1 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3 * config/nds32/nds32-md-auxiliary.c
4 (nds32_valid_smw_lwm_base_p): Refine.
5 (nds32_output_smw_single_word): Refine.
6 (nds32_output_smw_double_word): New.
7 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
9 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
12 (nds32_output_stack_pop): Refine.
13 (nds32_expand_unaligned_load): Refine.
14 (nds32_expand_unaligned_store): Refine.
16 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
17 Chung-Ju Wu <jasonwucj@gmail.com>
19 * config/nds32/constants.md: Add TP_REGNUM constant.
20 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
21 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
23 * config/nds32/nds32-doubleword.md: Consider flag_pic.
24 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
25 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
26 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
27 and PIC code generation.
28 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
30 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
32 * config/nds32/nds32.md: Support TLS and PIC.
33 * config/nds32/nds32.c: Support TLS and PIC.
34 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
35 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
38 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
40 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
43 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
44 Chung-Ju Wu <jasonwucj@gmail.com>
46 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
47 * config/nds32/nds32-md-auxiliary.c
48 (symbolic_reference_mentioned_p): New.
49 (nds32_legitimize_ict_address): New.
50 (nds32_expand_ict_move): New.
51 (nds32_indirect_call_referenced_p): New.
52 (nds32_symbol_binds_local_p): Delete.
53 (nds32_long_call_p): Modify.
54 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
55 * config/nds32/nds32-protos.h
56 (symbolic_reference_mentioned_p): Declare.
57 (nds32_legitimize_ict_address): Declare.
58 (nds32_expand_ict_move): Declare.
59 (nds32_indirect_call_referenced_p): Declare.
60 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
61 (nds32_relax_group): Use nds32_ict_const_p as condition.
62 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
63 (nds32_asm_file_start): Output ict_model directive in asm code.
64 (nds32_legitimate_address_p): Consider indirect call.
65 (nds32_print_operand): Consider indirect call.
66 (nds32_print_operand_address): Consider indirect call.
67 (nds32_insert_attributes): Handle "indirect_call" attribute.
68 (TARGET_LEGITIMATE_ADDRESS_P): Define.
69 (TARGET_LEGITIMATE_CONSTANT_P): Define.
70 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
71 (TARGET_DELEGITIMIZE_ADDRESS): Define.
72 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
73 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
74 (TARGET_ICT_MODEL_SMALL): Define.
75 (TARGET_ICT_MODEL_LARGE): Define.
76 * config/nds32/nds32.md (movsi): Consider ict model.
77 (call, call_value): Consider ict model.
78 (sibcall, sibcall_value): Consider ict model.
79 * config/nds32/nds32.opt (mict-model): New option.
80 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
83 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
84 Monk Chiang <sh.chiang04@gmail.com>
85 Jim Wilson <jimw@sifive.com>
87 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
88 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
89 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
90 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
91 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
92 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
93 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
94 compute save_libcall_adjustment properly.
95 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
96 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
97 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
98 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
99 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
100 (ABI_SPEC): Handle mabi=ilp32e.
101 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
103 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
104 <-march>: Add rv32e as an example.
106 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
109 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
110 (intra_create_variable_infos): Handle C++ constructors.
112 2018-05-18 Martin Liska <mliska@suse.cz>
114 * passes.def: Remove a redundant pass.
116 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
119 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
121 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
123 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
125 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
127 (armv2, armv2a, armv3, armv3m): Delete architectures.
128 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
129 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
130 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
132 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
133 (*mulsidi3adddi): Likewise.
134 (mulsidi3): Likewise.
135 (*mulsidi3_nov6): Likewise.
136 (umulsidi3): Likewise.
137 (umulsidi3_nov6): Likewise.
138 (umaddsidi4): Likewise.
139 (*umulsidi3adddi): Likewise.
140 (smulsi3_highpart): Likewise.
141 (*smulsi3_highpart_nov6): Likewise.
142 (umulsi3_highpart): Likewise.
143 (*umulsi3_highpart_nov6): Likewise.
144 * config/arm/arm.h (arm_arch3m): Delete.
145 * config/arm/arm.c (arm_arch3m): Delete.
146 (arm_option_override_internal): Update armv3-related comment.
147 (arm_configure_build_target): Delete use of isa_bit_mode32.
148 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
149 (arm_rtx_costs_internal): Delete check of arm_arch3m.
150 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
153 * config/arm/arm-protos.h (arm_arch3m): Delete.
154 * config/arm/arm-tables.opt: Regenerate.
155 * config/arm/arm-tune.md: Likewise.
156 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
157 deleted architectures.
159 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
161 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
162 (armv5t, armv5te): New features.
163 (ARMv5, ARMv5e): Delete fgroups.
164 (ARMv5t, ARMv5te): Adjust for above changes.
166 (armv5, armv5e): Delete arches.
167 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
169 (*call_reg_arm): Likewise.
170 (*call_value_reg_armv5): Likewise.
171 (*call_value_reg_arm): Likewise.
172 (*call_symbol): Likewise.
173 (*call_value_symbol): Likewise.
174 (*sibcall_insn): Likewise.
175 (*sibcall_value_insn): Likewise.
177 (prefetch): Likewise.
178 (define_split and define_peephole2 dependent on arm_arch5):
180 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
182 (TARGET_ARM_QBIT): Likewise.
183 (TARGET_DSP_MULTIPLY): Likewise.
184 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
185 (arm_arch5, arm_arch5e): Delete.
186 (arm_arch5t, arm_arch5te): Declare.
187 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
188 (arm_arch5t): Declare.
189 (arm_option_reconfigure_globals): Update for the above.
190 (arm_options_perform_arch_sanity_checks): Update comment, replace
191 use of arm_arch5 with arm_arch5t.
192 (use_return_insn): Likewise.
193 (arm_emit_call_insn): Likewise.
194 (output_return_instruction): Likewise.
195 (arm_final_prescan_insn): Likewise.
196 (arm_coproc_builtin_available): Likewise.
197 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
198 arm_arch5e with arm_arch5t and arm_arch5te.
199 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
200 (arm_arch5t, arm_arch5te): Declare.
201 * config/arm/arm-tables.opt: Regenerate.
202 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
203 * config/arm/t-multilib: Likewise.
204 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
205 instead of arm_arch5.
206 (*call_reg_thumb1): Likewise.
207 (*call_value_reg_thumb1_v5): Likewise.
208 (*call_value_reg_thumb1): Likewise.
209 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
211 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
213 2018-05-18 Martin Liska <mliska@suse.cz>
215 PR gcov-profile/84846
216 * doc/gcov.texi: Document -t option of gcov tool.
218 2018-05-18 Martin Liska <mliska@suse.cz>
220 PR gcov-profile/84846
221 * gcov.c (print_usage): Add new -t option.
222 (process_args): Handle the option.
223 (generate_results): Use stdout as output when requested by
226 2018-05-18 Martin Liska <mliska@suse.cz>
228 PR gcov-profile/84846
229 * coverage.c (coverage_init): Write PWD to .gcno file.
230 * doc/gcov.texi: Document how working directory is printed.
231 * gcov-dump.c (dump_gcov_file): Print PWD.
232 * gcov.c (output_intermediate_file): Likewise.
233 (read_graph_file): Read PWD string.
234 (output_lines): Print PWD.
236 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
239 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
240 for retval and return false if all args to phi are zero.
242 2018-05-18 Richard Biener <rguenther@suse.de>
244 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
246 (evrp_dom_walker::before_dom_children): Call it.
248 2018-05-18 Richard Biener <rguenther@suse.de>
250 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
251 results when processing array refs with variable index.
253 2018-05-18 Toon Moene <toon@moene.org>
255 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
256 directly after that of -floop-interchange. Indicate that both
257 options are enabled by default when specifying -O3.
259 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
261 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
262 iterator. Delete separate integer-mode vec_set<mode> expander.
263 (aarch64_simd_vec_setv2di): Delete.
264 (vec_setv2di): Delete.
265 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
266 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
267 the "w, r" alternative.
269 2018-05-18 Martin Liska <mliska@suse.cz>
271 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
272 * tree-pass.h (make_pass_lower_switch_O0): New function.
273 * tree-switch-conversion.c (node_has_low_bound): Remove.
274 (node_has_high_bound): Likewise.
275 (node_is_bounded): Likewise.
276 (class pass_lower_switch): Make it a template type and create
278 (pass_lower_switch::execute): Add template argument.
279 (make_pass_lower_switch): New function.
280 (make_pass_lower_switch_O0): New function.
281 (do_jump_if_equal): Remove.
282 (emit_case_nodes): Simplify to just handle all 3 cases and leave
283 all the hard work to tree optimization passes.
285 2018-05-18 Martin Liska <mliska@suse.cz>
287 * dbgcnt.c (limit_low): Renamed from limit.
288 (limit_high): New variable.
289 (dbg_cnt_is_enabled): Check for upper limit.
290 (dbg_cnt): Adjust dumping.
291 (dbg_cnt_set_limit_by_index): Add new argument for high
293 (dbg_cnt_set_limit_by_name): Likewise.
294 (dbg_cnt_process_single_pair): Parse new format.
295 (dbg_cnt_process_opt): Use strtok.
296 (dbg_cnt_list_all_counters): Remove 'value' and add
298 * doc/invoke.texi: Document changes.
300 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
302 * doc/sourcebuild.texi (scalar_all_fma): Document.
303 * tree.def (FMA_EXPR): Delete.
304 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
305 * internal-fn.c (ternary_direct): New macro.
306 (expand_ternary_optab_fn): Likewise.
307 (direct_ternary_optab_supported_p): Likewise.
308 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
309 * builtins.c (fold_builtin_fma): Delete.
310 (fold_builtin_3): Don't call it.
311 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
312 * expr.c (expand_expr_real_2): Likewise.
313 * fold-const.c (operand_equal_p): Likewise.
314 (fold_ternary_loc): Likewise.
315 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
316 * gimple.c (DEFTREECODE): Likewise.
317 * gimplify.c (gimplify_expr): Likewise.
318 * optabs-tree.c (optab_for_tree_code): Likewise.
319 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
320 * tree-eh.c (operation_could_trap_p): Likewise.
321 (stmt_could_throw_1_p): Likewise.
322 * tree-inline.c (estimate_operator_cost): Likewise.
323 * tree-pretty-print.c (dump_generic_node): Likewise.
324 (op_code_prio): Likewise.
325 * tree-ssa-loop-im.c (stmt_cost): Likewise.
326 * tree-ssa-operands.c (get_expr_operands): Likewise.
327 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
328 * fold-const-call.h (fold_fma): Delete.
329 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
330 CFN_FNMA and CFN_FNMS.
332 * genmatch.c (combined_fn): New enum.
333 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
334 (commutative_op): New function.
335 (commutate): Use it. Handle more than 2 operands.
336 (dt_operand::gen_gimple_expr): Use commutative_op.
337 (parser::parse_expr): Allow :c to be used with non-binary
338 operators if the commutative operand is known.
339 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
340 CFN_FMS, CFN_FNMA and CFN_FNMS.
341 (backprop::process_assign_use): Remove FMA_EXPR handling.
342 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
343 (gen_hsa_fma): New function.
