-wide
9842 reasons for disallowing fully-masked loops. Make the final decision
9843 about whether use a fully-masked loop or not.
9844 (vect_estimate_min_profitable_iters): Do not assume that peeling
9845 for the number of iterations will be needed for fully-masked loops.
9846 (vectorizable_reduction): Disable fully-masked loops.
9847 (vectorizable_live_operation): Likewise.
9848 (vect_halve_mask_nunits): New function.
9849 (vect_double_mask_nunits): Likewise.
9850 (vect_record_loop_mask): Likewise.
9851 (vect_get_loop_mask): Likewise.
9852 (vect_transform_loop): Handle the case in which the final loop
9853 iteration might handle a partial vector. Call vect_set_loop_condition
9854 instead of slpeel_make_loop_iterate_ntimes.
9855 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9856 (check_load_store_masking): New function.
9857 (prepare_load_store_mask): Likewise.
9858 (vectorizable_store): Handle fully-masked loops.
9859 (vectorizable_load): Likewise.
9860 (supportable_widening_operation): Use vect_halve_mask_nunits for
9862 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9863 (vect_gen_while): New function.
9864 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9865 (aarch64_uqdec<mode>): New insn.
9867 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9868 Alan Hayward <alan.hayward@arm.com>
9869 David Sherwood <david.sherwood@arm.com>
9871 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9872 (reduc_xor_scal_optab): New optabs.
9873 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9874 (reduc_xor_scal_@var{m}): Document.
9875 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9876 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9878 * fold-const-call.c (fold_const_call): Handle them.
9879 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9880 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9881 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9882 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9883 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9884 (UNSPEC_XORV): New unspecs.
9885 (optab): Add entries for them.
9886 (BITWISEV): New int iterator.
9887 (bit_reduc_op): New int attributes.
9889 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9890 Alan Hayward <alan.hayward@arm.com>
9891 David Sherwood <david.sherwood@arm.com>
9893 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9894 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9895 * optabs.def (vec_shl_insert_optab): New optab.
9896 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9897 (duplicate_and_interleave): Likewise.
9898 * tree-vect-loop.c: Include internal-fn.h.
9899 (neutral_op_for_slp_reduction): New function, split out from
9900 get_initial_defs_for_reduction.
9901 (get_initial_def_for_reduction): Handle option 2 for variable-length
9902 vectors by loading the neutral value into a vector and then shifting
9903 the initial value into element 0.
9904 (get_initial_defs_for_reduction): Replace the code argument with
9905 the neutral value calculated by neutral_op_for_slp_reduction.
9906 Use gimple_build_vector for constant-length vectors.
9907 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9908 but the first group_size elements have a neutral value.
9909 Use duplicate_and_interleave otherwise.
9910 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9911 Update call to get_initial_defs_for_reduction. Handle SLP
9912 reductions for variable-length vectors by creating one vector
9913 result for each scalar result, with the elements associated
9914 with other scalar results stubbed out with the neutral value.
9915 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9916 Require IFN_VEC_SHL_INSERT for double reductions on
9917 variable-length vectors, or SLP reductions that have
9918 a neutral value. Require can_duplicate_and_interleave_p
9919 support for variable-length unchained SLP reductions if there
9920 is no neutral value, such as for MIN/MAX reductions. Also require
9921 the number of vector elements to be a multiple of the number of
9922 SLP statements when doing variable-length unchained SLP reductions.
9923 Update call to vect_create_epilog_for_reduction.
9924 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9925 and remove initial values.
9926 (duplicate_and_interleave): Make public.
9927 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9928 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9930 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9931 Alan Hayward <alan.hayward@arm.com>
9932 David Sherwood <david.sherwood@arm.com>
9934 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9935 (can_duplicate_and_interleave_p): New function.
9936 (vect_get_and_check_slp_defs): Take the vector of statements
9937 rather than just the current one. Remove excess parentheses.
9938 Restriction rejectinon of vect_constant_def and vect_external_def
9939 for variable-length vectors to boolean types, or types for which
9940 can_duplicate_and_interleave_p is false.
9941 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9942 (duplicate_and_interleave): New function.
9943 (vect_get_constant_vectors): Use gimple_build_vector for
9944 constant-length vectors and suitable variable-length constant
9945 vectors. Use duplicate_and_interleave for other variable-length
9946 vectors. Don't defer the update when inserting new statements.
9948 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9949 Alan Hayward <alan.hayward@arm.com>
9950 David Sherwood <david.sherwood@arm.com>
9952 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9953 min_profitable_iters doesn't go negative.
9955 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9956 Alan Hayward <alan.hayward@arm.com>
9957 David Sherwood <david.sherwood@arm.com>
9959 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9960 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9961 * optabs.def (vec_mask_load_lanes_optab): New optab.
9962 (vec_mask_store_lanes_optab): Likewise.
9963 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9964 (MASK_STORE_LANES): Likewise.
9965 * internal-fn.c (mask_load_lanes_direct): New macro.
9966 (mask_store_lanes_direct): Likewise.
9967 (expand_mask_load_optab_fn): Handle masked operations.
9968 (expand_mask_load_lanes_optab_fn): New macro.
9969 (expand_mask_store_optab_fn): Handle masked operations.
9970 (expand_mask_store_lanes_optab_fn): New macro.
9971 (direct_mask_load_lanes_optab_supported_p): Likewise.
9972 (direct_mask_store_lanes_optab_supported_p): Likewise.
9973 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9975 (vect_load_lanes_supported): Likewise.
9976 * tree-vect-data-refs.c (strip_conversion): New function.
9977 (can_group_stmts_p): Likewise.
9978 (vect_analyze_data_ref_accesses): Use it instead of checking
9979 for a pair of assignments.
9980 (vect_store_lanes_supported): Take a masked_p parameter.
9981 (vect_load_lanes_supported): Likewise.
9982 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9983 vect_store_lanes_supported and vect_load_lanes_supported.
9984 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9985 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9986 parameter. Don't allow gaps for masked accesses.
9987 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9988 and vect_load_lanes_supported.
9989 (get_load_store_type): Take a masked_p parameter and update
9990 call to get_group_load_store_type.
9991 (vectorizable_store): Update call to get_load_store_type.
9992 Handle IFN_MASK_STORE_LANES.
9993 (vectorizable_load): Update call to get_load_store_type.
9994 Handle IFN_MASK_LOAD_LANES.
9996 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9997 Alan Hayward <alan.hayward@arm.com>
9998 David Sherwood <david.sherwood@arm.com>
10000 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10002 * config/aarch64/aarch64-protos.h
10003 (aarch64_sve_struct_memory_operand_p): Declare.
10004 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10005 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10006 (VPRED, vpred): Handle SVE structure modes.
10007 * config/aarch64/constraints.md (Utx): New constraint.
10008 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10009 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10010 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10011 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10012 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10013 structure modes. Split into pieces after RA.
10014 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10015 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10017 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10018 SVE structure modes.
10019 (aarch64_classify_address): Likewise.
10020 (sizetochar): Move earlier in file.
10021 (aarch64_print_operand): Handle SVE register lists.
10022 (aarch64_array_mode): New function.
10023 (aarch64_sve_struct_memory_operand_p): Likewise.
10024 (TARGET_ARRAY_MODE): Redefine.
10026 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10027 Alan Hayward <alan.hayward@arm.com>
10028 David Sherwood <david.sherwood@arm.com>
10030 * target.def (array_mode): New target hook.
10031 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10032 * doc/tm.texi: Regenerate.
10033 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10034 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10035 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10036 targetm.array_mode.
10037 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10040 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10041 Alan Hayward <alan.hayward@arm.com>
10042 David Sherwood <david.sherwood@arm.com>
10044 * fold-const.c (fold_binary_loc): Check the argument types
10045 rather than the result type when testing for a vector operation.
10047 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10049 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10050 * doc/tm.texi: Regenerate.
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 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10057 (sve): Document new AArch64 extension.
10058 * doc/md.texi (w): Extend the description of the AArch64
10059 constraint to include SVE vectors.
10060 (Upl, Upa): Document new AArch64 predicate constraints.
10061 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10063 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10064 (msve-vector-bits=): New option.
10065 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10066 SVE when these are disabled.
10067 (sve): New extension.
10068 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10069 modes. Adjust their number of units based on aarch64_sve_vg.
10070 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10071 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10072 aarch64_addr_query_type.
10073 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10074 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10075 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10076 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10077 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10078 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10079 (aarch64_simd_imm_zero_p): Delete.
10080 (aarch64_check_zero_based_sve_index_immediate): Declare.
10081 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10082 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10083 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10084 (aarch64_sve_float_mul_immediate_p): Likewise.
10085 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10086 rather than an rtx.
10087 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10088 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10089 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10090 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10091 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10092 (aarch64_regmode_natural_size): Likewise.
10093 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10094 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10096 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10097 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10098 for VG and the SVE predicate registers.
10099 (V_ALIASES): Add a "z"-prefixed alias.
10100 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10101 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10102 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10103 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10104 (REG_CLASS_NAMES): Add entries for them.
10105 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10106 and the predicate registers.
10107 (aarch64_sve_vg): Declare.
10108 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10109 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10110 (REGMODE_NATURAL_SIZE): Define.
10111 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10113 * config/aarch64/aarch64.c: Include cfgrtl.h.
10114 (simd_immediate_info): Add a constructor for series vectors,
10115 and an associated step field.
10116 (aarch64_sve_vg): New variable.
10117 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10118 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10119 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10120 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10121 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10122 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10123 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10124 (aarch64_get_mask_mode): New functions.
10125 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10126 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10127 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10128 predicate modes and predicate registers. Explicitly restrict
10129 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10130 to store a vector mode if it is recognized by
10131 aarch64_classify_vector_mode.
10132 (aarch64_regmode_natural_size): New function.
10133 (aarch64_hard_regno_caller_save_mode): Return the original mode
10135 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10136 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10137 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10138 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10140 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10141 does not overlap dest if the function is frame-related. Handle
10143 (aarch64_split_add_offset): New function.
10144 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10145 them aarch64_add_offset.
10146 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10147 and update call to aarch64_sub_sp.
10148 (aarch64_add_cfa_expression): New function.
10149 (aarch64_expand_prologue): Pass extra temporary registers to the
10150 functions above. Handle the case in which we need to emit new
10151 DW_CFA_expressions for registers that were originally saved
10152 relative to the stack pointer, but now have to be expressed
10153 relative to the frame pointer.
10154 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10156 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10157 IP0 and IP1 values for SVE frames.
10158 (aarch64_expand_vec_series): New function.
10159 (aarch64_expand_sve_widened_duplicate): Likewise.
10160 (aarch64_expand_sve_const_vector): Likewise.
10161 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10162 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10163 into the register, rather than emitting a SET directly.
10164 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10165 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10166 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10167 (offset_9bit_signed_scaled_p): New functions.
10168 (aarch64_replicate_bitmask_imm): New function.
10169 (aarch64_bitmask_imm): Use it.
10170 (aarch64_cannot_force_const_mem): Reject expressions involving
10171 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10172 (aarch64_classify_index): Handle SVE indices, by requiring
10173 a plain register index with a scale that matches the element size.
10174 (aarch64_classify_address): Handle SVE addresses. Assert that
10175 the mode of the address is VOIDmode or an integer mode.
10176 Update call to aarch64_classify_symbol.
10177 (aarch64_classify_symbolic_expression): Update call to
10178 aarch64_classify_symbol.
10179 (aarch64_const_vec_all_in_range_p): New function.
10180 (aarch64_print_vector_float_operand): Likewise.
10181 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10182 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10183 and the FP immediates 1.0 and 0.5.
10184 (aarch64_print_address_internal): Handle SVE addresses.
10185 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10186 (aarch64_regno_regclass): Handle predicate registers.
10187 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10189 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10190 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10191 (aarch64_convert_sve_vector_bits): New function.
10192 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10193 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10194 rather than an rtx.
10195 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10196 Handle SVE vector and predicate modes. Accept VL-based constants
10197 that need only one temporary register, and VL offsets that require
10198 no temporary registers.
10199 (aarch64_conditional_register_usage): Mark the predicate registers
10200 as fixed if SVE isn't available.
10201 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10202 Return true for SVE vector and predicate modes.
10203 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10204 rather than an unsigned int. Handle SVE modes.
10205 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10207 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10209 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10210 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10211 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10212 (aarch64_sve_float_mul_immediate_p): New functions.
10213 (aarch64_sve_valid_immediate): New function.
10214 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10215 Explicitly reject structure modes. Check for INDEX constants.
10216 Handle PTRUE and PFALSE constants.
10217 (aarch64_check_zero_based_sve_index_immediate): New function.
10218 (aarch64_simd_imm_zero_p): Delete.
10219 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10220 vector modes. Accept constants in the range of CNT[BHWD].
10221 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10222 ask for an Advanced SIMD mode.
10223 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10224 (aarch64_simd_vector_alignment): Handle SVE predicates.
10225 (aarch64_vectorize_preferred_vector_alignment): New function.
10226 (aarch64_simd_vector_alignment_reachable): Use it instead of
10228 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10229 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10231 (MAX_VECT_LEN): Delete.
10232 (expand_vec_perm_d): Add a vec_flags field.
10233 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10234 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10235 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10237 (aarch64_evpc_rev): Rename to...
10238 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10239 (aarch64_evpc_rev_global): New function.