344 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
345 IFN_FNMA and IFN_FNMS.
346 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
347 * gimple-fold.h (follow_all_ssa_edges): Declare.
348 * gimple-fold.c (follow_all_ssa_edges): New function.
349 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
350 gimple_build interface and use follow_all_ssa_edges to fold the result.
351 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
352 instead of checking for optabs directly.
353 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
354 rather than FMA_EXPRs.
355 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
356 call to IFN_FMA instead of an FMA_EXPR.
358 2018-05-17 Jim Wilson <jimw@sifive.com>
360 * expr.c (do_tablejump): When converting index to Pmode, if we have a
361 sign extended promoted subreg, and the range does not have the sign bit
362 set, then do a sign extend.
364 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
365 test, check for sign extended subreg and/or constant operands, and
366 do a sign extend in that case.
368 2018-05-17 Steve Ellcey <sellcey@cavium.com>
370 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
371 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
373 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
374 Change logics_shift_reg to logics_shift_imm.
375 (thunderx2t99_fp_loadpair_basic): Delete.
376 (thunderx2t99_fp_storepair_basic): Delete.
377 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
378 (thunderx2t99_asimd_polynomial): Delete.
379 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
380 and neon_fp_mul_d_scalar_q.
381 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
382 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
383 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
384 (thunderx2t99_asimd_lut): Add missing tbl types.
385 (thunderx2t99_asimd_ext): Delete.
386 (thunderx2t99_asimd_load1_1_mult): Delete.
387 (thunderx2t99_asimd_load1_2_mult): Delete.
388 (thunderx2t99_asimd_load1_ldp): New.
389 (thunderx2t99_asimd_load1): New.
390 (thunderx2t99_asimd_load2): Add missing *load2* types.
391 (thunderx2t99_asimd_load3): New.
392 (thunderx2t99_asimd_load4): New.
393 (thunderx2t99_asimd_store1_1_mult): Delete.
394 (thunderx2t99_asimd_store1_2_mult): Delete.
395 (thunderx2t99_asimd_store2_mult): Delete.
396 (thunderx2t99_asimd_store2_onelane): Delete.
397 (thunderx2t99_asimd_store_stp): New.
398 (thunderx2t99_asimd_store1): New.
399 (thunderx2t99_asimd_store2): New.
400 (thunderx2t99_asimd_store3): New.
401 (thunderx2t99_asimd_store4): New.
403 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
405 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
406 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
408 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
409 Segher Boessenkool <segher@kernel.crashing.org>
412 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
415 2018-05-17 Richard Biener <rguenther@suse.de>
417 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
418 for pruning loop and prune defs feeding only already visited PHIs.
420 2018-05-17 Richard Biener <rguenther@suse.de>
422 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
424 2018-05-17 Bin Cheng <bin.cheng@arm.com>
425 Richard Biener <rguenther@suse.de>
427 PR tree-optimization/85793
428 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
429 for VMAT_ELEMENTWISE.
431 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
433 * internal-fn.h (lookup_internal_fn): Declare
434 * internal-fn.c (lookup_internal_fn): New function.
435 * gimple.c (gimple_build_call_from_tree): Handle calls to
437 * gimple-pretty-print.c (dump_gimple_call): Print "." before
438 internal function names.
439 * tree-pretty-print.c (dump_generic_node): Likewise.
440 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
442 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
444 * gimple-fold.h (gimple_build): Make the function forms take
445 combined_fn rather than built_in_function.
446 (gimple_simplify): Likewise.
447 * gimple-match-head.c (gimple_simplify): Likewise.
448 * gimple-fold.c (gimple_build): Likewise.
449 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
450 rather than gimple_build_call_internal.
451 (get_initial_defs_for_reduction): Likewise.
452 (vect_create_epilog_for_reduction): Likewise.
453 (vectorizable_live_operation): Likewise.
455 2018-05-17 Martin Liska <mliska@suse.cz>
457 * gimple-ssa-sprintf.c (format_directive): Do not use
458 space in between 'G_' and '('.
460 2018-05-17 Jakub Jelinek <jakub@redhat.com>
463 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
464 even if the mask is not all ones.
467 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
469 (ix86_gimple_fold_builtin): Likewise.
472 * config/i386/i386.c: Include tree-vector-builder.h.
473 (ix86_vector_shift_count): New function.
474 (ix86_fold_builtin): Fold shift builtins by scalar count.
475 (ix86_gimple_fold_builtin): Likewise.
477 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
478 _mm512_setzero): New intrinsics.
480 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
481 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
483 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
484 code generation for cases where splatting a value is not useful.
485 * simplify-rtx.c (simplify_ternary_operation): Simplify
486 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
487 mode to a vec_concat.
489 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
491 * config.gcc: Support "goldmont-plus".
492 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
494 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
495 PROCESSOR_GOLDMONT_PLUS.
496 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
497 (processor_target_table): Add "goldmont-plus".
498 (PTA_GOLDMONT_PLUS): Define.
499 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
500 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
501 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
502 (fold_builtin_cpu): Add "goldmont-plus".
503 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
504 (ix86_option_override_internal): Add "goldmont-plus".
505 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
506 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
507 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
508 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
510 2018-05-17 Richard Biener <rguenther@suse.de>
512 PR tree-optimization/85757
513 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
514 remove defs that only feed that PHI from further processing.
516 2018-05-16 Jim Wilson <jimw@sifive.com>
518 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
520 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
522 2018-05-16 Mark Wielaard <mark@klomp.org>
524 * dwarf2out.c (count_index_strings): New function.
525 (output_indirect_strings): Call count_index_strings and generate
526 header for dwarf_version >= 5.
528 2018-05-16 Mark Wielaard <mark@klomp.org>
530 * dwarf2out.c (dwarf_FORM): New function.
531 (set_indirect_string): Use dwarf_FORM.
532 (reset_indirect_string): Likewise.
533 (size_of_die): Likewise.
534 (value_format): Likewise.
535 (output_die): Likewise.
536 (add_skeleton_AT_string): Likewise.
537 (output_macinfo_op): Likewise.
538 (index_string): Likewise.
539 (output_index_string_offset): Likewise.
540 (output_index_string): Likewise.
541 (count_index_strings): Likewise.
543 2018-05-16 Carl Love <cel@us.ibm.com>
545 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
546 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
548 2018-05-16 Martin Jambor <mjambor@suse.cz>
550 * ipa-prop.c (ipa_free_all_edge_args): Remove.
551 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
553 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
555 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
556 (fnma<mode>4): Likewise.
557 (fms<mode>4): Likewise.
558 (fnms<mode>4): Likewise.
559 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
560 (aarch64_fnma<mode>4): Likewise.
561 (aarch64_fms<mode>4): Likewise.
562 (aarch64_fnms<mode>4): Likewise.
563 (aarch64_fnmadd<mode>4): Likewise.
565 2018-05-16 Jason Merrill <jason@redhat.com>
567 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
569 2018-05-16 Richard Biener <rguenther@suse.de>
571 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
572 (dump_stmt_cost): Declare.
573 (add_stmt_cost): Dump cost we add.
574 (add_stmt_costs): New function.
575 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
577 (vect_analyze_stmt): Adjust prototype.
578 (vectorizable_condition): Likewise.
579 (vectorizable_live_operation): Likewise.
580 (vectorizable_reduction): Likewise.
581 (vectorizable_induction): Likewise.
582 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
583 cost vector to pass to vectorizable_ and record afterwards.
584 (vect_model_reduction_cost): Take cost vector argument and adjust.
585 (vect_model_induction_cost): Likewise.
586 (vectorizable_reduction): Likewise.
587 (vectorizable_induction): Likewise.
588 (vectorizable_live_operation): Likewise.
589 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
590 SLP_TREE_NUMBER_OF_VEC_STMTS.
591 (vect_analyze_slp_cost_1): Remove.
592 (vect_analyze_slp_cost): Likewise.
593 (vect_slp_analyze_node_operations): Take visited args and
594 a target cost vector. Avoid processing already visited stmt sets.
595 (vect_slp_analyze_operations): Use a local cost vector to gather
596 costs and register those of non-discarded instances.
597 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
598 (vect_schedule_slp_instance): Remove copying of
599 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
601 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
602 adding cost. Record cost entry location.
603 (vect_prologue_cost_for_slp_op): Function to compute cost of
604 a constant or invariant generated for SLP vect in the prologue,
605 split out from vect_analyze_slp_cost_1.
606 (vect_model_simple_cost): Make static. Adjust for SLP costing.
607 (vect_model_promotion_demotion_cost): Likewise.
608 (vect_model_store_cost): Likewise, make static.
609 (vect_model_load_cost): Likewise.
610 (vectorizable_bswap): Add cost vector arg and adjust.
611 (vectorizable_call): Likewise.
612 (vectorizable_simd_clone_call): Likewise.
613 (vectorizable_conversion): Likewise.
614 (vectorizable_assignment): Likewise.
615 (vectorizable_shift): Likewise.
616 (vectorizable_operation): Likewise.
617 (vectorizable_store): Likewise.
618 (vectorizable_load): Likewise.
619 (vectorizable_condition): Likewise.
620 (vectorizable_comparison): Likewise.
621 (can_vectorize_live_stmts): Likewise.
622 (vect_analyze_stmt): Likewise.
623 (vect_transform_stmt): Adjust calls to vectorizable_*.
624 * tree-vectorizer.c: Include gimple-pretty-print.h.
625 (dump_stmt_cost): New function.
627 2018-05-16 Richard Biener <rguenther@suse.de>
629 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
630 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
631 * tree-ssa-dse.c: Include tree-ssa-loop.h.
632 (check_name): New callback.
633 (dse_classify_store): Track cycles via a visited bitmap of PHI
634 defs and simplify handling of in-loop and across loop dead stores
635 and properly fail for loop-variant refs. Handle byte-tracking with
636 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
639 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
641 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
642 (vect_get_mask_type_for_stmt): Likewise.
643 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
645 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
646 to determine the statement's vector type and the vector type that
647 should be used for calculating nunits. Deal with cases in which
648 the type has to be deferred.
649 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
650 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
651 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
652 (vect_determine_vf_for_stmt): New functions, split out from...
653 (vect_determine_vectorization_factor): ...here.
654 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
655 (vect_get_mask_type_for_stmt): New functions, split out from
656 vect_determine_vectorization_factor.
658 2018-05-16 Richard Biener <rguenther@suse.de>
660 * tree-cfg.c (verify_gimple_assign_ternary): Properly
661 verify the [VEC_]COND_EXPR embedded comparison.
663 2018-05-15 Martin Sebor <msebor@redhat.com>
665 PR tree-optimization/85753
666 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
667 RECORD_TYPE in addition to ARRAY_TYPE.
669 2018-05-15 Martin Sebor <msebor@redhat.com>
672 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
674 2018-05-15 Richard Biener <rguenther@suse.de>
676 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
677 add by_clobber_p one. Change algorithm to collect all defs
678 representing uses we need to walk and try reducing them to
679 a single one before failing.
680 (dse_dom_walker::dse_optimize_stmt): Adjust.
682 2018-05-13 Mark Wielaard <mark@klomp.org>
684 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
685 (size_of_loc_descr): Likewise.
686 (output_loc_operands): Likewise.