10240 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10241 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10243 (aarch64_evpc_sve_tbl): New function.
10244 (aarch64_expand_vec_perm_const_1): Update after rename of
10245 aarch64_evpc_rev. Handle SVE permutes too, trying
10246 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10247 than aarch64_evpc_tbl.
10248 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10249 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10250 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10251 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10252 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10253 (aarch64_expand_sve_vcond): New functions.
10254 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10255 of aarch64_vector_mode_p.
10256 (aarch64_dwarf_poly_indeterminate_value): New function.
10257 (aarch64_compute_pressure_classes): Likewise.
10258 (aarch64_can_change_mode_class): Likewise.
10259 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10260 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10261 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10262 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10263 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10264 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10265 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10266 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10268 (Dn, Dl, Dr): Accept const as well as const_vector.
10269 (Dz): Likewise. Compare against CONST0_RTX.
10270 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10271 of "vector" where appropriate.
10272 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10273 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10274 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10275 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10276 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10277 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10278 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10279 (v_int_equiv): Extend to SVE modes.
10280 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10282 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10283 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10284 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10285 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10286 (SVE_COND_FP_CMP): New int iterators.
10287 (perm_hilo): Handle the new unpack unspecs.
10288 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10290 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10291 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10292 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10293 (aarch64_equality_operator, aarch64_constant_vector_operand)
10294 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10295 (aarch64_sve_nonimmediate_operand): Likewise.
10296 (aarch64_sve_general_operand): Likewise.
10297 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10298 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10299 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10300 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10301 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10302 (aarch64_sve_float_arith_immediate): Likewise.
10303 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10304 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10305 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10306 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10307 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10308 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10309 (aarch64_sve_float_arith_operand): Likewise.
10310 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10311 (aarch64_sve_float_mul_operand): Likewise.
10312 (aarch64_sve_vec_perm_operand): Likewise.
10313 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10314 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10315 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10316 as well as const_vector.
10317 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10318 in file. Use CONST0_RTX and CONSTM1_RTX.
10319 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10320 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10321 Use aarch64_simd_imm_zero.
10322 * config/aarch64/aarch64-sve.md: New file.
10323 * config/aarch64/aarch64.md: Include it.
10324 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10325 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10326 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10327 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10328 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10329 (sve): New attribute.
10330 (enabled): Disable instructions with the sve attribute unless
10332 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10333 aarch64_expand_mov_immediate.
10334 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10335 CNT[BHSD] immediates.
10336 (movti): Split CONST_POLY_INT moves into two halves.
10337 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10338 Split additions that need a temporary here if the destination
10339 is the stack pointer.
10340 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10341 (*add<mode>3_poly_1): New instruction.
10342 (set_clobber_cc): New expander.
10344 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10346 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10347 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10348 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10349 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10350 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10351 Change innermode from fixed_mode_size to machine_mode.
10352 (simplify_subreg): Update call accordingly. Handle a constant-sized
10353 subreg of a variable-length CONST_VECTOR.
10355 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10356 Alan Hayward <alan.hayward@arm.com>
10357 David Sherwood <david.sherwood@arm.com>
10359 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10360 (add_offset_to_base): New function, split out from...
10361 (create_mem_ref): ...here. When handling a scale other than 1,
10362 check first whether the address is valid without the offset.
10363 Add it into the base if so, leaving the index and scale as-is.
10365 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10368 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10369 fold_for_warn before checking if arg2 is INTEGER_CST.
10371 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10373 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10374 (store_multiple_operation): Delete.
10375 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10376 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10377 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10378 guarded by TARGET_STRING.
10379 (rs6000_output_load_multiple): Delete.
10380 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10381 OPTION_MASK_STRING / TARGET_STRING handling.
10382 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10383 (const rs6000_opt_masks) <"string">: Change mask to 0.
10384 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10385 (MASK_STRING): Delete.
10386 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10388 (load_multiple): Delete.
10395 (store_multiple): Delete.
10402 (movmemsi_8reg): Delete.
10403 (corresponding unnamed define_insn): Delete.
10404 (movmemsi_6reg): Delete.
10405 (corresponding unnamed define_insn): Delete.
10406 (movmemsi_4reg): Delete.
10407 (corresponding unnamed define_insn): Delete.
10408 (movmemsi_2reg): Delete.
10409 (corresponding unnamed define_insn): Delete.
10410 (movmemsi_1reg): Delete.
10411 (corresponding unnamed define_insn): Delete.
10412 * config/rs6000/rs6000.opt (mno-string): New.
10413 (mstring): Replace by deprecation warning stub.
10414 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10416 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10418 * regrename.c (regrename_do_replace): If replacing the same
10419 reg multiple times, try to reuse last created gen_raw_REG.
10422 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10423 main to workaround a bug in GDB.
10425 2018-01-12 Tom de Vries <tom@codesourcery.com>
10428 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10430 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10432 PR rtl-optimization/80481
10433 * ira-color.c (get_cap_member): New function.
10434 (allocnos_conflict_by_live_ranges_p): Use it.
10435 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10436 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10438 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10441 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10442 (*saddl_se_1): Ditto.
10443 (*ssubsi_1): Ditto.
10444 (*ssubl_se_1): Ditto.
10446 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10448 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10449 rather than wi::to_widest for DR_INITs.
10450 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10451 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10452 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10454 (vect_analyze_group_access_1): Note that here.
10456 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10458 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10459 polynomial type sizes.
10461 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10463 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10464 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10465 (gimple_add_tmp_var): Likewise.
10467 2018-01-12 Martin Liska <mliska@suse.cz>
10469 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10470 (gimple_alloc_sizes): Likewise.
10471 (dump_gimple_statistics): Use PRIu64 in printf format.
10472 * gimple.h: Change uint64_t to int.
10474 2018-01-12 Martin Liska <mliska@suse.cz>
10476 * tree-core.h: Use uint64_t instead of int.
10477 * tree.c (tree_node_counts): Likewise.
10478 (tree_node_sizes): Likewise.
10479 (dump_tree_statistics): Use PRIu64 in printf format.
10481 2018-01-12 Martin Liska <mliska@suse.cz>
10483 * Makefile.in: As qsort_chk is implemented in vec.c, add
10484 vec.o to linkage of gencfn-macros.
10485 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10486 passing the info to record_node_allocation_statistics.
10487 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10489 * ggc-common.c (struct ggc_usage): Add operator== and use
10490 it in operator< and compare function.
10491 * mem-stats.h (struct mem_usage): Likewise.
10492 * vec.c (struct vec_usage): Remove operator< and compare
10493 function. Can be simply inherited.
10495 2018-01-12 Martin Jambor <mjambor@suse.cz>
10498 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10499 * tree-ssa-math-opts.c: Include domwalk.h.
10500 (convert_mult_to_fma_1): New function.
10501 (fma_transformation_info): New type.
10502 (fma_deferring_state): Likewise.
10503 (cancel_fma_deferring): New function.
10504 (result_of_phi): Likewise.
10505 (last_fma_candidate_feeds_initial_phi): Likewise.
10506 (convert_mult_to_fma): Added deferring logic, split actual
10507 transformation to convert_mult_to_fma_1.
10508 (math_opts_dom_walker): New type.
10509 (math_opts_dom_walker::after_dom_children): New method, body moved
10510 here from pass_optimize_widening_mul::execute, added deferring logic
10512 (pass_optimize_widening_mul::execute): Moved most of code to
10513 math_opts_dom_walker::after_dom_children.
10514 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10515 * config/i386/i386.c (ix86_option_override_internal): Added
10516 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10518 2018-01-12 Richard Biener <rguenther@suse.de>
10521 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10522 inline instance vars.
10524 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10527 * config/rx/rx.c (rx_is_restricted_memory_address):
10528 Handle SUBREG case.
10530 2018-01-12 Richard Biener <rguenther@suse.de>
10532 PR tree-optimization/80846
10533 * target.def (split_reduction): New target hook.
10534 * targhooks.c (default_split_reduction): New function.
10535 * targhooks.h (default_split_reduction): Declare.
10536 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10537 target requests first reduce vectors by combining low and high
10539 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10540 (get_vectype_for_scalar_type_and_size): Export.
10541 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10542 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10543 * doc/tm.texi: Regenerate.
10544 * config/i386/i386.c (ix86_split_reduction): Implement
10545 TARGET_VECTORIZE_SPLIT_REDUCTION.
10547 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10550 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10551 in PIC mode except for TARGET_VXWORKS_RTP.
10552 * config/sparc/sparc.c: Include cfgrtl.h.
10553 (TARGET_INIT_PIC_REG): Define.
10554 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10555 (sparc_pic_register_p): New predicate.
10556 (sparc_legitimate_address_p): Use it.
10557 (sparc_legitimize_pic_address): Likewise.
10558 (sparc_delegitimize_address): Likewise.
10559 (sparc_mode_dependent_address_p): Likewise.
10560 (gen_load_pcrel_sym): Remove 4th parameter.
10561 (load_got_register): Adjust call to above. Remove obsolete stuff.
10562 (sparc_expand_prologue): Do not call load_got_register here.
10563 (sparc_flat_expand_prologue): Likewise.
10564 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10565 (sparc_use_pseudo_pic_reg): New function.
10566 (sparc_init_pic_reg): Likewise.
10567 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10568 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10570 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10572 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10573 Add item for branch_cost.
10575 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10577 PR rtl-optimization/83565
10578 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10579 not extend the result to a larger mode for rotate operations.
10580 (num_sign_bit_copies1): Likewise.
10582 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10585 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10587 Use values-Xc.o for -pedantic.
10588 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10590 2018-01-12 Martin Liska <mliska@suse.cz>
10593 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10595 (possible_polymorphic_call_targets): Use it.
10596 (ipa_devirt): Likewise.
10598 2018-01-12 Martin Liska <mliska@suse.cz>
10600 * profile-count.h (enum profile_quality): Use 0 as invalid
10601 enum value of profile_quality.
10603 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10605 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10606 -mext-string options.
10608 2018-01-12 Richard Biener <rguenther@suse.de>
10610 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10611 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10612 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10614 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10616 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10618 * configure.ac (--with-long-double-format): Add support for the
10619 configuration option to change the default long double format on
10621 * config.gcc (powerpc*-linux*-*): Likewise.
10622 * configure: Regenerate.
10623 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10624 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10625 used without modification.
10627 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10629 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10630 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10631 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10632 MISC_BUILTIN_SPEC_BARRIER.
10633 (rs6000_init_builtins): Likewise.
10634 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10636 (speculation_barrier): New define_insn.
10637 * doc/extend.texi: Document __builtin_speculation_barrier.
10639 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10642 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10643 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10644 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10646 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10647 integral modes instead of "ss" and "sd".
10648 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10649 vectors with 32-bit and 64-bit elements.
10650 (vecdupssescalarmodesuffix): New mode attribute.
10651 (vec_dup<mode>): Use it.
10653 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10656 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10657 frame if argument is passed on stack.
10659 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10662 * ree.c (combine_reaching_defs): Optimize also
10663 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10664 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10666 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10668 PR middle-end/83189
10669 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10671 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10673 PR middle-end/83718
10674 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10675 after they are computed.
10677 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10679 PR tree-optimization/83695
10680 * gimple-loop-linterchange.cc
10681 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10682 reset cached scev information after interchange.
10683 (pass_linterchange::execute): Remove call to scev_reset_htab.
10685 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10687 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10688 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10689 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10690 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10691 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10692 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10693 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10694 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10695 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10696 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10697 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10698 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10699 (V_lane_reg): Likewise.
10700 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10702 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10703 (vfmal_lane_low<mode>_intrinsic,
10704 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10705 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10706 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10707 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10708 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10709 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10711 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10713 * config/arm/arm-cpus.in (fp16fml): New feature.
10714 (ALL_SIMD): Add fp16fml.
10715 (armv8.2-a): Add fp16fml as an option.
10716 (armv8.3-a): Likewise.
10717 (armv8.4-a): Add fp16fml as part of fp16.
10718 * config/arm/arm.h (TARGET_FP16FML): Define.
10719 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10721 * config/arm/arm-modes.def (V2HF): Define.
10722 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10723 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10724 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10725 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10726 vfmsl_low, vfmsl_high): New set of builtins.
10727 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10728 (vfml_op): New code attribute.
10729 (VFMLHALVES): New int iterator.
10730 (VFML, VFMLSEL): New mode attributes.
10731 (V_reg): Define mapping for V2HF.
10732 (V_hi, V_lo): New mode attributes.
10733 (VF_constraint): Likewise.
10734 (vfml_half, vfml_half_selector): New int attributes.
10735 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10737 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10738 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10740 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10741 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10742 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10743 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10745 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10746 Document new effective target and option set.
10748 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10750 * config/arm/arm-cpus.in (armv8_4): New feature.
10751 (ARMv8_4a): New fgroup.
10752 (armv8.4-a): New arch.
10753 * config/arm/arm-tables.opt: Regenerate.
10754 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10755 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10756 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10757 Add matching rules for -march=armv8.4-a and extensions.
10758 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10760 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10763 * config/rx/rx.md (BW): New mode attribute.
10764 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10766 2018-01-11 Richard Biener <rguenther@suse.de>
10768 PR tree-optimization/83435
10769 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10770 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10771 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10773 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10774 Alan Hayward <alan.hayward@arm.com>
10775 David Sherwood <david.sherwood@arm.com>
10777 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10779 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10780 (aarch64_print_address_internal): Use it to check for a zero offset.