687 (output_loc_operands_raw): Likewise.
688 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
689 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
690 (hash_loc_operands): Likewise.
691 (compare_loc_operands): Likewise.
693 2018-05-14 Mark Wielaard <mark@klomp.org>
695 * dwarf2out.c (count_index_addrs): New function.
696 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
698 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
700 PR tree-optimization/83648
701 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
702 return value as malloc candidate.
704 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
707 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
708 as true in call to suggest_attribute.
710 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
712 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
715 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
717 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
719 (aarch64_crypto_aesd_fused): Likewise.
721 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
723 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
724 (movsi_aarch64): Likewise.
725 (load_pairsi): Likewise.
726 (load_pairdi): Likewise.
727 (store_pairsi): Likewise.
728 (store_pairdi): Likewise.
729 (load_pairsf): Likewise.
730 (load_pairdf): Likewise.
731 (store_pairsf): Likewise.
732 (store_pairdf): Likewise.
733 (zero_extend): Likewise.
734 (trunc): Swap alternatives.
735 (fcvt_target): Add '?' to prefer w over r.
737 2018-05-14 Jakub Jelinek <jakub@redhat.com>
740 * config/i386/i386.md: Disallow non-commutative arithmetics in
741 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
742 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
743 in the peephole2 before it.
745 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
747 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
748 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
749 (ix86_handle_option): Handle -mcldemote.
750 * config.gcc: New header.
751 * config/i386/cldemoteintrin.h: New file.
752 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
753 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
755 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
756 OPTION_MASK_ISA_CLDEMOTE.
757 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
758 (ix86_valid_target_attribute_inner_p): Ditto.
759 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
760 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
761 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
762 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
763 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
765 * config/i386/i386.opt: Add -mcldemote.
766 * config/i386/x86intrin.h: New header.
767 * doc/invoke.texi: Add -mcldemote.
769 2018-05-14 Richard Biener <rguenther@suse.de>
771 * doc/match-and-simplify.texi: Adjust :s documentation.
773 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
775 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
776 intended memcpy size.
777 (REORDER_45): Likewise.
779 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
782 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
783 * vec.c (qsort_chk): Use gcc_qsort.
784 * Makefile.in (OBJS-libcommon): Add sort.o.
785 (build/sort.o): New target. Use it...
786 (BUILD_RTL): ... here, and...
787 (build/gencfn-macros): ... here, and...
788 (build/genmatch): ... here.
790 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
791 Chung-Ju Wu <jasonwucj@gmail.com>
793 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
794 * config/nds32/nds32-graywolf.md: New file.
795 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
796 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
798 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
799 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
800 * config/nds32/nds32.md (pipeline_model): Add graywolf.
801 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
802 * config/nds32/pipelines.md: Include n15 settings.
804 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
805 Chung-Ju Wu <jasonwucj@gmail.com>
807 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
808 * config/nds32/nds32-n13.md: New file.
809 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
810 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
812 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
813 * config/nds32/nds32.md (pipeline_model): Add n13.
814 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
815 * config/nds32/pipelines.md: Include n13 settings.
817 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
818 Chung-Ju Wu <jasonwucj@gmail.com>
820 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
821 * config/nds32/nds32-n10.md: New file.
822 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
823 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
825 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
826 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
827 * config/nds32/nds32.md (pipeline_model): Add n10.
828 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
829 * config/nds32/pipelines.md: Include n10 settings.
831 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
832 Kito Cheng <kito.cheng@gmail.com>
833 Chung-Ju Wu <jasonwucj@gmail.com>
835 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
836 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
837 Add enum values for DSP extension instructions.
838 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
840 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
841 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
843 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
844 * config/nds32/nds32-dspext.md: New file for DSP implementation.
845 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
846 * config/nds32/nds32-intrinsic.md: Likewise.
847 * config/nds32/nds32_intrinsic.h: Likewise.
848 * config/nds32/nds32-md-auxiliary.c: Likewise.
849 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
850 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
851 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
852 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
853 * config/nds32/nds32-protos.h: New declarations for DSP extension.
854 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
855 TYPE_DMAC in switch statement.
856 * config/nds32/nds32.c: New checking and implementation for DSP
857 extension instructions.
858 * config/nds32/nds32.h: Likewise.
859 * config/nds32/nds32.md: Likewise.
860 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
861 * config/nds32/predicates.md: Implement new predicates for DSP
864 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
866 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
867 Reformat alternatives and attributes so it is easier to identify
868 which constraints/attributes go with which instruction.
869 (mov<mode>_hardfloat32, FMOVE64): Likewise.
870 (mov<mode>_softfloat32, FMOVE64): Likewise.
871 (mov<mode>_hardfloat64, FMOVE64): Likewise.
872 (mov<mode>_softfloat64, FMOVE64): Likewise.
874 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
876 * doc/extend.texi (PowerPC Built-in Functions): Rename this
878 (Basic PowerPC Built-in Functions): The new name of the
879 subsection previously known as "PowerPC Built-in Functions".
880 (Basic PowerPC Built-in Functions Available on all Configurations):
882 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
883 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
884 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
885 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
887 2018-05-11 Martin Jambor <mjambor@suse.cz>
890 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
893 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
896 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
898 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
900 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
901 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
902 (ix86_handle_option): Handle -mwaitpkg.
903 * config.gcc: New header.
904 * config/i386/cpuid.h (bit_WAITPKG): New bit.
905 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
906 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
908 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
909 OPTION_MASK_ISA_WAITPKG.
910 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
911 (ix86_option_override_internal): Add PTA_WAITPKG.
912 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
913 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
915 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
916 __builtin_ia32_umwait and __builtin_ia32_tpause.
917 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
918 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
919 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
920 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
921 UNSPECV_TPAUSE): New.
922 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
923 * config/i386/i386.opt: Add -mwaitpkg.
924 * config/i386/waitpkgintrin.h: New file.
925 * config/i386/x86intrin.h: New header.
926 * doc/invoke.texi: Add -mwaitpkg.
928 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
931 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
933 (cortex-m0): Use armv6s-m isa.
934 (cortex-m0plus): Likewise.
935 (cortex-m1): Likewise.
936 (cortex-m0.small-multiply): Likewise.
937 (cortex-m0plus.small-multiply): Likewise.
938 (cortex-m1.small-multiply): Likewise.
940 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
941 Jakub Jelinek <jakub@redhat.com>
943 PR tree-optimization/85692
944 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
945 source permute as well.
947 2018-05-11 Martin Liska <mliska@suse.cz>
950 * doc/extend.texi: Document LLVM style format for no_sanitize
953 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
955 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
956 mode_supports_vsx_dform_quad to mode_supports_dq_form.
957 (mode_supports_vsx_dform_quad): Likewise.
958 (mode_supports_vmx_dform): Move these functions to be next to the
959 other mode_supports functions.
960 (mode_supports_dq_form): Likewise.
961 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
962 mode_supports_dq_form.
963 (reg_offset_addressing_ok_p): Likewise.
964 (offsettable_ok_by_alignment): Likewise.
965 (rs6000_legitimate_offset_address_p): Likewise.
966 (legitimate_lo_sum_address_p): Likewise.
967 (rs6000_legitimize_address): Likewise.
968 (rs6000_legitimize_reload_address): Likewise.
969 (rs6000_secondary_reload_inner): Likewise.
970 (rs6000_preferred_reload_class): Likewise.
971 (rs6000_output_move_128bit): Likewise.
973 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
975 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
976 Generate SImode target register for null target.
977 <case IX86_BUILTIN_XGETBV>: Ditto.
978 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
979 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
981 2018-05-10 Carl Love <cel@us.ibm.com>
983 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
984 dcbtt and dcbtstt if operands[2] is 0.
986 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
989 * config/i386/sse.md (usadv64qi): New expander.
991 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
993 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
994 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
995 -maltivec=be support.
996 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
997 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
998 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
999 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1000 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1001 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1002 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1003 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1004 altivec_vsumsws): Adjust.
1005 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1006 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1007 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1009 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1010 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1011 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1012 (altivec_lve<VI_char>x): Delete expand.
1013 (*altivec_lve<VI_char>x_internal): Rename to...
1014 (altivec_lve<VI_char>x): ... this.
1015 (altivec_lvxl_<mode>): Delete expand.
1016 (*altivec_lvxl_<mode>_internal): Rename to ...
1017 (altivec_lvxl_<mode>): ... this.
1018 (altivec_stvxl_<mode>): Delete expand.
1019 (*altivec_stvxl_<mode>_internal): Rename to ...
1020 (altivec_stvxl_<mode>): ... this.
1021 (altivec_stve<VI_char>x): Delete expand.
1022 (*altivec_stve<VI_char>x_internal): Rename to ...
1023 (altivec_stve<VI_char>x): ... this.
1024 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1025 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1026 reduc_plus_scal_<mode>): Adjust.
1027 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1029 (rs6000_cpu_cpp_builtins): Adjust.
1030 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1031 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1032 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1033 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1034 -maltivec=be support.
1035 (rs6000_split_vec_extract_var): Adjust.
1036 (rs6000_split_v4si_init): Adjust.
1037 (swap_selector_for_mode): Delete.
1038 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1039 altivec_expand_stvex_be): Delete.
1040 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1041 -maltivec=be support.
1042 (rs6000_gimple_fold_builtin): Ditto.
1043 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1045 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1046 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1047 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1048 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1049 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1050 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1051 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1052 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1053 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1054 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1055 anonymous split): Adjust.
1056 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1057 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1059 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1061 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1062 when --with-gxx-include-dir is also specified.
1063 * configure: Regenerate.
1065 2018-05-09 Jim Wilson <jimw@sifive.com>
1068 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1069 * config/riscv/t-withmultilib: New.
1070 * config/riscv/withmultilib.h: New.
1071 * doc/install.texi: Document RISC-V --with-multilib-list support.
1073 2018-05-09 Richard Biener <rguenther@suse.de>
1075 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1077 (vect_bb_vectorization_profitable_p): Adjust. Compute
1078 actual scalar cost using the cost vector and the add_stmt_cost
1081 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1083 PR rtl-optimization/85645
1084 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1085 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1087 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1089 PR rtl-optimization/85645
1090 * shrink-wrap.c (spread_components): Return a boolean saying if
1091 anything was changed.
1092 (try_shrink_wrapping_separate): Iterate spread_components until
1093 nothing changes anymore.
1095 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1097 PR rtl-optimization/85645
1098 * regrename.c (build_def_use): Also kill the chains that include the
1099 destination of a REG_CFA_REGISTER note.
1101 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1103 PR rtl-optimization/85645
1104 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1105 insn that has a REG_CFA_REGISTER note.
1107 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1109 * cfgexpand.c (expand_clobber): New function.
1110 (expand_gimple_stmt_1): Use it.
1111 * tree-vect-stmts.c (vect_clobber_variable): New function,
1113 (vectorizable_simd_clone_call): ...here.
1114 (vectorizable_store): Emit a clobber either side of an
1115 IFN_STORE_LANES sequence.
1116 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1118 2018-05-09 Tom de Vries <tom@codesourcery.com>
1121 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1122 (define_insn "trap_if_false"): Add exit after trap.
1124 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1126 PR rtl-optimization/85638
1127 * bb-reorder.c: Include common/common-target.h.
1128 (create_forwarder_block): New function extracted from...
1129 (fix_up_crossing_landing_pad): ...here. Rename into...
1130 (dw2_fix_up_crossing_landing_pad): ...this.
1131 (sjlj_fix_up_crossing_landing_pad): New function.
1132 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1133 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1134 from both partitions and exit the loop after one iteration.
1136 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1139 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1141 (Basic PowerPC Built-in Functions): The new name of the
1142 subsection previously known as "PowerPC Built-in Functions".
1143 (Basic PowerPC Built-in Functions Available on all Configurations):
1145 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1147 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1149 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1151 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1154 2018-05-08 Jim Wilson <jimw@sifive.com>
1156 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1157 (LD_EMUL_SUFFIX): New.
1158 (LINK_SPEC): Use it.
1160 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1162 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1164 (Basic PowerPC Built-in Functions): The new name of the
1165 subsection previously known as "PowerPC Built-in Functions".
1166 (Basic PowerPC Built-in Functions Available on all Configurations):
1168 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1170 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1172 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1174 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1177 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1180 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1181 after cmpelim optimization.
1183 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1185 * config.gcc: Support "goldmont".
1186 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1187 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1189 * config/i386/i386.c (m_GOLDMONT): Define.
1190 (processor_target_table): Add "goldmont".
1191 (PTA_GOLDMONT): Define.
1192 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1193 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1194 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1195 (fold_builtin_cpu): Add "goldmont".
1196 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1197 (ix86_option_override_internal): Add "goldmont".
1198 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1199 (processor_type): Add PROCESSOR_GOLDMONT.
1200 * config/i386/i386.md: Add CPU "glm".
1201 * config/i386/glm.md: New file.
1202 * config/i386/x86-tune.def: Add m_GOLDMONT.
1203 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1205 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1208 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1210 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1211 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1212 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1215 * config/i386/i386.c (ix86_fold_builtin): Handle
1216 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1219 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1220 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1222 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1225 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1226 (check_arch): Likewise.
1227 (check_fpu): Return the result rather than printing it.
1228 (end arch): Fix operator precedence.
1229 (end cpu): Likewise.
1230 (END): Print the result from check_fpu.
1232 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1233 Alan Hayward <alan.hayward@arm.com>
1234 David Sherwood <david.sherwood@arm.com>
1236 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1237 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1238 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1239 (*fcmuo<mode>_and): New patterns.
1241 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1243 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1244 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1245 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1246 (cmp_op, sve_imm_con): New code attributes.
1247 (SVE_COND_INT_CMP, imm_con): Delete.
1248 (cmp_op): Remove above unspecs from int attribute.
1249 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1251 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1252 comparison-specific unspecs.
1253 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1254 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1255 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1256 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1257 (*vec_fcm<cmp_op><mode>): Rename to...
1258 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1259 (*vec_fcmuo<mode>): Rename to...
1260 (*fcmuo<mode>): ...this and adjust likewise.
1261 (*pred_fcm<cmp_op><mode>): New pattern.
1262 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1263 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1265 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1267 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1268 (aarch64_emit_sve_predicated_cond): New function.
1269 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1270 (aarch64_emit_unspec_cond_or): Replace with...
1271 (aarch64_emit_sve_or_conds): ...this new function. Use
1272 aarch64_emit_sve_ptrue_op for the individual comparisons and
1273 aarch64_emit_binop to OR them together.
1274 (aarch64_emit_inverted_unspec_cond): Replace with...
1275 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1276 aarch64_emit_sve_ptrue_op for the comparison and
1277 aarch64_emit_unop to invert the result.
1278 (aarch64_expand_sve_vec_cmp_float): Update after the above
1279 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1281 2018-05-07 Nathan Sidwell <nathan@acm.org>
1283 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1284 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1285 (Backwards Compatibility): Likewise.
1287 2018-05-07 Luis Machado <luis.machado@linaro.org>
1291 2018-05-07 Luis Machado <luis.machado@linaro.org>
1293 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1294 <prefetch_dynamic_strides>: New const bool field.
1295 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1296 prefetch_dynamic_strides.
1297 (exynosm1_prefetch_tune): Likewise.
1298 (thunderxt88_prefetch_tune): Likewise.
1299 (thunderx_prefetch_tune): Likewise.
1300 (thunderx2t99_prefetch_tune): Likewise.
1301 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1302 (aarch64_override_options_internal): Update to set
1303 PARAM_PREFETCH_DYNAMIC_STRIDES.
1304 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1305 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1306 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1307 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1308 prefetch-dynamic-strides setting.
1310 2018-05-07 Luis Machado <luis.machado@linaro.org>
1312 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1313 <minimum_stride>: New const int field.
1314 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1315 minimum_stride field.
1316 (exynosm1_prefetch_tune): Likewise.
1317 (thunderxt88_prefetch_tune): Likewise.
1318 (thunderx_prefetch_tune): Likewise.
1319 (thunderx2t99_prefetch_tune): Likewise.
1320 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1321 (aarch64_override_options_internal): Update to set
1322 PARAM_PREFETCH_MINIMUM_STRIDE.
1323 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1324 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1325 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1326 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1327 stride is constant and is below the minimum stride threshold.
1329 2018-05-07 Luis Machado <luis.machado@linaro.org>
1331 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1334 2018-05-07 Luis Machado <luis.machado@linaro.org>
1336 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1337 <prefetch_dynamic_strides>: New const bool field.
1338 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1339 prefetch_dynamic_strides.
1340 (exynosm1_prefetch_tune): Likewise.
1341 (thunderxt88_prefetch_tune): Likewise.
1342 (thunderx_prefetch_tune): Likewise.
1343 (thunderx2t99_prefetch_tune): Likewise.
1344 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1345 (aarch64_override_options_internal): Update to set
1346 PARAM_PREFETCH_DYNAMIC_STRIDES.
1347 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1348 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1349 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1350 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1351 prefetch-dynamic-strides setting.
1353 2018-05-07 Luis Machado <luis.machado@linaro.org>
1355 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1356 <minimum_stride>: New const int field.
1357 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1358 minimum_stride field.
1359 (exynosm1_prefetch_tune): Likewise.
1360 (thunderxt88_prefetch_tune): Likewise.
1361 (thunderx_prefetch_tune): Likewise.
1362 (thunderx2t99_prefetch_tune): Likewise.
1363 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1364 (aarch64_override_options_internal): Update to set
1365 PARAM_PREFETCH_MINIMUM_STRIDE.
1366 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1367 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1368 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1369 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1370 stride is constant and is below the minimum stride threshold.
1372 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1375 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1376 the type is addressable. Don't force op into register if it has
1379 2018-05-05 Roland McGrath <mcgrathr@google.com>
1382 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1383 any section for which we don't know a specific type it should have,
1384 regardless of name. Previously this was done only for the exact
1385 names ".init_array", ".fini_array", and ".preinit_array".
1386 (default_elf_asm_named_section): Add comment about
1387 relationship with default_section_type_flags and SECTION_NOTYPE.
1388 (get_section): Don't consider it a type conflict if one side has
1389 SECTION_NOTYPE and the other doesn't, as long as neither has the
1390 SECTION_BSS et al used in the default_section_type_flags logic.
1392 2018-05-05 Tom de Vries <tom@codesourcery.com>
1395 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1396 (workaround_barsyncs): New function.
1397 (nvptx_reorg): Use workaround_barsyncs.
1398 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1399 (define_expand "nvptx_membar_cta"): New define_expand.
1400 (define_insn "*nvptx_membar_cta"): New insn.
1402 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1404 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1405 To improve optimization opportunities.
1406 * builtin-types.def: The new needed builtin types for the above.
1408 2018-05-04 Richard Biener <rguenther@suse.de>
1410 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1411 * gimple-ssa-store-merging.c
1412 (imm_store_chain_info::output_merged_store): Remove redundant create,
1413 release split_store vector contents on failure.
1414 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1415 scalar stmt vector on cache hit.
1417 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1419 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1421 * config.gcc (powerpc-xilinx-eabi*): Remove.
1422 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1424 (fusion_addis_mem_combo_load): Ditto.
1425 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1427 (rs6000_cpu_cpp_builtins): Ditto.
1428 * config/rs6000/rs6000-linux.c
1429 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1430 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1431 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1433 (rs6000_setup_reg_addr_masks): Ditto.
1434 (rs6000_init_hard_regno_mode_ok): Ditto.
1435 (rs6000_option_override_internal): Ditto.
1436 (legitimate_lo_sum_address_p): Ditto.
1437 (rs6000_legitimize_address): Ditto.
1438 (rs6000_legitimize_reload_address): Ditto.
1439 (rs6000_legitimate_address_p): Ditto.
1440 (abi_v4_pass_in_fpr): Ditto.
1441 (setup_incoming_varargs): Ditto.
1442 (rs6000_gimplify_va_arg): Ditto.
1443 (rs6000_split_multireg_move): Ditto.
1444 (rs6000_savres_strategy): Ditto.
1445 (rs6000_emit_prologue_components): Ditto.
1446 (rs6000_emit_epilogue_components): Ditto.
1447 (rs6000_emit_prologue): Ditto.
1448 (rs6000_emit_epilogue): Ditto.
1449 (rs6000_elf_file_end): Ditto.
1450 (rs6000_function_value): Ditto.
1451 (rs6000_libcall_value): Ditto.
1452 * config/rs6000/rs6000.h: Ditto.
1453 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1454 (TARGET_MINMAX): ... this. New.
1455 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1456 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1457 (*movsi_internal1_single): Delete.
1458 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1459 mfpu=, mxilinx-fpu): Delete.
1460 * config/rs6000/singlefp.h: Delete.
1461 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1462 * config/rs6000/t-rs6000: Ditto.
1463 * config/rs6000/t-xilinx: Delete.
1464 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1465 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1466 (VStype_simple): Delete.
1467 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1468 * config/rs6000/xfpu.h: Delete.
1469 * config/rs6000/xfpu.md: Delete.
1470 * config/rs6000/xilinx.h: Delete.
1471 * config/rs6000/xilinx.opt: Delete.
1472 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1473 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1475 2018-05-04 Tom de Vries <tom@codesourcery.com>
1478 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1481 2018-05-04 Richard Biener <rguenther@suse.de>
1484 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1485 (expand_complex_div_wide): Likewise.
1486 (expand_complex_operations_1): Likewise.
1487 (expand_complex_libcall): Preserve EH info of the original stmt.
1488 (tree_lower_complex): Handle removed blocks.
1489 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1490 on complex multiplication and division libcall builtins.
1492 2018-05-04 Richard Biener <rguenther@suse.de>
1495 * fold-const.c (negate_expr_p): Restrict negation of operand
1496 zero of a division to when we know that can happen without
1498 (fold_negate_expr_1): Likewise.
1500 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1503 * real.h (real_nextafter): Declare.
1504 * real.c (real_nextafter): New function.
1505 * fold-const-call.c (fold_const_nextafter): New function.
1506 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1507 CASE_CFN_NEXTTOWARD.
1508 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1509 even when arg1_mode is different from arg0_mode.