10782 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10783 Alan Hayward <alan.hayward@arm.com>
10784 David Sherwood <david.sherwood@arm.com>
10786 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10787 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10788 Return a poly_int64 rather than a HOST_WIDE_INT.
10789 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10790 rather than a HOST_WIDE_INT.
10791 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10792 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10793 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10794 final_offset from HOST_WIDE_INT to poly_int64.
10795 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10796 to_constant when getting the number of units in an Advanced SIMD
10798 (aarch64_builtin_vectorized_function): Check for a constant number
10800 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10802 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10803 attribute instead of GET_MODE_NUNITS.
10804 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10805 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10806 GET_MODE_SIZE for fixed-size registers.
10807 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10808 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10809 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10810 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10811 (aarch64_print_operand, aarch64_print_address_internal)
10812 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10813 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10814 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10815 Handle polynomial GET_MODE_SIZE.
10816 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10817 wider than SImode without modification.
10818 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10819 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10820 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10821 passing and returning SVE modes.
10822 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10823 rather than GEN_INT.
10824 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10825 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10826 (aarch64_allocate_and_probe_stack_space): Likewise.
10827 (aarch64_layout_frame): Cope with polynomial offsets.
10828 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10829 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10830 polynomial offsets.
10831 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10832 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10833 poly_int64 rather than a HOST_WIDE_INT.
10834 (aarch64_get_separate_components, aarch64_process_components)
10835 (aarch64_expand_prologue, aarch64_expand_epilogue)
10836 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10837 (aarch64_anchor_offset): New function, split out from...
10838 (aarch64_legitimize_address): ...here.
10839 (aarch64_builtin_vectorization_cost): Handle polynomial
10840 TYPE_VECTOR_SUBPARTS.
10841 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10843 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10844 number of elements from the PARALLEL rather than the mode.
10845 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10846 rather than GET_MODE_BITSIZE.
10847 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10848 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10849 (aarch64_expand_vec_perm_const_1): Handle polynomial
10850 d->perm.length () and d->perm elements.
10851 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10852 Apply to_constant to d->perm elements.
10853 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10854 polynomial CONST_VECTOR_NUNITS.
10855 (aarch64_move_pointer): Take amount as a poly_int64 rather
10857 (aarch64_progress_pointer): Avoid temporary variable.
10858 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10859 the mode attribute instead of GET_MODE.
10861 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10862 Alan Hayward <alan.hayward@arm.com>
10863 David Sherwood <david.sherwood@arm.com>
10865 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10866 x exists before using it.
10867 (aarch64_add_constant_internal): Rename to...
10868 (aarch64_add_offset_1): ...this. Replace regnum with separate
10869 src and dest rtxes. Handle the case in which they're different,
10870 including when the offset is zero. Replace scratchreg with an rtx.
10871 Use 2 additions if there is no spare register into which we can
10872 move a 16-bit constant.
10873 (aarch64_add_constant): Delete.
10874 (aarch64_add_offset): Replace reg with separate src and dest
10875 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10876 Use aarch64_add_offset_1.
10877 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10878 an rtx rather than an int. Take the delta as a poly_int64
10879 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10880 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10881 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10882 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10883 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10884 and aarch64_add_sp.
10885 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10886 aarch64_add_constant.
10888 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10890 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10891 Use scalar_float_mode.
10893 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10895 * config/aarch64/aarch64-simd.md
10896 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10897 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10898 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10899 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10900 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10901 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10902 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10903 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10904 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10905 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10907 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10910 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10911 targ_options->x_arm_arch_string is non NULL.
10913 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10915 * config/aarch64/aarch64.h
10916 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10918 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10921 * expmed.c (emit_store_flag_force): Swap if const op0
10922 and change VOIDmode to mode of op0.
10924 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10926 PR rtl-optimization/83761
10927 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10928 than bytes to mode_for_size.
10930 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10932 PR middle-end/83189
10933 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10934 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10937 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10939 PR middle-end/83575
10940 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10941 when in layout mode.
10942 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10943 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10946 2018-01-10 Michael Collison <michael.collison@arm.com>
10948 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10949 * config/aarch64/aarch64-option-extension.def: Add
10950 AARCH64_OPT_EXTENSION of 'fp16fml'.
10951 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10952 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10953 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10954 * config/aarch64/constraints.md (Ui7): New constraint.
10955 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10956 (VFMLA_SEL_W): Ditto.
10959 (VFMLA16_LOW): New int iterator.
10960 (VFMLA16_HIGH): Ditto.
10961 (UNSPEC_FMLAL): New unspec.
10962 (UNSPEC_FMLSL): Ditto.
10963 (UNSPEC_FMLAL2): Ditto.
10964 (UNSPEC_FMLSL2): Ditto.
10965 (f16mac): New code attribute.
10966 * config/aarch64/aarch64-simd-builtins.def
10967 (aarch64_fmlal_lowv2sf): Ditto.
10968 (aarch64_fmlsl_lowv2sf): Ditto.
10969 (aarch64_fmlalq_lowv4sf): Ditto.
10970 (aarch64_fmlslq_lowv4sf): Ditto.
10971 (aarch64_fmlal_highv2sf): Ditto.
10972 (aarch64_fmlsl_highv2sf): Ditto.
10973 (aarch64_fmlalq_highv4sf): Ditto.
10974 (aarch64_fmlslq_highv4sf): Ditto.
10975 (aarch64_fmlal_lane_lowv2sf): Ditto.
10976 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10977 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10978 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10979 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10980 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10981 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10982 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10983 (aarch64_fmlal_lane_highv2sf): Ditto.
10984 (aarch64_fmlsl_lane_highv2sf): Ditto.
10985 (aarch64_fmlal_laneq_highv2sf): Ditto.
10986 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10987 (aarch64_fmlalq_lane_highv4sf): Ditto.
10988 (aarch64_fmlsl_lane_highv4sf): Ditto.
10989 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10990 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10991 * config/aarch64/aarch64-simd.md:
10992 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10993 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10994 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10995 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10996 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10997 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10998 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10999 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11000 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11001 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11002 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11003 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11004 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11005 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11006 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11007 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11008 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11009 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11010 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11011 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11012 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11013 (vfmlsl_low_u32): Ditto.
11014 (vfmlalq_low_u32): Ditto.
11015 (vfmlslq_low_u32): Ditto.
11016 (vfmlal_high_u32): Ditto.
11017 (vfmlsl_high_u32): Ditto.
11018 (vfmlalq_high_u32): Ditto.
11019 (vfmlslq_high_u32): Ditto.
11020 (vfmlal_lane_low_u32): Ditto.
11021 (vfmlsl_lane_low_u32): Ditto.
11022 (vfmlal_laneq_low_u32): Ditto.
11023 (vfmlsl_laneq_low_u32): Ditto.
11024 (vfmlalq_lane_low_u32): Ditto.
11025 (vfmlslq_lane_low_u32): Ditto.
11026 (vfmlalq_laneq_low_u32): Ditto.
11027 (vfmlslq_laneq_low_u32): Ditto.
11028 (vfmlal_lane_high_u32): Ditto.
11029 (vfmlsl_lane_high_u32): Ditto.
11030 (vfmlal_laneq_high_u32): Ditto.
11031 (vfmlsl_laneq_high_u32): Ditto.
11032 (vfmlalq_lane_high_u32): Ditto.
11033 (vfmlslq_lane_high_u32): Ditto.
11034 (vfmlalq_laneq_high_u32): Ditto.
11035 (vfmlslq_laneq_high_u32): Ditto.
11036 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11037 (AARCH64_FL_FOR_ARCH8_4): New.
11038 (AARCH64_ISA_F16FML): New ISA flag.
11039 (TARGET_F16FML): New feature flag for fp16fml.
11040 (doc/invoke.texi): Document new fp16fml option.
11042 2018-01-10 Michael Collison <michael.collison@arm.com>
11044 * config/aarch64/aarch64-builtins.c:
11045 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11046 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11047 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11048 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11049 (AARCH64_ISA_SHA3): New ISA flag.
11050 (TARGET_SHA3): New feature flag for sha3.
11051 * config/aarch64/iterators.md (sha512_op): New int attribute.
11052 (CRYPTO_SHA512): New int iterator.
11053 (UNSPEC_SHA512H): New unspec.
11054 (UNSPEC_SHA512H2): Ditto.
11055 (UNSPEC_SHA512SU0): Ditto.
11056 (UNSPEC_SHA512SU1): Ditto.
11057 * config/aarch64/aarch64-simd-builtins.def
11058 (aarch64_crypto_sha512hqv2di): New builtin.
11059 (aarch64_crypto_sha512h2qv2di): Ditto.
11060 (aarch64_crypto_sha512su0qv2di): Ditto.
11061 (aarch64_crypto_sha512su1qv2di): Ditto.
11062 (aarch64_eor3qv8hi): Ditto.
11063 (aarch64_rax1qv2di): Ditto.
11064 (aarch64_xarqv2di): Ditto.
11065 (aarch64_bcaxqv8hi): Ditto.
11066 * config/aarch64/aarch64-simd.md:
11067 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11068 (aarch64_crypto_sha512su0qv2di): Ditto.
11069 (aarch64_crypto_sha512su1qv2di): Ditto.
11070 (aarch64_eor3qv8hi): Ditto.
11071 (aarch64_rax1qv2di): Ditto.
11072 (aarch64_xarqv2di): Ditto.
11073 (aarch64_bcaxqv8hi): Ditto.
11074 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11075 (vsha512h2q_u64): Ditto.
11076 (vsha512su0q_u64): Ditto.
11077 (vsha512su1q_u64): Ditto.
11078 (veor3q_u16): Ditto.
11079 (vrax1q_u64): Ditto.
11080 (vxarq_u64): Ditto.
11081 (vbcaxq_u16): Ditto.
11082 * config/arm/types.md (crypto_sha512): New type attribute.
11083 (crypto_sha3): Ditto.
11084 (doc/invoke.texi): Document new sha3 option.
11086 2018-01-10 Michael Collison <michael.collison@arm.com>
11088 * config/aarch64/aarch64-builtins.c:
11089 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11090 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11091 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11092 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11093 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11094 (AARCH64_ISA_SM4): New ISA flag.
11095 (TARGET_SM4): New feature flag for sm4.
11096 * config/aarch64/aarch64-simd-builtins.def
11097 (aarch64_sm3ss1qv4si): Ditto.
11098 (aarch64_sm3tt1aq4si): Ditto.
11099 (aarch64_sm3tt1bq4si): Ditto.
11100 (aarch64_sm3tt2aq4si): Ditto.
11101 (aarch64_sm3tt2bq4si): Ditto.
11102 (aarch64_sm3partw1qv4si): Ditto.
11103 (aarch64_sm3partw2qv4si): Ditto.
11104 (aarch64_sm4eqv4si): Ditto.
11105 (aarch64_sm4ekeyqv4si): Ditto.
11106 * config/aarch64/aarch64-simd.md:
11107 (aarch64_sm3ss1qv4si): Ditto.
11108 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11109 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11110 (aarch64_sm4eqv4si): Ditto.
11111 (aarch64_sm4ekeyqv4si): Ditto.
11112 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11113 (sm3part_op): Ditto.
11114 (CRYPTO_SM3TT): Ditto.
11115 (CRYPTO_SM3PART): Ditto.
11116 (UNSPEC_SM3SS1): New unspec.
11117 (UNSPEC_SM3TT1A): Ditto.
11118 (UNSPEC_SM3TT1B): Ditto.
11119 (UNSPEC_SM3TT2A): Ditto.
11120 (UNSPEC_SM3TT2B): Ditto.
11121 (UNSPEC_SM3PARTW1): Ditto.
11122 (UNSPEC_SM3PARTW2): Ditto.
11123 (UNSPEC_SM4E): Ditto.
11124 (UNSPEC_SM4EKEY): Ditto.
11125 * config/aarch64/constraints.md (Ui2): New constraint.
11126 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11127 * config/arm/types.md (crypto_sm3): New type attribute.
11128 (crypto_sm4): Ditto.
11129 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11130 (vsm3tt1aq_u32): Ditto.
11131 (vsm3tt1bq_u32): Ditto.
11132 (vsm3tt2aq_u32): Ditto.
11133 (vsm3tt2bq_u32): Ditto.
11134 (vsm3partw1q_u32): Ditto.
11135 (vsm3partw2q_u32): Ditto.
11136 (vsm4eq_u32): Ditto.
11137 (vsm4ekeyq_u32): Ditto.
11138 (doc/invoke.texi): Document new sm4 option.
11140 2018-01-10 Michael Collison <michael.collison@arm.com>
11142 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11143 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11144 (AARCH64_FL_FOR_ARCH8_4): New.
11145 (AARCH64_FL_V8_4): New flag.
11146 (doc/invoke.texi): Document new armv8.4-a option.
11148 2018-01-10 Michael Collison <michael.collison@arm.com>
11150 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11151 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11152 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11153 * config/aarch64/aarch64-option-extension.def: Add
11154 AARCH64_OPT_EXTENSION of 'sha2'.
11155 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11156 (crypto): Disable sha2 and aes if crypto disabled.
11157 (crypto): Enable aes and sha2 if enabled.
11158 (simd): Disable sha2 and aes if simd disabled.
11159 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11161 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11162 (TARGET_SHA2): New feature flag for sha2.
11163 (TARGET_AES): New feature flag for aes.
11164 * config/aarch64/aarch64-simd.md:
11165 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11166 conditional on TARGET_AES.
11167 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11168 (aarch64_crypto_sha1hsi): Make pattern conditional
11170 (aarch64_crypto_sha1hv4si): Ditto.
11171 (aarch64_be_crypto_sha1hv4si): Ditto.
11172 (aarch64_crypto_sha1su1v4si): Ditto.
11173 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11174 (aarch64_crypto_sha1su0v4si): Ditto.
11175 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11176 (aarch64_crypto_sha256su0v4si): Ditto.
11177 (aarch64_crypto_sha256su1v4si): Ditto.
11178 (doc/invoke.texi): Document new aes and sha2 options.
11180 2018-01-10 Martin Sebor <msebor@redhat.com>
11182 PR tree-optimization/83781
11183 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11186 2018-01-11 Martin Sebor <msebor@gmail.com>
11187 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11189 PR tree-optimization/83501
11190 PR tree-optimization/81703
11192 * tree-ssa-strlen.c (get_string_cst): Rename...
11193 (get_string_len): ...to this. Handle global constants.
11194 (handle_char_store): Adjust.
11196 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11197 Jim Wilson <jimw@sifive.com>
11199 * config/riscv/riscv-protos.h (riscv_output_return): New.
11200 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11201 (riscv_attribute_table, riscv_output_return),
11202 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11203 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11204 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11206 (riscv_expand_prologue): Add early return for naked function.
11207 (riscv_expand_epilogue): Likewise.
11208 (riscv_function_ok_for_sibcall): Return false for naked function.
11209 (riscv_set_current_function): New.
11210 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11211 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11212 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11213 * doc/extend.texi (RISC-V Function Attributes): New.
11215 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11217 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11218 check for 128-bit long double before checking TCmode.
11219 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11220 128-bit long doubles before checking TFmode or TCmode.
11221 (FLOAT128_IBM_P): Likewise.
11223 2018-01-10 Martin Sebor <msebor@redhat.com>
11225 PR tree-optimization/83671
11226 * builtins.c (c_strlen): Unconditionally return zero for the empty
11228 Use -Warray-bounds for warnings.
11229 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11230 for non-constant array indices with COMPONENT_REF, arrays of
11231 arrays, and pointers to arrays.
11232 (gimple_fold_builtin_strlen): Determine and set length range for
11233 non-constant character arrays.
11235 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11237 PR middle-end/81897
11238 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11241 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11243 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11245 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11248 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11249 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11250 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11251 indexed_or_indirect_operand predicate.
11252 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11253 (*vsx_le_perm_load_v8hi): Likewise.
11254 (*vsx_le_perm_load_v16qi): Likewise.
11255 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11256 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11257 (*vsx_le_perm_store_v8hi): Likewise.
11258 (*vsx_le_perm_store_v16qi): Likewise.
11259 (eight unnamed splitters): Likewise.
11261 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11263 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11264 * config/rs6000/emmintrin.h: Likewise.
11265 * config/rs6000/mmintrin.h: Likewise.
11266 * config/rs6000/xmmintrin.h: Likewise.
11268 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11271 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11273 * tree.c (tree_nop_conversion): Return true for location wrapper
11275 (maybe_wrap_with_location): New function.
11276 (selftest::check_strip_nops): New function.
11277 (selftest::test_location_wrappers): New function.
11278 (selftest::tree_c_tests): Call it.
11279 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11280 (maybe_wrap_with_location): New decl.
11281 (EXPR_LOCATION_WRAPPER_P): New macro.
11282 (location_wrapper_p): New inline function.
11283 (tree_strip_any_location_wrapper): New inline function.
11285 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11288 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11289 stack_realign_offset for the largest alignment of stack slot
11291 (ix86_find_max_used_stack_alignment): New function.
11292 (ix86_finalize_stack_frame_flags): Use it. Set
11293 max_used_stack_alignment if we don't realign stack.
11294 * config/i386/i386.h (machine_function): Add
11295 max_used_stack_alignment.
11297 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11299 * config/arm/arm.opt (-mbranch-cost): New option.
11300 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11303 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11306 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11307 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11309 2018-01-10 Richard Biener <rguenther@suse.de>
11312 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11313 early out so it also covers the case where we have a non-NULL
11316 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11318 PR tree-optimization/83753
11319 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11320 for non-strided grouped accesses if the number of elements is 1.
11322 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11325 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11326 * i386.h (TARGET_USE_GATHER): Define.
11327 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11329 2018-01-10 Martin Liska <mliska@suse.cz>
11332 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11333 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11335 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11336 CLEANUP_NO_PARTITIONING is not set.
11338 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11340 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11341 for vectors, as a partial revert of r254296.
11342 * rtl.h (const_vec_p): Delete.
11343 (const_vec_duplicate_p): Don't test for vector CONSTs.
11344 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11345 * expmed.c (make_tree): Likewise.
11348 * common.md (E, F): Use CONSTANT_P instead of checking for
11350 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11351 checking for CONST_VECTOR.
11353 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11355 PR middle-end/83575
11356 * predict.c (force_edge_cold): Handle in more sane way edges
11357 with no prediction.
11359 2018-01-09 Carl Love <cel@us.ibm.com>
11361 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11363 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11364 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11365 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11366 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11367 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11368 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11369 * config/rs6000/rs6000-protos.h: Add extern defition for
11370 rs6000_generate_float2_double_code.
11371 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11373 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11374 (float2_v2df): Add define_expand.
11376 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11379 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11380 op_mode in the force_to_mode call.
11382 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11384 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11385 instead of checking each element individually.
11386 (aarch64_evpc_uzp): Likewise.
11387 (aarch64_evpc_zip): Likewise.
11388 (aarch64_evpc_ext): Likewise.
11389 (aarch64_evpc_rev): Likewise.
11390 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11391 instead of checking each element individually. Return true without
11393 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11394 whether all selected elements come from the same input, instead of
11395 checking each element individually. Remove calls to gen_rtx_REG,
11396 start_sequence and end_sequence and instead assert that no rtl is
11399 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11401 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11402 order of HIGH and CONST checks.
11404 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11406 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11407 if the destination isn't an SSA_NAME.
11409 2018-01-09 Richard Biener <rguenther@suse.de>
11411 PR tree-optimization/83668
11412 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11414 (canonicalize_loop_form): ... here, renamed from ...
11415 (canonicalize_loop_closed_ssa_form): ... this and amended to
11416 swap successor edges for loop exit blocks to make us use
11417 the RPO order we need for initial schedule generation.
11419 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11421 PR tree-optimization/64811
11422 * match.pd: When optimizing comparisons with Inf, avoid
11423 introducing or losing exceptions from comparisons with NaN.
11425 2018-01-09 Martin Liska <mliska@suse.cz>
11428 * asan.c (shadow_mem_size): Add gcc_assert.
11430 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11432 Don't save registers in main().
11435 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11436 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11437 * config/avr/avr.c (avr_set_current_function): Don't error if
11438 naked, OS_task or OS_main are specified at the same time.
11439 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11441 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11443 * common/config/avr/avr-common.c (avr_option_optimization_table):
11444 Switch on -mmain-is-OS_task for optimizing compilations.
11446 2018-01-09 Richard Biener <rguenther@suse.de>
11448 PR tree-optimization/83572
11449 * graphite.c: Include cfganal.h.
11450 (graphite_transform_loops): Connect infinite loops to exit
11451 and remove fake edges at the end.
11453 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11455 * ipa-inline.c (edge_badness): Revert accidental checkin.
11457 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11460 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11461 symbols; not inline clones.
11463 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11466 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11467 hard registers. Formatting fixes.
11469 PR preprocessor/83722
11470 * gcc.c (try_generate_repro): Pass
11471 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11472 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11475 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11476 Kito Cheng <kito.cheng@gmail.com>
11478 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11479 (riscv_leaf_function_p): Delete.
11480 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11482 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11484 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11486 (do_ifelse): New function.
11487 (do_isel): New function.
11488 (do_sub3): New function.
11489 (do_add3): New function.
11490 (do_load_mask_compare): New function.
11491 (do_overlap_load_compare): New function.
11492 (expand_compare_loop): New function.
11493 (expand_block_compare): Call expand_compare_loop() when appropriate.
11494 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11495 option description.
11496 (-mblock-compare-inline-loop-limit): New option.
11498 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11501 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11502 Reverse order of second and third operands in first alternative.
11503 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11504 of first and second elements in UNSPEC_VPERMR vector.
11505 (altivec_expand_vec_perm_le): Likewise.
11507 2018-01-08 Jeff Law <law@redhat.com>
11509 PR rtl-optimizatin/81308
11510 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11511 (process_switch): If group_case_labels makes a change, then set
11513 (pass_convert_switch::execute): If a switch is converted, then
11514 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11516 PR rtl-optimization/81308
11517 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11520 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11522 PR target/83663 - Revert r255946
11523 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11524 generation for cases where splatting a value is not useful.
11525 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11526 across a vec_duplicate and a paradoxical subreg forming a vector
11527 mode to a vec_concat.
11529 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11531 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11532 -march=armv8.3-a variants.
11533 * config/arm/t-multilib: Likewise.
11534 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11536 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11538 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11540 (cceq_ior_compare_complement): Give it a name so I can use it, and
11541 change boolean_or_operator predicate to boolean_operator so it can
11542 be used to generate a crand.
11543 (eqne): New code iterator.
11544 (bd/bd_neg): New code_attrs.
11545 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11546 a single define_insn.
11547 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11548 decrement (bdnzt/bdnzf/bdzt/bdzf).
11549 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11550 with the new names of the branch decrement patterns, and added the
11551 names of the branch decrement conditional patterns.
11553 2018-01-08 Richard Biener <rguenther@suse.de>
11555 PR tree-optimization/83563
11556 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11559 2018-01-08 Richard Biener <rguenther@suse.de>
11561 PR middle-end/83713
11562 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11564 2018-01-08 Richard Biener <rguenther@suse.de>
11566 PR tree-optimization/83685
11567 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11568 references to abnormals.
11570 2018-01-08 Richard Biener <rguenther@suse.de>
11573 * dwarf2out.c (output_indirect_strings): Handle empty
11574 skeleton_debug_str_hash.
11575 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11577 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11579 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11580 (emit_store_direct): Likewise.
11581 (arc_trampoline_adjust_address): Likewise.
11582 (arc_asm_trampoline_template): New function.
11583 (arc_initialize_trampoline): Use asm_trampoline_template.
11584 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11585 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11586 * config/arc/arc.md (flush_icache): Delete pattern.
11588 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11590 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11591 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11594 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11597 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11598 by not USED_FOR_TARGET.
11599 (make_pass_resolve_sw_modes): Likewise.
11601 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11603 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11606 2018-01-08 Richard Biener <rguenther@suse.de>
11608 PR middle-end/83580
11609 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11611 2018-01-08 Richard Biener <rguenther@suse.de>
11613 PR middle-end/83517
11614 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11616 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11618 PR middle-end/81897
11619 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11620 basic blocks with a small number of successors.
11621 (convert_control_dep_chain_into_preds): Improve handling of
11623 (dump_predicates): Split apart into...
11624 (dump_pred_chain): ...here...
11625 (dump_pred_info): ...and here.
11626 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11627 (can_chain_union_be_invalidated_p): Improve check for invalidation
11629 (uninit_uses_cannot_happen): Avoid unnecessary if
11630 convert_control_dep_chain_into_preds yielded nothing.
11632 2018-01-06 Martin Sebor <msebor@redhat.com>
11634 PR tree-optimization/83640
11635 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11636 subtracting negative offset from size.
11637 (builtin_access::overlap): Adjust offset bounds of the access to fall
11638 within the size of the object if possible.
11640 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11642 PR rtl-optimization/83699
11643 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11644 extract_bit_field_as_subreg to cases in which the extracted
11645 value is also a vector.
11647 * lra-constraints.c (process_alt_operands): Test for the equivalence
11648 substitutions when detecting a possible reload cycle.
11650 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11653 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11654 by default if flag_selective_schedling{,2}. Formatting fixes.
11656 PR rtl-optimization/83682
11657 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11658 if it has non-VECTOR_MODE element mode.
11659 (vec_duplicate_p): Likewise.
11661 PR middle-end/83694
11662 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11663 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11665 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11668 * config/i386/i386-builtin.def
11669 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11670 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11671 Require also OPTION_MASK_ISA_AVX512F in addition to
11672 OPTION_MASK_ISA_GFNI.
11673 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11674 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11675 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11676 to OPTION_MASK_ISA_GFNI.
11677 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11678 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11679 OPTION_MASK_ISA_AVX512BW.
11680 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11681 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11682 addition to OPTION_MASK_ISA_GFNI.
11683 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11684 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11685 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11686 to OPTION_MASK_ISA_GFNI.
11687 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11688 a requirement for all ISAs rather than any of them with a few
11690 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11692 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11693 bitmasks to be enabled with 3 exceptions, instead of requiring any
11694 enabled ISA with lots of exceptions.
11695 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11696 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11697 Change avx512bw in isa attribute to avx512f.
11698 * config/i386/sgxintrin.h: Add license boilerplate.
11699 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11700 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11701 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11702 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11704 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11705 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11706 temporarily sse2 rather than sse if not enabled already.
11709 * config/i386/sse.md (VI248_VLBW): Rename to ...