1511 2018-05-03 Nathan Sidwell <nathan@acm.org>
1513 * doc/extend.texi (Deprecated Features): Remove
1515 (Backwards Compatibility): Likewise.
1516 * doc/invoke.texi (C++ Language Options): Likewise.
1517 (C++ Dialect Options): Likewise.
1519 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1522 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1523 _mm512_mask_mullox_epi64): New intrinsics.
1525 2018-05-03 Tom de Vries <tom@codesourcery.com>
1528 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1529 dump files): Add offload-tree.
1531 2018-05-03 Richard Biener <rguenther@suse.de>
1533 PR tree-optimization/85615
1534 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1535 to loops not nested in BBs loop father to avoid creating multi-entry
1538 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1540 PR tree-optimization/70291
1541 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1542 arguments. Change return type to tree. Emit libcall as a new
1543 statement rather than replacing existing one when inplace_p is true.
1544 (expand_complex_multiplication_components): New function.
1545 (expand_complex_multiplication): Expand floating-point complex
1546 multiplication using the above.
1547 (expand_complex_division): Rename inner_type parameter to type.
1548 Update expand_complex_libcall call-site.
1549 (expand_complex_operations_1): Update expand_complex_multiplication
1550 and expand_complex_division call-sites.
1552 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1555 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1556 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1557 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1558 the highest significant bit of the shift count mask is clear. In
1559 check whether and[sq]i3 is needed verify that all significant bits
1560 of the shift count other than the highest are set.
1562 2018-05-02 Tom de Vries <tom@codesourcery.com>
1565 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1566 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1567 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1568 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1569 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1570 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1571 __builtin_goacc_parlevel_size.
1573 2018-05-02 Richard Biener <rguenther@suse.de>
1575 PR tree-optimization/85597
1576 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1577 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1580 2018-05-02 Tom de Vries <tom@codesourcery.com>
1583 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1584 dump files): Add ltrans-tree.
1586 2018-05-02 Tom de Vries <tom@codesourcery.com>
1589 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1590 dump files): Add wpa-ipa.
1592 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1594 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1595 powerpc*-*-linux*paired* target.
1596 * config/rs6000/750cl.h: Delete.
1597 * config/rs6000/paired.h: Delete.
1598 * config/rs6000/paired.md: Delete.
1599 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1601 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1602 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1603 comment. Remove paired float support.
1604 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1605 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1607 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1608 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1610 * config/rs6000/rs6000.c: Remove paired float support.
1611 (paired_expand_vector_init, paired_expand_vector_move,
1612 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1613 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1614 paired_expand_builtin, paired_expand_predicate_builtin,
1615 paired_init_builtins): Delete.
1616 * config/rs6000/rs6000.h: Remove paired float support.
1617 * config/rs6000/rs6000.md: Remove paired float support.
1618 (move_from_CR_ov_bit): Delete.
1619 * config/rs6000/rs6000.opt (mpaired): Delete.
1620 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1621 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1623 2018-05-02 Richard Biener <rguenther@suse.de>
1626 * gimplify.c (gimplify_save_expr): When in SSA form allow
1627 SAVE_EXPRs to compute to SSA vars.
1629 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1632 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1633 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1634 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1635 clobber operands[2], instead use a new pseudo. Formatting fixes.
1637 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1639 PR tree-optimization/85586
1640 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1641 exit early for statements in the same group if the accesses are
1644 2018-05-02 Tom de Vries <tom@codesourcery.com>
1647 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1650 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1652 PR tree-optimization/85143
1653 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1655 2018-05-01 Tom de Vries <tom@codesourcery.com>
1658 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1659 not found" error message.
1661 2018-05-01 Tom de Vries <tom@codesourcery.com>
1664 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1665 * vec.c (test_ordered_remove_if): New function.
1666 (vec_c_tests): Call test_ordered_remove_if.
1667 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1668 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1669 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1670 VEC_ORDERED_REMOVE_IF.
1672 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1674 PR tree-optimization/82665
1675 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1676 pointer subtraction where arguments come from a memchr call.
1678 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1680 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1681 --push-state --as-needed and --pop-state instead of --as-needed and
1682 --no-as-needed if ld supports it.
1683 * configure: Regenerated.
1686 * doc/install.texi2html: Replace _002d with - and _002a with * in
1687 generated html files using sed.
1689 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1692 * gcc-rich-location.c (blank_line_before_p): New function.
1693 (use_new_line): New function.
1694 (gcc_rich_location::add_fixit_insert_formatted): New function.
1695 * gcc-rich-location.h
1696 (gcc_rich_location::add_fixit_insert_formatted): New function.
1698 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1700 * selftest.c (assert_streq): Rename "expected" and "actual" to
1701 "val1" and "val2". Extend NULL-handling to cover both inputs
1702 symmetrically, while still requiring both to be non-NULL for a pass.
1703 * selftest.h (assert_streq): Rename "expected" and "actual" to
1705 (ASSERT_EQ): Likewise.
1706 (ASSERT_EQ_AT): Likewise.
1707 (ASSERT_KNOWN_EQ): Likewise.
1708 (ASSERT_KNOWN_EQ_AT): Likewise.
1709 (ASSERT_NE): Likewise.
1710 (ASSERT_MAYBE_NE): Likewise.
1711 (ASSERT_MAYBE_NE_AT): Likewise.
1712 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1713 the assertion to pass.
1714 (ASSERT_STREQ_AT): Likewise.
1716 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1718 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1719 interaction with -pie.
1721 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1723 * selftest.h: Fix alphabetization of per-source-file selftest
1726 2018-04-30 Jason Merrill <jason@redhat.com>
1728 PR c++/61982 - dead stores to destroyed objects.
1729 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1732 2018-04-30 Jason Merrill <jason@redhat.com>
1734 * tree.c (build_clobber): New.
1735 * tree.h: Declare it.
1736 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1738 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1740 * diagnostic-show-locus.c (layout::layout): Update for
1741 location_get_source_line returning a char_span.
1742 (struct char_span): Move to input.h.
1743 (struct correction): Update for fields in char_span becoming
1745 (struct source_line): Update for location_get_source_line
1746 returning a char_span.
1747 (layout::print_line): Likewise.
1748 * edit-context.c (edited_file::print_content): Likewise.
1749 (edited_file::print_diff_hunk): Likewise.
1750 (edited_file::print_run_of_changed_lines): Likewise.
1751 (edited_file::get_num_lines): Likewise.
1752 (edited_line::edited_line): Likewise.
1753 * final.c (asm_show_source): Likewise.
1754 * input.c (location_get_source_line): Convert return type
1755 from const char * to char_span, losing the final "line_len"
1757 (dump_location_info): Update for the above.
1758 (get_substring_ranges_for_loc): Likewise. Use a char_span
1759 when handling the literal within the line.
1760 (test_reading_source_line): Update for location_get_source_line
1761 returning a char_span.
1762 * input.h (class char_span): Move here from
1763 diagnostic-show-locus.c, converting from a struct to a class.
1764 Make data members private.
1765 (char_span::operator bool): New.
1766 (char_span::length): New.
1767 (char_span::get_buffer): New.
1768 (char_span::operator[]): New.
1769 (char_span::subspan): Make const.
1770 (char_span::xstrdup): New.
1771 (location_get_source_line): Convert return type from const char *
1772 to char_span, losing the final "line_size" param.
1774 2018-04-30 Jan Hubicka <jh@suse.cz>
1776 * lto-wrapper.c (ltrans_priorities): New static var.
1777 (cmp_priority): New.
1778 (run_gcc): Read priorities and if doing parallel build order
1779 the Makefile by them.
1781 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1783 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1785 2018-04-30 Richard Biener <rguenther@suse.de>
1787 * tree-cfg.c (verify_address): Remove base argument, add
1788 flag whether to check TREE_ADDRESSABLE and do that.
1789 (verify_expr): Remove.
1790 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1791 (verify_gimple_assign_single): Likewise.
1792 (verify_gimple_switch): Likewise.
1793 (verify_expr_location_1): Dereference tp once. Add (disabled)
1794 piece from verify_expr.
1795 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1797 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1799 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1801 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1803 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1804 (small_data_pattern): Likewise.
1805 (arc_rewrite_small_data): Likewise.
1806 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1807 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1808 (get_symbol_alignment): New function.
1809 (legitimate_small_data_address_p): Likewise.
1810 (legitimate_scaled_address): Update, call
1811 legitimate_small_data_address_p.
1812 (output_sdata): New static variable.
1813 (arc_print_operand): Update how we handle small data operands.
1814 (arc_print_operand_address): Likewise.
1815 (arc_legitimate_address_p): Update, use
1816 legitimate_small_data_address_p.
1817 (arc_rewrite_small_data_p): Remove.
1818 (arc_rewrite_small_data_1): Likewise.
1819 (arc_rewrite_small_data): Likewise.
1820 (small_data_pattern): Likewise.
1821 (compact_sda_memory_operand): Update to use
1822 legitimate_small_data_address_p and get_symbol_alignment.
1823 (prepare_move_operands): Don't rewite sdata pattern.
1824 (prepare_extend_operands): Remove.
1825 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1827 (zero_extendqisi2): Likewise.
1828 (zero_extendhisi2): Likewise.
1829 (extendqihi2): Likewise.
1830 (extendqisi2): Likewise.
1831 (extendhisi2): Likewise.
1835 * config/arc/constraints.md (Usd): Change it to memory constraint.
1837 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1839 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1840 as source of std instructions.
1841 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1842 allow 6-bit constants as source for store instructions.
1843 (movdi_insn): Update instruction pattern to allow 6-bit constants
1844 as source for store instructions.
1846 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1848 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1850 2018-04-30 Nathan Sidwell <nathan@acm.org>
1851 Sandra Loosemore <sandra@codesourcery.com>
1853 * dumpfile.c (dump_open): Allow '-' for stdout.
1854 * doc/invoke.texi (Developer Options): Document dump filename
1855 determination early. Document stdin/stdout selection.
1857 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1859 Microblaze Target: PIC data text relative
1861 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1862 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1864 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1865 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1866 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1867 New addressing mode for data-text relative position indepenedent code.
1868 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1869 'ADDRESS_SYMBOLIC_TXT_REL'.
1870 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1871 (microblaze_legitimate_pic_operand): Exclude function calls from
1872 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1873 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1875 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1876 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1877 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1878 for 'address + offset'.
1879 (microblaze_expand_prologue): Add new function prologue call for
1881 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1882 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1883 table in case of TARGET_PIC_DATA_TEXT_REL.
1884 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1885 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1886 Add new macros 'UNSPEC_TEXT',
1887 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1888 + exclude function calls from 'UNSPEC_PLT' in case of data text
1890 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1891 new target hook for generating address diff vector tables in case of
1893 * doc/tm.texi : Regenerate.
1894 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1895 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1896 of addr diff vector generation.
1897 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1898 target hook definition.
1899 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1900 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1901 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1902 option for data text relative.
1904 2018-04-30 Richard Biener <rguenther@suse.de>
1906 * tree-chrec.h (evolution_function_is_constant_p): Remove
1908 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1910 2018-04-30 Richard Biener <rguenther@suse.de>
1913 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1915 2018-04-30 Richard Biener <rguenther@suse.de>
1917 PR tree-optimization/28364
1918 PR tree-optimization/85275
1919 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1920 copying first exit test.
1922 2018-04-28 Mark Wielaard <mark@klomp.org>
1924 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1926 (dwarf_AT): Handle DW_AT_addr_base.
1927 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1929 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1932 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1933 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1934 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1935 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1937 2018-04-28 Richard Biener <rguenther@suse.de>
1939 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1940 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1941 to reflect use. Only add interesting stmts.
1943 2018-04-27 Martin Jambor <mjambor@suse.cz>
1946 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1947 the jump function allows for passing through aggregate values.
1949 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1951 * input.h (in_system_header_at): Convert from macro to inline
1953 (from_macro_expansion_at): Likewise.
1954 (from_macro_definition_at): Likewise.
1956 2018-04-27 Jeff Law <law@redhat.com>
1958 * config.gcc: Mark tile* targets as deprecated/obsolete.
1960 2018-04-27 Richard Biener <rguenther@suse.de>
1962 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1965 2018-04-27 Richard Biener <rguenther@suse.de>
1967 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1969 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1971 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1972 with Yd constraint. Set "preferred_for_speed" attribute from
1973 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1975 (*movdi_internal): Ditto.
1976 (movti_interunit splitters): Remove
1977 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1978 (movdi_interunit splitters): Ditto.
1979 * config/i386/constraints.md (Ye): Remove.
1980 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1982 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1985 * config/aarch64/constraints.md (Usg): Limit to 31.
1988 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1990 PR tree-optimization/85529
1991 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1992 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1993 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1994 zero extension or masking of the MSB bit.
1995 (optimize_range_tests): Add FIRST_BB argument, pass it through
1996 to optimize_range_tests_var_bound.
1997 (maybe_optimize_range_tests, reassociate_bb): Adjust
1998 optimize_range_tests callers.
2000 2018-04-26 Richard Biener <rguenther@suse.de>
2001 Jakub Jelinek <jakub@redhat.com>
2003 * cgraph.h (symbol_table): Just declare debug method here.
2004 * symtab.c (symbol_table::debug): Define.
2006 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2008 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2010 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2012 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2013 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2014 (*movdi_internal): Substitute Yi and Yj constraint with x
2015 and Ym and Yn constraint with y constraint. Update "isa"
2016 attribute and set "preferred_for_speed" attribute from
2017 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2018 (*movsi_internal): Ditto.
2019 (*movdf_internal): Ditto.
2020 (*movsf_internal): Ditto.
2021 (*zero_extendsidi2): Ditto.
2022 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2023 (sse2_loadld): Ditto.
2024 (*vec_extract<ssevecmodelower>_0): Ditto.
2025 (*vec_extractv4si_0_zext_sse4): Ditto.
2026 (vec_concatv2di): Ditto.
2027 (*vec_dup<mode>): Ditto.
2028 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2029 * config/i386/constraints.md (Yi): Remove.
2034 2018-04-26 Nathan Sidwell <nathan@acm.org>
2036 * dumpfile.c (dump_open): New.
2037 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2038 (dump_finish): Detect stdio/stderr by value not name.
2040 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2042 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2044 2018-04-26 Tom de Vries <tom@codesourcery.com>
2047 * config/nvptx/nvptx.c (verify_neutering_jumps)
2048 (verify_neutering_labels): New function
2049 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2051 2018-04-26 Tom de Vries <tom@codesourcery.com>
2054 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2055 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2058 2018-04-26 Richard Biener <rguenther@suse.de>
2059 Tom de Vries <tom@codesourcery.com>
2062 * lto-streamer-out.c (output_function): Fixup loops if required to match
2063 discovery done in the reader.
2065 2018-04-26 Richard Biener <rguenther@suse.de>
2067 PR tree-optimization/85116
2068 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2069 have a loop exit from the single latch predecessor. Remove
2070 case of header with just condition.
2071 (ch_base::copy_headers): Exclude infinite loops from any
2073 (pass_ch::execute): Record exits.
2075 2018-04-26 Richard Biener <rguenther@suse.de>
2077 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2078 prologue cost vector and pass it to vect_get_load_cost.
2079 (vect_get_peeling_costs_all_drs): Likewise.
2080 (vect_peeling_hash_get_lowest_cost): Likewise.
2081 (vect_enhance_data_refs_alignment): Likewise.
2083 2018-04-26 Richard Biener <rguenther@suse.de>
2086 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2087 checking of integer<->pointer conversions.
2088 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2089 sign-/zero-extending pointer types.
2090 (expand_omp_for_static_chunk): Likewise.
2092 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2093 Jean Lee <xiaoyur347@gmail.com>
2095 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2096 (TARGET_ASAN_SHADOW_OFFSET): Define.
2097 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2098 true for -fsanitize=address.
2100 2018-04-25 Mark Wielaard <mark@klomp.org>
2102 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2105 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2107 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2108 than "alu", remove explicit "memory" and "imm_disp" attributes.
2109 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2112 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2113 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2114 gen_lowpart_no_emit.
2116 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2119 * config/i386/i386.c (ix86_expand_builtin): Change memory
2120 operand to XI, extend p0 to Pmode.
2121 * config/i386/i386.md: Change unspec volatile and operand
2122 1 mode to XI, change operand 0 mode to P.
2124 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2126 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2127 GET_MODE_MASK before any checking.
2128 (nds32_can_use_bset_p): Likewise.
2129 (nds32_can_use_btgl_p): Likewise.
2131 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2133 * config/nds32/nds32-doubleword.md: New define_split pattern for
2134 illegal register number.
2136 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2138 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2140 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2142 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2144 2018-04-25 Richard Biener <rguenther@suse.de>
2146 * lto-streamer.h (LTO_major_version): Bump to 8.
2148 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2150 * BASE-VER: Set to 9.0.0.
2152 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2154 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2155 in __abskf2 and __powikf2.
2157 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2160 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2161 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2162 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2163 Use the above on operand 2. Reindent.
2164 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2166 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2169 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2171 * config/i386/i386.opt (mcet): Removed.
2172 * doc/install.texi: Remove -mcet documentation.
2173 * doc/invoke.texi: Likewise.
2175 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2178 * doc/install.texi: Remove -mcet from bootstrap-cet.
2180 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2183 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2184 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2188 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2189 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2190 containing a CONST_VECTOR.
2192 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2194 * doc/install.texi: Update newlib dependency for nvptx.
2196 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2199 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2200 instead of INTVAL when shifting x left.
2202 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2204 PR tree-optimization/85478
2205 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2206 vect_grouped_store_supported for single element vectors.
2208 2018-04-24 Richard Biener <rguenther@suse.de>
2211 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2212 load cost increase to the case of non-constant step.
2214 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2217 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2218 destination if any_malformed_asm.
2220 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2223 * expr.c (store_field): In the bitfield case, if the value comes from
2224 a function call and is returned in registers by means of a PARALLEL,
2225 do not change the mode of the temporary unless BLKmode and VOIDmode.
2227 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2229 PR rtl-optimization/85423
2230 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2231 dependencies to debug insns when the previous insn is non-debug.
2233 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2235 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2236 enums into a single definition.
2237 (fls): Fix predicates and printing.
2240 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2242 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2243 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2244 and short u6 immediate.
2245 (check_if_valid_sleep_operand): Remove.
2246 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2248 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2250 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2251 flag_always_save_lp condition.
2252 * config/nds32/nds32.opt (malways-save-lp): New option.
2254 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2256 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2257 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2258 * config/nds32/nds32.h
2259 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2260 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2262 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2264 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2265 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2267 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2268 Chung-Ju Wu <jasonwucj@gmail.com>
2270 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2272 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2273 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2275 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2277 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2279 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2281 * config/nds32/nds32-protos.h (nds32_data_alignment,
2282 nds32_local_alignment): Declare.
2283 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2284 nds32_local_alignment): New functions.
2285 (TARGET_CONSTANT_ALIGNMENT): Define.
2286 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2288 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2290 * config/nds32/nds32.c
2291 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2292 (TARGET_MODES_TIEABLE_P): Likewise.
2294 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2296 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2299 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2300 Chung-Ju Wu <jasonwucj@gmail.com>
2302 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2303 for unaligned access.
2304 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2306 * config/nds32/nds32-intrinsic.md: Likewise.
2307 * config/nds32/nds32_intrinsic.h: Likewise.
2308 * config/nds32/nds32.h (nds32_builtins): Likewise.
2309 * config/nds32/nds32.opt (munaligned-access): New option.
2310 * config/nds32/nds32.c (nds32_asm_file_start): Display
2311 flag_unaligned_access status.
2313 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2315 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2316 -mno-relax is present.
2317 * config/riscv/linux.h (LINK_SPEC): Ditto.
2319 2018-04-20 Martin Sebor <msebor@redhat.com>
2322 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2324 (gimple_fold_builtin_stxcpy_chk): Same.
2325 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2327 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2330 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2331 __powikf2 when long double is IEEE 128-bit.
2333 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2335 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2336 step to make sure stack always aligned.
2338 2018-04-20 Carl Love <cel@us.ibm.com>
2341 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2342 size check for arg0.
2344 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2345 Tom de Vries <tom@codesourcery.com>
2348 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2349 Emit insns for calls too.
2350 (nvptx_find_par): Always look for worker-level predecessor insn.
2351 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2353 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2354 (nvptx_process_pars): Propagate frames for calls.
2356 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2359 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2361 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2362 (ix86_handle_option): Don't handle OPT_mibt.
2363 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2365 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2367 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2368 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2369 (ix86_target_macros): Define __CET__ with flag_cf_protection
2370 for -fcf-protection.
2371 * config/i386/i386.c (isa2_opts): Remove -mibt.
2372 * config/i386/i386.h (TARGET_IBT): Removed.
2373 (TARGET_IBT_P): Likewise.
2374 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2375 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2376 * config/i386/i386.opt (mcet): Update help message.
2379 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2380 -mcet as an alias for -mshstk.
2382 2018-04-20 Richard Biener <rguenther@suse.de>
2385 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2386 complexity by forcing a single use of the multiply operand.
2388 2018-04-20 Martin Jambor <mjambor@suse.cz>
2391 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2392 recursion dependency to only apply to non-clones.
2394 2018-04-20 Martin Jambor <mjambor@suse.cz>
2397 * ipa-cp.c (create_specialized_node): Check that clones of
2398 self-recursive edges exist during IPA-CP.
2400 2018-04-19 Toon Moene <toon@moene.org>
2402 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2405 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2407 PR tree-optimization/85467
2408 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2409 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2410 VECTOR_CST element to type.
2412 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2415 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2416 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2417 (builtin_longjmp): Likewise.
2418 (save_stack_nonlocal): New pattern.
2419 (restore_stack_nonlocal): Likewise.
2421 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2424 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2425 Replace ASM_OUTPUT_LABEL with fprintf.
2427 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2430 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2431 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2432 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2433 define __IBT__ and __SHSTK__ for -fcf-protection.
2434 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2436 (ix86_trampoline_init): Likewise.
2437 (x86_output_mi_thunk): Likewise.
2438 (ix86_notrack_prefixed_insn_p): Likewise.
2439 (ix86_option_override_internal): Don't disallow -fcf-protection.