11710 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11711 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11712 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11713 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11714 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11715 mode iterator instead of VI248_VLBW.
11717 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11719 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11720 (record_modified): Skip clobbers; add debug output.
11721 (param_change_prob): Use sreal frequencies.
11723 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11725 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11726 punt for user-aligned variables.
11728 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11730 * tree-chrec.c (chrec_contains_symbols): Return true for
11733 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11736 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11737 of (x|y) == x for BICS pattern.
11739 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11741 PR tree-optimization/83605
11742 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11743 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11746 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11748 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11749 * config/epiphany/rtems.h: New file.
11751 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11752 Uros Bizjak <ubizjak@gmail.com>
11755 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11756 QIreg_operand instead of register_operand predicate.
11757 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11758 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11759 comments instead of -fmitigate[-_]rop.
11761 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11764 * cgraphunit.c (symbol_table::compile): Switch to text_section
11765 before calling assembly_start debug hook.
11766 * run-rtl-passes.c (run_rtl_passes): Likewise.
11769 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11771 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11772 range_int_cst_p rather than !symbolic_range_p before calling
11773 extract_range_from_multiplicative_op_1.
11775 2018-01-04 Jeff Law <law@redhat.com>
11777 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11778 redundant test in assertion.
11780 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11782 * doc/rtl.texi: Document machine_mode wrapper classes.
11784 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11786 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11787 using tree_to_uhwi.
11789 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11791 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11792 the VEC_PERM_EXPR fold to fail.
11794 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11797 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11798 to switched_sections.
11800 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11803 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11804 test for d.testing.
11806 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11809 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11810 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11812 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11815 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11816 is BLKmode and bitpos not zero or mode change is needed.
11818 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11821 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11824 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11827 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11828 instead of MULT rtx. Update all corresponding splitters.
11829 (*saddl_se): Ditto.
11830 (*ssub<modesuffix>): Ditto.
11831 (*ssubl_se): Ditto.
11832 (*cmp_sadd_di): Update split patterns.
11833 (*cmp_sadd_si): Ditto.
11834 (*cmp_sadd_sidi): Ditto.
11835 (*cmp_ssub_di): Ditto.
11836 (*cmp_ssub_si): Ditto.
11837 (*cmp_ssub_sidi): Ditto.
11838 * config/alpha/predicates.md (const23_operand): New predicate.
11839 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11840 Look for ASHIFT, not MULT inner operand.
11841 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11843 2018-01-04 Martin Liska <mliska@suse.cz>
11845 PR gcov-profile/83669
11846 * gcov.c (output_intermediate_file): Add version to intermediate
11848 * doc/gcov.texi: Document new field 'version' in intermediate
11849 file format. Fix location of '-k' option of gcov command.
11851 2018-01-04 Martin Liska <mliska@suse.cz>
11854 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11856 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11858 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11860 2018-01-03 Martin Sebor <msebor@redhat.com>
11862 PR tree-optimization/83655
11863 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11864 checking calls with invalid arguments.
11866 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11868 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11869 (vectorizable_mask_load_store): Delete.
11870 (vectorizable_call): Return false for masked loads and stores.
11871 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11872 instead of gimple_assign_rhs1.
11873 (vectorizable_load): Handle IFN_MASK_LOAD.
11874 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11878 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11880 (vectorizable_mask_load_store): ...here.
11881 (vectorizable_load): ...and here.
11883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11885 * tree-vect-stmts.c (vect_build_all_ones_mask)
11886 (vect_build_zero_merge_argument): New functions, split out from...
11887 (vectorizable_load): ...here.
11889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11891 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11893 (vectorizable_mask_load_store): ...here.
11894 (vectorizable_store): ...and here.
11896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11898 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11900 (vectorizable_mask_load_store): ...here.
11902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11904 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11905 (vect_model_store_cost): Take a vec_load_store_type instead of a
11907 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11908 (vect_model_store_cost): Take a vec_load_store_type instead of a
11910 (vectorizable_mask_load_store): Update accordingly.
11911 (vectorizable_store): Likewise.
11912 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11914 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11916 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11917 IFN_MASK_LOAD calls here rather than...
11918 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11921 Alan Hayward <alan.hayward@arm.com>
11922 David Sherwood <david.sherwood@arm.com>
11924 * expmed.c (extract_bit_field_1): For vector extracts,
11925 fall back to extract_bit_field_as_subreg if vec_extract
11928 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11929 Alan Hayward <alan.hayward@arm.com>
11930 David Sherwood <david.sherwood@arm.com>
11932 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11933 they are variable or constant sized.
11934 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11935 slots for constant-sized data.
11937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11938 Alan Hayward <alan.hayward@arm.com>
11939 David Sherwood <david.sherwood@arm.com>
11941 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11942 handling COND_EXPRs with boolean comparisons, try to find a better
11943 basis for the mask type than the boolean itself.
11945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11947 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11948 is calculated and how it can be overridden.
11949 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11950 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11952 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11956 Alan Hayward <alan.hayward@arm.com>
11957 David Sherwood <david.sherwood@arm.com>
11959 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11960 Remove the mode argument.
11961 (aarch64_simd_valid_immediate): Remove the mode and inverse
11963 * config/aarch64/iterators.md (bitsize): New iterator.
11964 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11965 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11966 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11967 aarch64_simd_valid_immediate.
11968 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11969 (aarch64_reg_or_bic_imm): Likewise.
11970 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11971 with an insn_type enum and msl with a modifier_type enum.
11972 Replace element_width with a scalar_mode. Change the shift
11973 to unsigned int. Add constructors for scalar_float_mode and
11974 scalar_int_mode elements.
11975 (aarch64_vect_float_const_representable_p): Delete.
11976 (aarch64_can_const_movi_rtx_p)
11977 (aarch64_simd_scalar_immediate_valid_for_move)
11978 (aarch64_simd_make_constant): Update call to
11979 aarch64_simd_valid_immediate.
11980 (aarch64_advsimd_valid_immediate_hs): New function.
11981 (aarch64_advsimd_valid_immediate): Likewise.
11982 (aarch64_simd_valid_immediate): Remove mode and inverse
11983 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11984 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11985 and aarch64_float_const_representable_p on the result.
11986 (aarch64_output_simd_mov_immediate): Remove mode argument.
11987 Update call to aarch64_simd_valid_immediate and use of
11988 simd_immediate_info.
11989 (aarch64_output_scalar_simd_mov_immediate): Update call
11992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11993 Alan Hayward <alan.hayward@arm.com>
11994 David Sherwood <david.sherwood@arm.com>
11996 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11997 (mode_nunits): Likewise CONST_MODE_NUNITS.
11998 * machmode.def (ADJUST_NUNITS): Document.
11999 * genmodes.c (mode_data::need_nunits_adj): New field.
12000 (blank_mode): Update accordingly.
12001 (adj_nunits): New variable.
12002 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12004 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12005 listed in adj_nunits.
12006 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12007 listed in adj_nunits. Don't emit case statements for such modes.
12008 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12009 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12010 nothing if adj_nunits is nonnull.
12011 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12012 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12013 (emit_mode_fbit): Update use of print_maybe_const_decl.
12014 (emit_move_size): Likewise. Treat the array as non-const
12016 (emit_mode_adjustments): Handle adj_nunits.
12018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12020 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12021 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12022 (VECTOR_MODES): Use it.
12023 (make_vector_modes): Take the prefix as an argument.
12025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12026 Alan Hayward <alan.hayward@arm.com>
12027 David Sherwood <david.sherwood@arm.com>
12029 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12030 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12031 for MODE_VECTOR_BOOL.
12032 * machmode.def (VECTOR_BOOL_MODE): Document.
12033 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12034 (make_vector_bool_mode): New function.
12035 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12037 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12038 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12040 * stor-layout.c (int_mode_for_mode): Likewise.
12041 * tree.c (build_vector_type_for_mode): Likewise.
12042 * varasm.c (output_constant_pool_2): Likewise.
12043 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12044 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12045 for MODE_VECTOR_BOOL.
12046 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12047 of mode class checks.
12048 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12049 instead of a list of mode class checks.
12050 (expand_vector_scalar_condition): Likewise.
12051 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12054 Alan Hayward <alan.hayward@arm.com>
12055 David Sherwood <david.sherwood@arm.com>
12057 * machmode.h (mode_size): Change from unsigned short to
12059 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12060 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12061 or if measurement_type is not polynomial.
12062 (fixed_size_mode::includes_p): Check for constant-sized modes.
12063 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12064 return a poly_uint16 rather than an unsigned short.
12065 (emit_mode_size): Change the type of mode_size from unsigned short
12066 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12067 (emit_mode_adjustments): Cope with polynomial vector sizes.
12068 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12070 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12072 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12073 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12074 * caller-save.c (setup_save_areas): Likewise.
12075 (replace_reg_with_saved_mem): Likewise.
12076 * calls.c (emit_library_call_value_1): Likewise.
12077 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12078 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12079 (gen_lowpart_for_combine): Likewise.
12080 * convert.c (convert_to_integer_1): Likewise.
12081 * cse.c (equiv_constant, cse_insn): Likewise.
12082 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12083 (cselib_subst_to_values): Likewise.
12084 * dce.c (word_dce_process_block): Likewise.
12085 * df-problems.c (df_word_lr_mark_ref): Likewise.
12086 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12087 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12088 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12089 (rtl_for_decl_location): Likewise.
12090 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12091 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12092 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12093 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12094 (expand_expr_real_1): Likewise.
12095 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12096 (pad_below): Likewise.
12097 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12098 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12099 * ira.c (get_subreg_tracking_sizes): Likewise.
12100 * ira-build.c (ira_create_allocno_objects): Likewise.
12101 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12102 (ira_sort_regnos_for_alter_reg): Likewise.
12103 * ira-costs.c (record_operand_costs): Likewise.
12104 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12105 (resolve_simple_move): Likewise.
12106 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12107 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12108 (lra_constraints): Likewise.
12109 (CONST_POOL_OK_P): Reject variable-sized modes.
12110 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12111 (add_pseudo_to_slot, lra_spill): Likewise.
12112 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12113 * optabs-query.c (get_best_extraction_insn): Likewise.
12114 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12115 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12116 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12117 * recog.c (offsettable_address_addr_space_p): Likewise.
12118 * regcprop.c (maybe_mode_change): Likewise.
12119 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12120 * regrename.c (build_def_use): Likewise.
12121 * regstat.c (dump_reg_info): Likewise.
12122 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12123 (find_reloads, find_reloads_subreg_address): Likewise.
12124 * reload1.c (eliminate_regs_1): Likewise.
12125 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12126 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12127 (simplify_binary_operation_1, simplify_subreg): Likewise.
12128 * targhooks.c (default_function_arg_padding): Likewise.
12129 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12130 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12131 (verify_gimple_assign_ternary): Likewise.
12132 * tree-inline.c (estimate_move_cost): Likewise.
12133 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12134 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12135 (get_address_cost_ainc): Likewise.
12136 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12137 (vect_supportable_dr_alignment): Likewise.
12138 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12139 (vectorizable_reduction): Likewise.
12140 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12141 (vectorizable_operation, vectorizable_load): Likewise.
12142 * tree.c (build_same_sized_truth_vector_type): Likewise.
12143 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12144 * var-tracking.c (emit_note_insn_var_location): Likewise.
12145 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12146 (ADDR_VEC_ALIGN): Likewise.
12148 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12149 Alan Hayward <alan.hayward@arm.com>
12150 David Sherwood <david.sherwood@arm.com>
12152 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12154 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12155 or if measurement_type is polynomial.
12156 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12157 * combine.c (make_extraction): Likewise.
12158 * dse.c (find_shift_sequence): Likewise.
12159 * dwarf2out.c (mem_loc_descriptor): Likewise.
12160 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12161 (extract_bit_field, extract_low_bits): Likewise.
12162 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12163 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12164 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12165 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12166 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12167 * reload.c (find_reloads): Likewise.
12168 * reload1.c (alter_reg): Likewise.
12169 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12170 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12171 * tree-if-conv.c (predicate_mem_writes): Likewise.
12172 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12173 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12174 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12175 * valtrack.c (dead_debug_insert_temp): Likewise.
12176 * varasm.c (mergeable_constant_section): Likewise.
12177 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12180 Alan Hayward <alan.hayward@arm.com>
12181 David Sherwood <david.sherwood@arm.com>
12183 * expr.c (expand_assignment): Cope with polynomial mode sizes
12184 when assigning to a CONCAT.
12186 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12187 Alan Hayward <alan.hayward@arm.com>
12188 David Sherwood <david.sherwood@arm.com>
12190 * machmode.h (mode_precision): Change from unsigned short to
12192 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12194 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12195 or if measurement_type is not polynomial.
12196 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12197 in which the mode is already known to be a scalar_int_mode.
12198 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12199 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12201 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12202 for GET_MODE_PRECISION.
12203 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12204 for GET_MODE_PRECISION.
12205 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12207 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12208 (expand_field_assignment, make_extraction): Likewise.
12209 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12210 (get_last_value): Likewise.
12211 * convert.c (convert_to_integer_1): Likewise.
12212 * cse.c (cse_insn): Likewise.
12213 * expr.c (expand_expr_real_1): Likewise.
12214 * lra-constraints.c (simplify_operand_subreg): Likewise.
12215 * optabs-query.c (can_atomic_load_p): Likewise.
12216 * optabs.c (expand_atomic_load): Likewise.