2440 * config/i386/i386.md (rdssp<mode>): Also enable for
2442 (incssp<mode>): Likewise.
2443 (nop_endbr): Likewise.
2444 * config/i386/i386.opt (mcet): Change help message to built-in
2448 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2449 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2450 enable CET built-in functions.
2452 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2454 * common/config/i386/i386-common.c
2455 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2456 OPTION_MASK_ISA_MOVDIRI_UNSET,
2457 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2458 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2459 * config.gcc (movdirintrin.h): New header.
2460 * config/i386/cpuid.h (bit_MOVDIRI,
2461 bit_MOVDIR64B): New bits.
2462 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2464 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2465 (VOID, PVOID, PCVOID)): New function types.
2466 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2467 __builtin_ia32_directstoreu_u64,
2468 __builtin_ia32_movdir64b): New builtins.
2469 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2470 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2472 (ix86_valid_target_attribute_inner_p): Ditto.
2473 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2474 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2475 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2476 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2477 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2478 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2479 (movdiri<mode>, movdir64b_<mode>): New.
2480 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2481 * config/i386/immintrin.h: Include movdirintrin.h.
2482 * config/i386/movdirintrin.h: New file.
2483 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2485 2018-04-19 Richard Biener <rguenther@suse.de>
2488 * cfg.c (clear_bb_flags): When loop state says we have
2489 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2491 2018-04-19 Richard Biener <rguenther@suse.de>
2493 PR tree-optimization/84737
2494 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2495 copying restrict info.
2496 (vect_setup_realignment): Use it.
2497 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2498 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2499 the first DR to all generated stores.
2500 (vectorizable_load): Likewise for loads.
2502 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2504 PR tree-optimization/85446
2505 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2506 the integral and pointer types to have the same precision.
2508 * doc/install.texi: Document --disable-cet being the default and
2511 2018-04-18 Martin Liska <mliska@suse.cz>
2513 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2516 2018-04-18 Martin Liska <mliska@suse.cz>
2519 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2522 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2523 arguments if they are comparable.
2525 2018-04-18 Martin Liska <mliska@suse.cz>
2528 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2531 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2534 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2537 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2538 ENDBR after calling __morestack.
2540 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2543 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2544 by using gcc_base_ver to generate a gcc_driver_version, and use
2545 it when generating GCC_DRIVER_NAME.
2546 * configure: Regenerate.
2548 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2551 * config.gcc: Obsolete powerpc*-*-*spe*.
2553 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2556 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2557 (stabstr_D): Change type of unum from unsigned int to
2558 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2561 2018-04-17 Jim Wilson <jimw@sifive.com>
2564 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2565 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2566 Set arg_pointer_offset after using pretend_args_size.
2568 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2570 PR rtl-optimization/85431
2571 * dse.c (record_store): Ignore zero width stores.
2574 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2575 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2576 __builtin_stack_restore rather than after it.
2577 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2578 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2579 argument instead of virtual_dynamic_stack_rtx.
2581 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2583 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2585 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2586 Add note to error message to explain internal mapping of overloaded
2587 built-in function name to non-overloaded built-in function name.
2588 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2591 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2594 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2595 where the inputs overlap with the output.
2597 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2599 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2600 (=v, v) alternative and explicit "memory" attribute.
2601 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2602 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2604 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2605 "sselog1" type instead of "sselog".
2606 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2607 "sselog". Remove explicit "memory" attribute.
2608 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2609 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2611 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2612 "sselog1" type instead of "sselog". Remove explicit "memory"
2614 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2615 use "sselog1" type instead of "sselog". Remove explicit "memory"
2617 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2618 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2620 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2621 "sselog1" type instead of "sselog". Remove explicit "memory"
2623 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2624 use "sselog1" type instead of "sselog". Remove explicit "memory"
2628 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2631 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2634 2018-04-17 Martin Jambor <mjambor@suse.cz>
2637 * ipa-cp.c (create_specialized_node): Call
2638 expand_all_artificial_thunks if necessary.
2640 2018-04-17 Martin Liska <mliska@suse.cz>
2643 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2644 in message, remote space in between '_G' and '('.
2646 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2649 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2650 avx512f_vmcmp<mode>3<round_saeonly_name>,
2651 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2652 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2653 avx512f_rndscale<mode><round_saeonly_name>,
2654 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2655 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2656 Use %<iptr>2 instead of %2 for -masm=intel.
2657 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2658 avx512f_vcvttss2usi<round_saeonly_name>,
2659 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2661 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2662 avx512f_vcvttsd2usi<round_saeonly_name>,
2663 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2664 Use %q1 instead of %1 for -masm=intel.
2665 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2666 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2667 of %3 for -masm=intel.
2668 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2670 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2672 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2674 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2675 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2677 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2679 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2680 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2681 %g1 and one with %0 and %1.
2682 (avx512er_vmrcp28<mode><round_saeonly_name>,
2683 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2685 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2686 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2687 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2688 of %0 and %{%4%} for -masm=intel.
2689 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2690 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2691 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2692 order of %0 and %{%5%}%{z%} for -masm=intel.
2694 2018-04-17 Jan Hubicka <jh@suse.cz>
2697 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2699 2018-04-17 Martin Liska <mliska@suse.cz>
2702 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2703 for target_clone error message. Make default implementation
2704 clone to be a local declaration.
2705 (separate_attrs): Add new argument and check for an empty
2707 (expand_target_clones): Handle it.
2708 (ipa_target_clone): Make redirection just for target_clones
2711 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2712 Tom de Vries <tom@codesourcery.com>
2715 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2716 tiled basic blocks without omp continue statements.
2718 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2721 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2722 vec_extract expression as having side effects to make sure it gets
2725 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2728 * config/i386/i386.c (get_builtin_code_for_version): Check
2731 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2734 * gcc/config.gcc: Support "skylake".
2735 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2737 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2738 (processor_target_table): Add "skylake".
2739 (ix86_option_override_internal): Add "skylake".
2740 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2741 PROCESSOR_CANNONLAKE.
2742 (get_builtin_code_for_version): Fix priority for
2743 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2744 PROCESSOR_SKYLAKE-AVX512.
2745 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2746 (processor_type): Add PROCESSOR_SKYLAKE.
2748 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2749 Jason Merrill <jason@redhat.com>
2752 * convert.c (convert_to_integer_1): Use direct recursion for
2753 enumeral types and types with a precision less than the number
2754 of bits in their mode.
2756 2018-04-16 Julia Koval <julia.koval@intel.com>
2759 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2760 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2762 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2765 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2766 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2767 and -mno-direct-move.
2769 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2772 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2773 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2774 Ensure negative shifts result in {0}.
2776 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2778 PR rtl-optimization/79916
2779 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2780 regs (if any) to define how to gnerate SD moves when LRA is in
2783 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2785 PR rtl-optimization/85393
2786 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2787 * except.c (expand_dw2_landing_pad_for_region): Make static.
2788 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2789 a label and unconditional jump to old_bb, rather than
2790 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2793 PR rtl-optimization/85376
2794 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2795 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2796 instead of a specific value.
2798 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2799 Bin Cheng <bin.cheng@arm.com>
2801 PR tree-optimization/82965
2802 PR tree-optimization/83991
2803 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2804 by_profile_only parameter.
2805 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2806 information if the loop was predicted to iterate too many times.
2807 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2809 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2812 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2815 2018-04-13 Martin Liska <mliska@suse.cz>
2816 Jakub Jelinek <jakub@redhat.com>
2819 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2820 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2821 * builtins.c (expand_builtin_memory_copy_args): Use
2822 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2823 handle dest_addr == pc_rtx.
2825 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2828 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2829 asked to not generate direct moves.
2830 (fix_trunc<mode>si2_stfiwx): Similar.
2831 (fix_trunc<mode>si2_internal): Similar.
2833 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2836 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2837 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2838 lookup if dest in some wider mode is known to be const0_rtx and
2839 if so, record permanent equivalence for it to be ZERO_EXTEND of
2840 the narrower mode destination.
2842 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2844 * lto-streamer-out.c (output_function): Revert 259346.
2845 * omp-expand.c (expand_oacc_for): Likewise.
2847 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2849 PR rtl-optimization/85354
2850 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2851 * sel-sched.c (sel_global_init): ... here.
2853 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2856 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2857 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2858 mode for PE-COFF targets.
2859 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2860 (i386_pe_asm_lto_end): Likewise.
2861 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2862 (TARGET_ASM_LTO_END): Likewise.
2863 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2864 (i386_pe_asm_lto_start): New function.
2865 (i386_pe_asm_lto_end): Likewise.
2867 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2868 Richard Biener <rguenther@suse.de>
2871 * lto-streamer-out.c (output_function): Fix CFG loop state before
2873 * omp-expand.c (expand_oacc_for): Handle calls to internal
2874 functions like regular functions.
2876 2018-04-12 Richard Biener <rguenther@suse.de>
2879 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2880 for the early LTO debug to properly generate references to it
2881 during DIE emission. Do not re-use that for the skeleton for
2883 (dwarf2out_early_finish): Likewise.
2885 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2888 * config/i386/sse.md
2889 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2890 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2891 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2892 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2893 and output is a reg, avoid creating invalid lowpart subreg, but
2894 instead split into a 512-bit move. Don't split if not AVX512VL,
2895 input is xmm16+ reg and output is a mem.
2896 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2897 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2898 xmm16+ reg and output is a mem.
2900 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2902 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2903 also for flag_dwarf2_cfi_asm.
2905 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2907 PR rtl-optimization/85342
2908 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2909 a bool scalar var inside of the loop instead. Don't try to update
2910 recog_data.operand after failed apply_change_group.
2912 2018-04-12 Tom de Vries <tom@codesourcery.com>
2915 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2916 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2917 array with flexible array member as array without given dimension.
2918 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2919 argument for undefined param to true.
2921 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2924 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2925 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2926 from PowerPC section.
2927 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2928 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2929 help text that is too long.
2930 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2931 help text that is too long.
2932 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2933 help text that is too long.
2935 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2937 * config/alpha/alpha.md (stack_probe_internal): Rename
2938 from "probe_stack". Update all callers.
2940 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2942 PR rtl-optimization/84566
2943 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2944 sched_macro_fuse_insns.
2946 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2949 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2950 (compute_block_dependences): ... from here.
2952 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2954 PR tree-optimization/85331
2955 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2956 from int to HOST_WIDE_INT.
2958 2018-04-11 Martin Jambor <mjambor@suse.cz>
2961 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2962 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2963 not the same as the source val.
2964 (cgraph_edge_brings_value_p): New parameter.
2965 (gather_edges_for_value): Pass destination value to
2966 cgraph_edge_brings_value_p.
2967 (perhaps_add_new_callers): Likewise.
2968 (get_info_about_necessary_edges): Likewise and exclude values brought
2969 only by self-recursive edges.
2970 (create_specialized_node): Redirect only clones of self-calling edges.
2971 (+self_recursive_pass_through_p): New function.
2972 (find_more_scalar_values_for_callers_subset): Use it.
2973 (find_aggregate_values_for_callers_subset): Likewise.
2974 (known_aggs_to_agg_replacement_list): Removed.
2975 (decide_whether_version_node): Re-calculate known constants for all
2976 remaining context clones.