12217 (expand_atomic_store): Likewise.
12218 * ree.c (combine_reaching_defs): Likewise.
12219 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12220 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12221 * tree.h (type_has_mode_precision_p): Likewise.
12222 * ubsan.c (instrument_si_overflow): Likewise.
12224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12225 Alan Hayward <alan.hayward@arm.com>
12226 David Sherwood <david.sherwood@arm.com>
12228 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12229 polynomial numbers of units.
12230 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12231 (valid_vector_subparts_p): New function.
12232 (build_vector_type): Remove temporary shim and take the number
12233 of units as a poly_uint64 rather than an int.
12234 (build_opaque_vector_type): Take the number of units as a
12235 poly_uint64 rather than an int.
12236 * tree.c (build_vector_from_ctor): Handle polynomial
12237 TYPE_VECTOR_SUBPARTS.
12238 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12239 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12240 (build_vector_from_val): If the number of units is variable,
12241 use build_vec_duplicate_cst for constant operands and
12242 VEC_DUPLICATE_EXPR otherwise.
12243 (make_vector_type): Remove temporary is_constant ().
12244 (build_vector_type, build_opaque_vector_type): Take the number of
12245 units as a poly_uint64 rather than an int.
12246 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12248 * cfgexpand.c (expand_debug_expr): Likewise.
12249 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12250 (store_constructor, expand_expr_real_1): Likewise.
12251 (const_scalar_mask_from_tree): Likewise.
12252 * fold-const-call.c (fold_const_reduction): Likewise.
12253 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12254 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12255 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12256 (fold_relational_const): Likewise.
12257 (native_interpret_vector): Likewise. Change the size from an
12258 int to an unsigned int.
12259 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12260 TYPE_VECTOR_SUBPARTS.
12261 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12262 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12263 duplicating a non-constant operand into a variable-length vector.
12264 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12265 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12266 * ipa-icf.c (sem_variable::equals): Likewise.
12267 * match.pd: Likewise.
12268 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12269 * print-tree.c (print_node): Likewise.
12270 * stor-layout.c (layout_type): Likewise.
12271 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12272 * tree-cfg.c (verify_gimple_comparison): Likewise.
12273 (verify_gimple_assign_binary): Likewise.
12274 (verify_gimple_assign_ternary): Likewise.
12275 (verify_gimple_assign_single): Likewise.
12276 * tree-pretty-print.c (dump_generic_node): Likewise.
12277 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12278 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12279 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12280 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12281 (vect_shift_permute_load_chain): Likewise.
12282 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12283 (expand_vector_condition, optimize_vector_constructor): Likewise.
12284 (lower_vec_perm, get_compute_type): Likewise.
12285 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12286 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12287 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12288 (vect_recog_mask_conversion_pattern): Likewise.
12289 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12290 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12291 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12292 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12293 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12294 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12295 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12296 (supportable_widening_operation): Likewise.
12297 (supportable_narrowing_operation): Likewise.
12298 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12300 * varasm.c (output_constant): Likewise.
12302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12303 Alan Hayward <alan.hayward@arm.com>
12304 David Sherwood <david.sherwood@arm.com>
12306 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12307 so that both the length == 3 and length != 3 cases set up their
12308 own permute vectors. Add comments explaining why we know the
12309 number of elements is constant.
12310 (vect_permute_load_chain): Likewise.
12312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12313 Alan Hayward <alan.hayward@arm.com>
12314 David Sherwood <david.sherwood@arm.com>
12316 * machmode.h (mode_nunits): Change from unsigned char to
12318 (ONLY_FIXED_SIZE_MODES): New macro.
12319 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12320 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12321 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12323 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12324 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12325 or if measurement_type is not polynomial.
12326 * genmodes.c (ZERO_COEFFS): New macro.
12327 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12329 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12330 Use ZERO_COEFFS when emitting initializers.
12331 * data-streamer.h (bp_pack_poly_value): New function.
12332 (bp_unpack_poly_value): Likewise.
12333 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12334 for GET_MODE_NUNITS.
12335 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12336 for GET_MODE_NUNITS.
12337 * tree.c (make_vector_type): Remove temporary shim and make
12338 the real function take the number of units as a poly_uint64
12339 rather than an int.
12340 (build_vector_type_for_mode): Handle polynomial nunits.
12341 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12342 * emit-rtl.c (const_vec_series_p_1): Likewise.
12343 (gen_rtx_CONST_VECTOR): Likewise.
12344 * fold-const.c (test_vec_duplicate_folding): Likewise.
12345 * genrecog.c (validate_pattern): Likewise.
12346 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12347 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12348 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12349 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12350 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12351 * rtlanal.c (subreg_get_info): Likewise.
12352 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12353 (vect_grouped_load_supported): Likewise.
12354 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12355 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12356 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12357 (simplify_const_unary_operation, simplify_binary_operation_1)
12358 (simplify_const_binary_operation, simplify_ternary_operation)
12359 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12360 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12361 instead of CONST_VECTOR_NUNITS.
12362 * varasm.c (output_constant_pool_2): Likewise.
12363 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12364 explicit-encoded elements in the XVEC for variable-length vectors.
12366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12368 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12371 Alan Hayward <alan.hayward@arm.com>
12372 David Sherwood <david.sherwood@arm.com>
12374 * coretypes.h (fixed_size_mode): Declare.
12375 (fixed_size_mode_pod): New typedef.
12376 * builtins.h (target_builtins::x_apply_args_mode)
12377 (target_builtins::x_apply_result_mode): Change type to
12378 fixed_size_mode_pod.
12379 * builtins.c (apply_args_size, apply_result_size, result_vector)
12380 (expand_builtin_apply_args_1, expand_builtin_apply)
12381 (expand_builtin_return): Update accordingly.
12383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12385 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12386 * cselib.c (cselib_hash_rtx): Likewise.
12387 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12388 CONST_VECTOR encoding.
12390 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12391 Jeff Law <law@redhat.com>
12394 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12395 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12396 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12397 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12400 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12401 explicitly probe *sp in a noreturn function if there were any callee
12402 register saves or frame pointer is needed.
12404 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12407 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12408 BLKmode for ternary, binary or unary expressions.
12411 * var-tracking.c (delete_vta_debug_insn): New inline function.
12412 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12413 insns from get_insns () to NULL instead of each bb separately.
12414 Use delete_vta_debug_insn. No longer static.
12415 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12416 delete_vta_debug_insns callers.
12417 * rtl.h (delete_vta_debug_insns): Declare.
12418 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12419 instead of variable_tracking_main.
12421 2018-01-03 Martin Sebor <msebor@redhat.com>
12423 PR tree-optimization/83603
12424 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12425 arguments past the endof the argument list in functions declared
12426 without a prototype.
12427 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12428 Avoid checking when arguments are null.
12430 2018-01-03 Martin Sebor <msebor@redhat.com>
12433 * doc/extend.texi (attribute const): Fix a typo.
12434 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12435 issuing -Wsuggest-attribute for void functions.
12437 2018-01-03 Martin Sebor <msebor@redhat.com>
12439 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12440 offset_int::from instead of wide_int::to_shwi.
12441 (maybe_diag_overlap): Remove assertion.
12442 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12443 * gimple-ssa-sprintf.c (format_directive): Same.
12444 (parse_directive): Same.
12445 (sprintf_dom_walker::compute_format_length): Same.
12446 (try_substitute_return_value): Same.
12448 2018-01-03 Jeff Law <law@redhat.com>
12450 PR middle-end/83654
12451 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12452 non-constant residual for zero at runtime and avoid probing in
12453 that case. Reorganize code for trailing problem to mirror handling
12456 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12458 PR tree-optimization/83501
12459 * tree-ssa-strlen.c (get_string_cst): New.
12460 (handle_char_store): Call get_string_cst.
12462 2018-01-03 Martin Liska <mliska@suse.cz>
12464 PR tree-optimization/83593
12465 * tree-ssa-strlen.c: Include tree-cfg.h.
12466 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12467 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12468 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12470 (strlen_dom_walker::before_dom_children): Call
12471 gimple_purge_dead_eh_edges. Dump tranformation with details
12473 (strlen_dom_walker::before_dom_children): Update call by adding
12474 new argument cleanup_eh.
12475 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12477 2018-01-03 Martin Liska <mliska@suse.cz>
12480 * cif-code.def (VARIADIC_THUNK): New enum value.
12481 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12484 2018-01-03 Jan Beulich <jbeulich@suse.com>
12486 * sse.md (mov<mode>_internal): Tighten condition for when to use
12487 vmovdqu<ssescalarsize> for TI and OI modes.
12489 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12491 Update copyright years.
12493 2018-01-03 Martin Liska <mliska@suse.cz>
12496 * ipa-visibility.c (function_and_variable_visibility): Skip
12497 functions with noipa attribure.
12499 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12501 * gcc.c (process_command): Update copyright notice dates.
12502 * gcov-dump.c (print_version): Ditto.
12503 * gcov.c (print_version): Ditto.
12504 * gcov-tool.c (print_version): Ditto.
12505 * gengtype.c (create_file): Ditto.
12506 * doc/cpp.texi: Bump @copying's copyright year.
12507 * doc/cppinternals.texi: Ditto.
12508 * doc/gcc.texi: Ditto.
12509 * doc/gccint.texi: Ditto.
12510 * doc/gcov.texi: Ditto.
12511 * doc/install.texi: Ditto.
12512 * doc/invoke.texi: Ditto.
12514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12516 * vector-builder.h (vector_builder::m_full_nelts): Change from
12517 unsigned int to poly_uint64.
12518 (vector_builder::full_nelts): Update prototype accordingly.
12519 (vector_builder::new_vector): Likewise.
12520 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12521 (vector_builder::operator ==): Likewise.
12522 (vector_builder::finalize): Likewise.
12523 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12524 Take the number of elements as a poly_uint64 rather than an
12526 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12527 from unsigned int to poly_uint64.
12528 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12529 (vec_perm_indices::new_vector): Likewise.
12530 (vec_perm_indices::length): Likewise.
12531 (vec_perm_indices::nelts_per_input): Likewise.
12532 (vec_perm_indices::input_nelts): Likewise.
12533 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12534 number of elements per input as a poly_uint64 rather than an
12535 unsigned int. Use the original encoding for variable-length
12536 vectors, rather than clamping each individual element.
12537 For the second and subsequent elements in each pattern,
12538 clamp the step and base before clamping their sum.
12539 (vec_perm_indices::series_p): Handle polynomial element counts.
12540 (vec_perm_indices::all_in_range_p): Likewise.
12541 (vec_perm_indices_to_tree): Likewise.
12542 (vec_perm_indices_to_rtx): Likewise.
12543 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12544 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12545 (tree_vector_builder::new_binary_operation): Handle polynomial
12546 element counts. Return false if we need to know the number
12547 of elements at compile time.
12548 * fold-const.c (fold_vec_perm): Punt if the number of elements
12549 isn't known at compile time.
12551 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12553 * vec-perm-indices.h (vec_perm_builder): Change element type
12554 from HOST_WIDE_INT to poly_int64.
12555 (vec_perm_indices::element_type): Update accordingly.
12556 (vec_perm_indices::clamp): Handle polynomial element_types.
12557 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12558 (vec_perm_indices::all_in_range_p): Likewise.
12559 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12561 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12562 polynomial vec_perm_indices element types.
12563 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12564 * fold-const.c (fold_vec_perm): Likewise.
12565 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12566 * tree-vect-generic.c (lower_vec_perm): Likewise.
12567 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12568 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12569 element type to HOST_WIDE_INT.
12571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12572 Alan Hayward <alan.hayward@arm.com>
12573 David Sherwood <david.sherwood@arm.com>
12575 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12576 rather than an int. Use plus_constant.
12577 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12578 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12581 Alan Hayward <alan.hayward@arm.com>
12582 David Sherwood <david.sherwood@arm.com>
12584 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12585 a HOST_WIDE_INT to a poly_int64.
12587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12588 Alan Hayward <alan.hayward@arm.com>
12589 David Sherwood <david.sherwood@arm.com>
12591 * calls.c (load_register_parameters): Cope with polynomial
12592 mode sizes. Require a constant size for BLKmode parameters
12593 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12594 forces a parameter to be padded at the lsb end in order to
12595 fill a complete number of words, require the parameter size
12596 to be ordered wrt UNITS_PER_WORD.
12598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12599 Alan Hayward <alan.hayward@arm.com>
12600 David Sherwood <david.sherwood@arm.com>
12602 * reload1.c (spill_stack_slot_width): Change element type
12603 from unsigned int to poly_uint64_pod.
12604 (alter_reg): Treat mode sizes as polynomial.
12606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12607 Alan Hayward <alan.hayward@arm.com>
12608 David Sherwood <david.sherwood@arm.com>
12610 * reload.c (complex_word_subreg_p): New function.
12611 (reload_inner_reg_of_subreg, push_reload): Use it.
12613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12614 Alan Hayward <alan.hayward@arm.com>
12615 David Sherwood <david.sherwood@arm.com>
12617 * lra-constraints.c (process_alt_operands): Reject matched
12618 operands whose sizes aren't ordered.
12619 (match_reload): Refer to this check here.
12621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12622 Alan Hayward <alan.hayward@arm.com>
12623 David Sherwood <david.sherwood@arm.com>
12625 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12626 that the mode size is in the set {1, 2, 4, 8, 16}.
12628 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12629 Alan Hayward <alan.hayward@arm.com>
12630 David Sherwood <david.sherwood@arm.com>
12632 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12633 Use plus_constant instead of gen_rtx_PLUS.
12635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12636 Alan Hayward <alan.hayward@arm.com>
12637 David Sherwood <david.sherwood@arm.com>
12639 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12640 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12641 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12642 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12643 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12644 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12645 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12646 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12647 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12648 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12650 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12651 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12652 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12653 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12654 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12655 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12656 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12657 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12658 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12659 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12661 * expr.c (emit_move_resolve_push): Treat the input and result
12662 of PUSH_ROUNDING as a poly_int64.
12663 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12664 (emit_push_insn): Likewise.
12665 * lra-eliminations.c (mark_not_eliminable): Likewise.
12666 * recog.c (push_operand): Likewise.
12667 * reload1.c (elimination_effects): Likewise.
12668 * rtlanal.c (nonzero_bits1): Likewise.
12669 * calls.c (store_one_arg): Likewise. Require the padding to be
12670 known at compile time.
12672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12673 Alan Hayward <alan.hayward@arm.com>
12674 David Sherwood <david.sherwood@arm.com>
12676 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12677 Use plus_constant instead of gen_rtx_PLUS.
12679 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12680 Alan Hayward <alan.hayward@arm.com>
12681 David Sherwood <david.sherwood@arm.com>
12683 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12684 rather than an int.
12686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12687 Alan Hayward <alan.hayward@arm.com>
12688 David Sherwood <david.sherwood@arm.com>
12690 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12691 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12692 via stack temporaries. Treat the mode size as polynomial too.
12694 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12695 Alan Hayward <alan.hayward@arm.com>
12696 David Sherwood <david.sherwood@arm.com>
12698 * expr.c (expand_expr_real_2): When handling conversions involving
12699 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12700 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12701 as a poly_uint64 too.
12703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12704 Alan Hayward <alan.hayward@arm.com>
12705 David Sherwood <david.sherwood@arm.com>
12707 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12709 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12710 Alan Hayward <alan.hayward@arm.com>
12711 David Sherwood <david.sherwood@arm.com>
12713 * combine.c (can_change_dest_mode): Handle polynomial
12714 REGMODE_NATURAL_SIZE.
12715 * expmed.c (store_bit_field_1): Likewise.
12716 * expr.c (store_constructor): Likewise.
12717 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12718 and polynomial REGMODE_NATURAL_SIZE.
12719 (gen_lowpart_common): Likewise.
12720 * reginfo.c (record_subregs_of_mode): Likewise.
12721 * rtlanal.c (read_modify_subreg_p): Likewise.
12723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12724 Alan Hayward <alan.hayward@arm.com>
12725 David Sherwood <david.sherwood@arm.com>
12727 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12728 numbers of elements.
12730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12731 Alan Hayward <alan.hayward@arm.com>
12732 David Sherwood <david.sherwood@arm.com>
12734 * match.pd: Cope with polynomial numbers of vector elements.
12736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12737 Alan Hayward <alan.hayward@arm.com>
12738 David Sherwood <david.sherwood@arm.com>
12740 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12741 in a POINTER_PLUS_EXPR.
12743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12744 Alan Hayward <alan.hayward@arm.com>
12745 David Sherwood <david.sherwood@arm.com>
12747 * omp-simd-clone.c (simd_clone_subparts): New function.
12748 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12749 (ipa_simd_modify_function_body): Likewise.
12751 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12752 Alan Hayward <alan.hayward@arm.com>
12753 David Sherwood <david.sherwood@arm.com>
12755 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12756 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12757 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12758 (expand_vector_condition, vector_element): Likewise.
12759 (subparts_gt): New function.
12760 (get_compute_type): Use subparts_gt.
12761 (count_type_subparts): Delete.
12762 (expand_vector_operations_1): Use subparts_gt instead of
12763 count_type_subparts.
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 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12770 (vect_compile_time_alias): ...this new function. Do the calculation
12771 on poly_ints rather than trees.
12772 (vect_prune_runtime_alias_test_list): Update call accordingly.
12774 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12775 Alan Hayward <alan.hayward@arm.com>
12776 David Sherwood <david.sherwood@arm.com>
12778 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12780 (vect_schedule_slp_instance): Likewise.
12782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12783 Alan Hayward <alan.hayward@arm.com>
12784 David Sherwood <david.sherwood@arm.com>
12786 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12787 constant and extern definitions for variable-length vectors.
12788 (vect_get_constant_vectors): Note that the number of units
12789 is known to be constant.
12791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12792 Alan Hayward <alan.hayward@arm.com>
12793 David Sherwood <david.sherwood@arm.com>
12795 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12796 of units as polynomial. Choose between WIDE and NARROW based
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 * tree-vect-stmts.c (simd_clone_subparts): New function.
12804 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12807 Alan Hayward <alan.hayward@arm.com>
12808 David Sherwood <david.sherwood@arm.com>
12810 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12811 vectors as polynomial. Use build_index_vector for
12812 IFN_GOMP_SIMD_LANE.
12814 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12815 Alan Hayward <alan.hayward@arm.com>
12816 David Sherwood <david.sherwood@arm.com>
12818 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12819 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12820 for variable-length vectors.
12821 (vectorizable_mask_load_store): Treat the number of units as
12822 polynomial, asserting that it is constant if the condition has
12823 already been enforced.
12824 (vectorizable_store, vectorizable_load): Likewise.
12826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12827 Alan Hayward <alan.hayward@arm.com>
12828 David Sherwood <david.sherwood@arm.com>
12830 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12831 of units as polynomial. Punt if we can't tell at compile time
12832 which vector contains the final result.
12834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12835 Alan Hayward <alan.hayward@arm.com>
12836 David Sherwood <david.sherwood@arm.com>
12838 * tree-vect-loop.c (vectorizable_induction): Treat the number
12839 of units as polynomial. Punt on SLP inductions. Use an integer
12840 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12841 cast of such a series for variable-length floating-point
12844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12845 Alan Hayward <alan.hayward@arm.com>
12846 David Sherwood <david.sherwood@arm.com>
12848 * tree.h (build_index_vector): Declare.
12849 * tree.c (build_index_vector): New function.
12850 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12851 of units as polynomial, forcibly converting it to a constant if
12852 vectorizable_reduction has already enforced the condition.
12853 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12854 to create a {1,2,3,...} vector.
12855 (vectorizable_reduction): Treat the number of units as polynomial.
12856 Choose vectype_in based on the largest scalar element size rather
12857 than the smallest number of units. Enforce the restrictions
12860 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12861 Alan Hayward <alan.hayward@arm.com>
12862 David Sherwood <david.sherwood@arm.com>
12864 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12865 number of units as polynomial.
12867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12868 Alan Hayward <alan.hayward@arm.com>
12869 David Sherwood <david.sherwood@arm.com>
12871 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12872 * target.def (autovectorize_vector_sizes): Return the vector sizes
12873 by pointer, using vector_sizes rather than a bitmask.
12874 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12875 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12876 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12878 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12879 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12880 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12881 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12882 * omp-general.c (omp_max_vf): Likewise.
12883 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12884 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12885 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12886 * tree-vect-slp.c (vect_slp_bb): Likewise.
12887 * doc/tm.texi: Regenerate.
12888 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12890 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12891 the vector size as a poly_uint64 rather than an unsigned int.
12892 (current_vector_size): Change from an unsigned int to a poly_uint64.
12893 (get_vectype_for_scalar_type): Update accordingly.
12894 * tree.h (build_truth_vector_type): Take the size and number of
12895 units as a poly_uint64 rather than an unsigned int.
12896 (build_vector_type): Add a temporary overload that takes
12897 the number of units as a poly_uint64 rather than an unsigned int.
12898 * tree.c (make_vector_type): Likewise.
12899 (build_truth_vector_type): Take the number of units as a poly_uint64
12900 rather than an unsigned int.
12902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12903 Alan Hayward <alan.hayward@arm.com>
12904 David Sherwood <david.sherwood@arm.com>
12906 * target.def (get_mask_mode): Take the number of units and length
12907 as poly_uint64s rather than unsigned ints.
12908 * targhooks.h (default_get_mask_mode): Update accordingly.
12909 * targhooks.c (default_get_mask_mode): Likewise.
12910 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12911 * doc/tm.texi: Regenerate.
12913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12914 Alan Hayward <alan.hayward@arm.com>
12915 David Sherwood <david.sherwood@arm.com>
12917 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12918 * omp-general.c (omp_max_vf): Likewise.
12919 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12920 (expand_omp_simd): Handle polynomial safelen.
12921 * omp-low.c (omplow_simd_context): Add a default constructor.
12922 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12923 (lower_rec_simd_input_clauses): Update accordingly.
12924 (lower_rec_input_clauses): Likewise.
12926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12927 Alan Hayward <alan.hayward@arm.com>
12928 David Sherwood <david.sherwood@arm.com>
12930 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12931 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12932 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12933 (vect_analyze_slp_cost): Likewise.
12934 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12935 (vect_model_load_cost): Likewise.
12937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12938 Alan Hayward <alan.hayward@arm.com>
12939 David Sherwood <david.sherwood@arm.com>
12941 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12942 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12943 from an unsigned int * to a poly_uint64_pod *.
12944 (calculate_unrolling_factor): New function.
12945 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12948 Alan Hayward <alan.hayward@arm.com>
12949 David Sherwood <david.sherwood@arm.com>
12951 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12952 from an unsigned int to a poly_uint64.
12953 (_loop_vec_info::slp_unrolling_factor): Likewise.
12954 (_loop_vec_info::vectorization_factor): Change from an int
12956 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12957 (vect_get_num_vectors): New function.
12958 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12959 (vect_get_num_copies): Use vect_get_num_vectors.
12960 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12961 to an unsigned int *.
12962 (vect_analyze_data_refs): Change min_vf from an int * to a
12964 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12965 than an unsigned HOST_WIDE_INT.
12966 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12967 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12968 to an unsigned int *.
12969 (vect_analyze_data_ref_dependences): Likewise.
12970 (vect_compute_data_ref_alignment): Handle polynomial vf.
12971 (vect_enhance_data_refs_alignment): Likewise.
12972 (vect_prune_runtime_alias_test_list): Likewise.
12973 (vect_shift_permute_load_chain): Likewise.
12974 (vect_supportable_dr_alignment): Likewise.
12975 (dependence_distance_ge_vf): Take the vectorization factor as a
12976 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12977 (vect_analyze_data_refs): Change min_vf from an int * to a
12979 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12980 vfm1 as a poly_uint64 rather than an int. Make the same change
12981 for the returned bound_scalar.
12982 (vect_gen_vector_loop_niters): Handle polynomial vf.
12983 (vect_do_peeling): Likewise. Update call to
12984 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12985 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12987 * tree-vect-loop.c (vect_determine_vectorization_factor)
12988 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12989 (vect_get_known_peeling_cost): Likewise.
12990 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12991 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12992 (vect_transform_loop): Likewise. Use the lowest possible VF when
12993 updating the upper bounds of the loop.
12994 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12995 rather than an int.
12996 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12997 polynomial unroll factors.
12998 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12999 (vect_make_slp_decision): Likewise.
13000 (vect_supported_load_permutation_p): Likewise, and polynomial
13002 (vect_analyze_slp_cost): Handle polynomial vf.
13003 (vect_slp_analyze_node_operations): Likewise.
13004 (vect_slp_analyze_bb_1): Likewise.
13005 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13006 than an unsigned HOST_WIDE_INT.
13007 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13008 (vectorizable_load): Handle polynomial vf.
13009 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13011 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13013 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13014 Alan Hayward <alan.hayward@arm.com>
13015 David Sherwood <david.sherwood@arm.com>
13017 * match.pd: Handle bit operations involving three constants
13018 and try to fold one pair.
13020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13022 * tree-vect-loop-manip.c: Include gimple-fold.h.
13023 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13024 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13025 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13026 Add a path that uses a step of VF instead of 1, but disable it
13028 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13029 and niters_no_overflow parameters. Update calls to
13030 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13031 Create a new SSA name if the latter choses to use a ste other
13032 than zero, and return it via niters_vector_mult_vf_var.
13033 * tree-vect-loop.c (vect_transform_loop): Update calls to
13034 vect_do_peeling, vect_gen_vector_loop_niters and
13035 slpeel_make_loop_iterate_ntimes.
13036 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13037 (vect_gen_vector_loop_niters): Update declarations after above changes.
13039 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13041 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13042 128-bit round to integer instructions.
13043 (ceil<mode>2): Likewise.
13044 (btrunc<mode>2): Likewise.
13045 (round<mode>2): Likewise.
13047 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13049 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13050 unaligned VSX load/store on P8/P9.
13051 (expand_block_clear): Allow the use of unaligned VSX
13052 load/store on P8/P9.
13054 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13056 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13058 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13059 swap associated with both a load and a store.
13061 2018-01-02 Andrew Waterman <andrew@sifive.com>
13063 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13064 * config/riscv/riscv.md (clear_cache): Use it.
13066 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13068 * web.c: Remove out-of-date comment.
13070 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13072 * expr.c (fixup_args_size_notes): Check that any existing
13073 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13074 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13075 (emit_single_push_insn): ...here.
13077 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13079 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13080 (const_vector_encoded_nelts): New function.
13081 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13082 (const_vector_int_elt, const_vector_elt): Declare.
13083 * emit-rtl.c (const_vector_int_elt_1): New function.
13084 (const_vector_elt): Likewise.
13085 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13086 of CONST_VECTOR_ELT.
13088 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13090 * expr.c: Include rtx-vector-builder.h.
13091 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13092 directly on the tree encoding.
13093 (const_vector_from_tree): Likewise.
13094 * optabs.c: Include rtx-vector-builder.h.
13095 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13096 sequence of "u" values.
13097 * vec-perm-indices.c: Include rtx-vector-builder.h.
13098 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13099 directly on the vec_perm_indices encoding.
13101 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13103 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13104 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13105 * rtx-vector-builder.h: New file.
13106 * rtx-vector-builder.c: Likewise.
13107 * rtl.h (rtx_def::u2): Add a const_vector field.
13108 (CONST_VECTOR_NPATTERNS): New macro.
13109 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13110 (CONST_VECTOR_DUPLICATE_P): Likewise.
13111 (CONST_VECTOR_STEPPED_P): Likewise.
13112 (CONST_VECTOR_ENCODED_ELT): Likewise.
13113 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13114 (unwrap_const_vec_duplicate): Likewise.
13115 (const_vec_series_p): Check for a non-duplicated vector encoding.
13116 Say that the function only returns true for integer vectors.
13117 * emit-rtl.c: Include rtx-vector-builder.h.
13118 (gen_const_vec_duplicate_1): Delete.
13119 (gen_const_vector): Call gen_const_vec_duplicate instead of
13120 gen_const_vec_duplicate_1.
13121 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13122 (gen_const_vec_duplicate): Use rtx_vector_builder.
13123 (gen_const_vec_series): Likewise.
13124 (gen_rtx_CONST_VECTOR): Likewise.
13125 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13126 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13127 Build a new vector rather than modifying a CONST_VECTOR in-place.
13128 (handle_special_swappables): Update call accordingly.
13129 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13130 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13131 Build a new vector rather than modifying a CONST_VECTOR in-place.
13132 (handle_special_swappables): Update call accordingly.
13134 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13136 * simplify-rtx.c (simplify_const_binary_operation): Use
13137 CONST_VECTOR_ELT instead of XVECEXP.
13139 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13141 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13142 the selector elements to be different from the data elements
13143 if the selector is a VECTOR_CST.
13144 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13145 ssizetype for the selector.
13147 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13149 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13150 before testing each element individually.
13151 * tree-vect-generic.c (lower_vec_perm): Likewise.
13153 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13155 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13156 * selftest-run-tests.c (selftest::run_tests): Call it.
13157 * vector-builder.h (vector_builder::operator ==): New function.
13158 (vector_builder::operator !=): Likewise.
13159 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13160 (vec_perm_indices::all_from_input_p): New function.
13161 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13162 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13163 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13164 instead of reading the VECTOR_CST directly. Detect whether both
13165 vector inputs are the same before constructing the vec_perm_indices,
13166 and update the number of inputs argument accordingly. Use the
13167 utility functions added above. Only construct sel2 if we need to.
13169 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13171 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13172 the broadcast of the low byte.
13173 (expand_mult_highpart): Use an explicit encoding for the permutes.
13174 * optabs-query.c (can_mult_highpart_p): Likewise.
13175 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13176 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13177 (vectorizable_bswap): Likewise.
13178 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13179 explicit encoding for the power-of-2 permutes.
13180 (vect_permute_store_chain): Likewise.
13181 (vect_grouped_load_supported): Likewise.
13182 (vect_permute_load_chain): Likewise.
13184 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13186 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13187 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13188 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13189 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13190 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13191 (vect_gen_perm_mask_any): Likewise.
13193 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13195 * int-vector-builder.h: New file.
13196 * vec-perm-indices.h: Include int-vector-builder.h.
13197 (vec_perm_indices): Redefine as an int_vector_builder.
13198 (auto_vec_perm_indices): Delete.
13199 (vec_perm_builder): Redefine as a stand-alone class.
13200 (vec_perm_indices::vec_perm_indices): New function.
13201 (vec_perm_indices::clamp): Likewise.
13202 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13203 (vec_perm_indices::new_vector): New function.
13204 (vec_perm_indices::new_expanded_vector): Update for new
13205 vec_perm_indices class.
13206 (vec_perm_indices::rotate_inputs): New function.
13207 (vec_perm_indices::all_in_range_p): Operate directly on the
13208 encoded form, without computing elided elements.
13209 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13210 encoding. Update for new vec_perm_indices class.
13211 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13212 the given vec_perm_builder.
13213 (expand_vec_perm_var): Update vec_perm_builder constructor.
13214 (expand_mult_highpart): Use vec_perm_builder instead of
13215 auto_vec_perm_indices.
13216 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13217 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13218 or double series encoding as appropriate.
13219 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13220 vec_perm_indices instead of auto_vec_perm_indices.
13221 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13222 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13223 (vect_permute_store_chain): Likewise.
13224 (vect_grouped_load_supported): Likewise.
13225 (vect_permute_load_chain): Likewise.
13226 (vect_shift_permute_load_chain): Likewise.
13227 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13228 (vect_transform_slp_perm_load): Likewise.
13229 (vect_schedule_slp_instance): Likewise.
13230 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13231 (vectorizable_mask_load_store): Likewise.
13232 (vectorizable_bswap): Likewise.
13233 (vectorizable_store): Likewise.
13234 (vectorizable_load): Likewise.
13235 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13236 vec_perm_indices instead of auto_vec_perm_indices. Use
13237 tree_to_vec_perm_builder to read the vector from a tree.
13238 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13239 vec_perm_builder instead of a vec_perm_indices.
13240 (have_whole_vector_shift): Use vec_perm_builder and
13241 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13242 truncation to calc_vec_perm_mask_for_shift.
13243 (vect_create_epilog_for_reduction): Likewise.
13244 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13245 from auto_vec_perm_indices to vec_perm_indices.
13246 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13247 instead of changing individual elements.
13248 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13249 the vector in d.perm.
13250 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13251 from auto_vec_perm_indices to vec_perm_indices.
13252 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13253 instead of changing individual elements.
13254 (arm_vectorize_vec_perm_const): Use new_vector to install
13255 the vector in d.perm.
13256 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13257 Update vec_perm_builder constructor.
13258 (rs6000_expand_interleave): Likewise.
13259 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13260 (rs6000_expand_interleave): Likewise.
13262 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13264 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13265 to qimode could truncate the indices.
13266 * optabs.c (expand_vec_perm_var): Likewise.
13268 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13270 * Makefile.in (OBJS): Add vec-perm-indices.o.
13271 * vec-perm-indices.h: New file.
13272 * vec-perm-indices.c: Likewise.
13273 * target.h (vec_perm_indices): Replace with a forward class
13275 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13276 * optabs.h: Include vec-perm-indices.h.
13277 (expand_vec_perm): Delete.
13278 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13279 (expand_vec_perm_const): Declare.
13280 * target.def (vec_perm_const_ok): Replace with...
13281 (vec_perm_const): ...this new hook.
13282 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13283 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13284 * doc/tm.texi: Regenerate.
13285 * optabs.def (vec_perm_const): Delete.
13286 * doc/md.texi (vec_perm_const): Likewise.
13287 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13288 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13289 expand_vec_perm for constant permutation vectors. Assert that
13290 the mode of variable permutation vectors is the integer equivalent
13291 of the mode that is being permuted.
13292 * optabs-query.h (selector_fits_mode_p): Declare.
13293 * optabs-query.c: Include vec-perm-indices.h.
13294 (selector_fits_mode_p): New function.
13295 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13296 is defined, instead of checking whether the vec_perm_const_optab
13297 exists. Use targetm.vectorize.vec_perm_const instead of
13298 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13299 fit in the vector mode before using a variable permute.
13300 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13301 vec_perm_indices instead of an rtx.
13302 (expand_vec_perm): Replace with...
13303 (expand_vec_perm_const): ...this new function. Take the selector
13304 as a vec_perm_indices rather than an rtx. Also take the mode of
13305 the selector. Update call to shift_amt_for_vec_perm_mask.
13306 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13307 Use vec_perm_indices::new_expanded_vector to expand the original
13308 selector into bytes. Check whether the indices fit in the vector
13309 mode before using a variable permute.
13310 (expand_vec_perm_var): Make global.
13311 (expand_mult_highpart): Use expand_vec_perm_const.
13312 * fold-const.c: Includes vec-perm-indices.h.
13313 * tree-ssa-forwprop.c: Likewise.
13314 * tree-vect-data-refs.c: Likewise.
13315 * tree-vect-generic.c: Likewise.
13316 * tree-vect-loop.c: Likewise.
13317 * tree-vect-slp.c: Likewise.
13318 * tree-vect-stmts.c: Likewise.
13319 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13321 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13322 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13323 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13324 (aarch64_vectorize_vec_perm_const): ...this new function.
13325 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13326 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13327 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13328 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13329 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13330 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13331 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13333 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13334 check for NEON modes.
13335 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13336 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13337 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13338 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13340 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13341 the old VEC_PERM_CONST conditions.
13342 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13343 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13344 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13345 (ia64_vectorize_vec_perm_const_ok): Merge into...
13346 (ia64_vectorize_vec_perm_const): ...this new function.
13347 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13348 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13349 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13350 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13351 * config/mips/mips.c (mips_expand_vec_perm_const)
13352 (mips_vectorize_vec_perm_const_ok): Merge into...
13353 (mips_vectorize_vec_perm_const): ...this new function.
13354 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13355 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13356 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13357 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13358 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13359 (rs6000_expand_vec_perm_const): Delete.
13360 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13362 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13363 (altivec_expand_vec_perm_const_le): Take each operand individually.
13364 Operate on constant selectors rather than rtxes.
13365 (altivec_expand_vec_perm_const): Likewise. Update call to
13366 altivec_expand_vec_perm_const_le.
13367 (rs6000_expand_vec_perm_const): Delete.
13368 (rs6000_vectorize_vec_perm_const_ok): Delete.
13369 (rs6000_vectorize_vec_perm_const): New function.
13370 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13371 an element count and rtx array.
13372 (rs6000_expand_extract_even): Update call accordingly.
13373 (rs6000_expand_interleave): Likewise.
13374 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13375 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13376 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13377 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13378 (rs6000_expand_vec_perm_const): Delete.
13379 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13380 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13381 (altivec_expand_vec_perm_const_le): Take each operand individually.
13382 Operate on constant selectors rather than rtxes.
13383 (altivec_expand_vec_perm_const): Likewise. Update call to
13384 altivec_expand_vec_perm_const_le.
13385 (rs6000_expand_vec_perm_const): Delete.
13386 (rs6000_vectorize_vec_perm_const_ok): Delete.
13387 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13388 reference to the SPE evmerge intructions.
13389 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13390 an element count and rtx array.
13391 (rs6000_expand_extract_even): Update call accordingly.
13392 (rs6000_expand_interleave): Likewise.
13393 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13394 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13396 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13398 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13400 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13401 vector mode and that that mode matches the mode of the data
13403 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13404 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13405 directly using expand_vec_perm_1 when forcing selectors into
13407 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13409 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13411 * optabs-query.h (can_vec_perm_p): Delete.
13412 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13413 * optabs-query.c (can_vec_perm_p): Split into...
13414 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13415 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13416 particular selector is valid.
13417 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13418 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13419 (vect_grouped_load_supported): Likewise.
13420 (vect_shift_permute_load_chain): Likewise.
13421 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13422 (vect_transform_slp_perm_load): Likewise.
13423 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13424 (vectorizable_bswap): Likewise.
13425 (vect_gen_perm_mask_checked): Likewise.
13426 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13427 implementations of variable permutation vectors into account
13428 when deciding which selector to use.
13429 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13430 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13431 with a false third argument.
13432 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13433 to test whether the constant selector is valid and can_vec_perm_var_p
13434 to test whether a variable selector is valid.
13436 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13438 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13439 * optabs-query.c (can_vec_perm_p): Likewise.
13440 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13441 instead of vec_perm_indices.
13442 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13443 (vect_gen_perm_mask_checked): Likewise,
13444 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13445 (vect_gen_perm_mask_checked): Likewise,
13447 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13449 * optabs-query.h (qimode_for_vec_perm): Declare.
13450 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13451 (qimode_for_vec_perm): ...this new function.
13452 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13454 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13456 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13457 does not have a conditional at the top.
13459 2018-01-02 Richard Biener <rguenther@suse.de>
13461 * ipa-inline.c (big_speedup_p): Fix expression.
13463 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13466 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13469 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13473 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13474 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13475 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13476 cond_taken_branch_cost 3->4.
13478 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13480 PR tree-optimization/83581
13481 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13482 TODO_cleanup_cfg if any changes have been made.
13484 PR middle-end/83608
13485 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13486 convert_modes if target mode has the right side, but different mode
13489 PR middle-end/83609
13490 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13491 last argument when extracting from CONCAT. If either from_real or
13492 from_imag is NULL, use expansion through memory. If result is not
13493 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13494 the parts directly to inner mode, if even that fails, use expansion
13497 PR middle-end/83623
13498 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13499 check for bswap in mode rather than HImode and use that in expand_unop
13502 Copyright (C) 2018 Free Software Foundation, Inc.
13504 Copying and distribution of this file, with or without modification,
13505 are permitted in any medium without royalty provided the copyright
13506 notice and this notice are preserved.
projects / gcc.git / blob