2978 2018-04-11 Richard Biener <rguenther@suse.de>
2981 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2982 from early DWARF output.
2983 (dwarf2out_early_finish): Output line info unconditionally into
2984 early DWARF and add reference to it.
2986 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2989 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2990 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2991 other than V2DFmode using iptr mode attribute.
2992 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2994 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2996 PR rtl-optimization/84659
2997 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2999 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3002 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3004 (output_loc_list): Pass address of a dummy size variable even in the
3005 locview handling loop.
3006 (index_location_lists): Add comment on why skip_loc_list_entry can't
3009 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3012 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3015 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3018 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3019 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3020 and -mstring-compare-inline-limit.
3022 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3025 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3026 for stack clash protection in a register whenever we need it to be in
3029 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3031 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3032 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3034 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3037 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3039 (mlong-double-): Ditto.
3040 * config/rs6000/sysv4.opt (msdata=): Ditto.
3041 (mtls-size=): Ditto.
3043 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3045 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3046 erroneous entries for
3047 "vector int vec_ldl (int, long int *)", and
3048 "vector unsigned int vec_ldl (int, unsigned long int *)".
3049 Add comments and entries for
3050 "vector bool char vec_ldl (int, bool char *)",
3051 "vector bool short vec_ldl (int, bool short *)",
3052 "vector bool int vec_ldl (int, bool int *)",
3053 "vector bool long long vec_ldl (int, bool long long *)",
3054 "vector pixel vec_ldl (int, pixel *)",
3055 "vector long long vec_ldl (int, long long *)",
3056 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3057 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3058 type tree bool_long_long_type_node and correct definition of
3059 bool_V2DI_type_node to make reference to this new type tree.
3060 (rs6000_mangle_type): Replace erroneous reference to
3061 bool_long_type_node with bool_long_long_type_node.
3062 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3063 comments to emphasize sign distinctions for char and int types and
3064 replace RS6000_BTI_bool_long constant with
3065 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3066 use of RS6000_BTI_pixel.
3067 (bool_long_type_node): Remove this macro definition.
3068 (bool_long_long_type_node): New macro definition
3070 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3072 PR rtl-optimization/85300
3073 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3074 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3075 simplify_unary_operation fails.
3077 2018-04-10 Martin Liska <mliska@suse.cz>
3079 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3080 cgraph_edge and ipa_ref.
3082 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3086 * config/i386/sse.md
3087 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3088 computation of the VEC_MERGE selector from mask.
3089 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3090 Fix decoding of the VEC_MERGE selector into mask.
3092 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3094 PR tree-optimization/85286
3095 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3097 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3099 * final.c (final_1): Set insn_last_address as well as
3100 insn_current_address.
3102 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3105 * explow.c (emit_stack_probe): Call validize_mem on memory location
3106 before passing it to gen_probe_stack. Create address operand and
3107 legitimize it for the probe_stack_address case.
3109 2018-04-09 Jan Hubicka <jh@suse.cz>
3112 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3113 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3114 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3116 (free_lang_data): Rebuild type inheritance graph.
3118 2018-04-09 Martin Sebor <msebor@redhat.com>
3120 * invoke.texi (-finline-small-functions): Mention other optimization
3122 (-findirect-inlining, -fpartial-inlining): Same.
3123 (-finline-functions-called-once): Same.
3124 (-freorder-blocks-and-partition): Same.
3126 2018-04-09 Jan Hubicka <jh@suse.cz>
3129 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3130 jumps; choose last target that matches the criteria (i.e.
3131 no partition changes for non-crossing jumps).
3132 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3133 support for redirecting crossing jumps to non-crossing.
3135 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3137 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3138 also for naked functions.
3140 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3142 * config/arc/arc.md (add_shift): New pattern.
3143 (add_shift2): Likewise.
3144 (sub_shift): Likewise.
3145 (sub_shift_cmp0_noout): Likewise.
3146 (compare_si_ashiftsi): Likewise.
3147 (xbfu_cmp0_noout): New combine pattern.
3148 (xbfu_cmp0"): Likewise.
3149 (movsi_set_cc_insn): Place the predicable variant first.
3150 (commutative_binary_cmp0_noout): Remove clobber.
3151 (commutative_binary_cmp0): New pattern.
3152 (noncommutative_binary_cmp0): Likewise.
3153 (noncommutative_binary_cmp0_noout): Likewise.
3154 (noncommutative_binary_comparison_result_used): Removed.
3155 (rsub_cmp0): New pattern.
3156 (rsub_cmp0_noout): Likewise.
3157 (extzvsi): Changed, keep only meaningful variants.
3158 (SQH, SEZ): New iterators.
3159 (SQH_postfix): New mode attribute.
3160 (SEZ_prefix): New code attribute.
3161 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3162 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3163 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3165 (noncommutative_operator): Check the availability of barrel
3168 2018-04-09 Richard Biener <rguenther@suse.de>
3170 PR tree-optimization/85284
3171 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3172 Only use the niter constraining form of simple_iv when the exit
3175 2018-04-09 Tom de Vries <tom@codesourcery.com>
3178 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3179 (define_expand "*memory_barrier"): New define_expand.
3180 (define_insn "memory_barrier"): New insn.
3182 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3184 PR rtl-optimization/80463
3185 PR rtl-optimization/83972
3186 PR rtl-optimization/83480
3188 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3189 correct producer for the insn.
3190 (tidy_control_flow): Fixup seqnos in case of debug insns.
3192 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3194 PR rtl-optimization/83913
3196 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3197 different sched-times when merging exprs.
3199 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3201 PR rtl-optimization/83962
3203 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3204 tidy_fallthru_edge and tidy_control_flow.
3206 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3208 PR rtl-optimization/83530
3210 * sel-sched.c (force_next_insn): New global variable.
3211 (remove_insn_for_debug): When force_next_insn is true, also leave only
3212 next insn in the ready list.
3213 (sel_sched_region): When the region wasn't scheduled, make another pass
3214 over it with force_next_insn set to 1.
3216 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3218 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3220 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3221 for interrupt control.
3222 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3223 functions for interrupt control.
3224 * config/nds32/nds32-intrinsic.md: Likewise.
3225 * config/nds32/nds32_intrinsic.h: Likewise.
3226 * config/nds32/nds32.h (nds32_builtins): Likewise.
3228 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3230 * config/nds32/nds32.c (nds32_init_machine_status,
3231 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3232 strict_aligned_p field.
3233 (nds32_expand_to_rtl_hook): New function.
3234 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3235 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3237 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3238 Chung-Ju Wu <jasonwucj@gmail.com>
3240 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3241 * config/nds32/nds32-n7.md: New file.
3242 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3243 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3245 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3246 * config/nds32/nds32.md (pipeline_model): Add n7.
3247 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3248 * config/nds32/pipelines.md: Include n7 settings.
3250 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3251 Chung-Ju Wu <jasonwucj@gmail.com>
3253 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3254 * config/nds32/nds32-e8.md: New file.
3255 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3256 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3258 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3259 * config/nds32/nds32.md (pipeline_model): Add e8.
3260 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3261 * config/nds32/pipelines.md: Include e8 settings.
3263 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3264 Chung-Ju Wu <jasonwucj@gmail.com>
3266 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3267 * config/nds32/nds32-n8.md: New file.
3268 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3269 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3271 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3272 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3273 * config/nds32/nds32.md (pipeline_model): Add n8.
3274 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3275 * config/nds32/pipelines.md: Include n8 settings.
3277 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3278 Chung-Ju Wu <jasonwucj@gmail.com>
3280 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3281 * config/nds32/nds32-n9-2r1w.md: New file.
3282 * config/nds32/nds32-n9-3r2w.md: New file.
3283 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3284 nds32_register_ports): New or modify for cpu n9.
3285 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3287 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3288 * config/nds32/nds32-utils.c: New file.
3289 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3290 TARGET_MUL_SLOW): Define.
3291 * config/nds32/nds32.md (pipeline_model): New attribute.
3292 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3293 New options that support cpu n9.
3294 * config/nds32/pipelines.md: Include n9 settings.
3295 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3297 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3299 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3300 information if necessary.
3301 (output_cond_branch_compare_zero): Likewise.
3302 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3303 (nds32_target_alignment): Refine for alignment.
3304 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3305 (FUNCTION_BOUNDARY): Modify.
3306 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3308 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3310 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3312 * config/nds32/constants.md (unspec_volatile_element): Add values for
3313 TLB operation and data prefetch.
3314 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3315 functions for TLB operation and data prefetch.
3316 * config/nds32/nds32-intrinsic.md: Likewise.
3317 * config/nds32/nds32_intrinsic.h: Likewise.
3318 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3319 (nds32_print_operand): Likewise.
3320 * config/nds32/nds32.h (nds32_builtins): Likewise.
3322 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3323 Andrew Pinski <pinsika@gcc.gnu.org>
3326 * match.pd: Use constant_boolean_node of correct type instead of
3327 boolean_true_node or boolean_false_node for simplifying
3328 pointer comparisons to zero.
3330 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3332 PR tree-optimization/80021
3333 * tree.c (verify_type_variant): Make error call in verify_variant_match
3334 translatable and remove final full stop.
3336 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3338 * config/nds32/constants.md (unspec_volatile_element): Add
3339 UNSPEC_VOLATILE_EH_RETURN.
3340 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3341 nds32_output_stack_pop): Support dwarf exception handling process.
3342 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3343 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3344 exception handling process.
3345 (nds32_compute_stack_frame): Likewise.
3346 (nds32_return_addr_rtx): Likewise.
3347 (nds32_initial_elimination_offset): Likewise.
3348 (nds32_expand_prologue): Likewise.
3349 (nds32_expand_epilogue): Likewise.
3350 (nds32_dynamic_chain_address): New function.
3351 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3353 (DYNAMIC_CHAIN_ADDRESS): Define.
3354 (EH_RETURN_DATA_REGNO): Define.
3355 (EH_RETURN_STACKADJ_RTX): Define.
3356 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3357 patterns for dwarf exception handling.
3359 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3361 * config/nds32/nds32.h: Clean up obsolete macros.
3363 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3365 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3366 Add enum values for particular instructions.
3367 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3368 particular intrinsic functions.
3369 * config/nds32/nds32-intrinsic.md: Likewise.
3370 * config/nds32/nds32_intrinsic.h: Likewise.
3371 * config/nds32/nds32.h (nds32_builtins): Likewise.
3372 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3373 (btst, ave): New patterns for particular instructions.
3375 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3377 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3378 Add enum values for atomic load/store and memory sync.
3379 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3381 * config/nds32/nds32-intrinsic.md: Likewise.
3382 * config/nds32/nds32_intrinsic.h: Likewise.
3383 * config/nds32/nds32.h (nds32_builtins): Likewise.
3385 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3387 PR tree-optimization/85257
3388 * fold-const.c (native_encode_vector): If not all elts could fit
3389 and off is -1, return 0 rather than offset.
3390 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3391 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3392 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3393 adjust buffer in native_interpret_expr call.
3395 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3397 * config/nds32/constants.md (unspec_volatile_element): Add cache
3398 control enum values.
3399 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3400 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3401 * config/nds32/nds32.c (nds32_cctl_names): New.
3402 (nds32_print_operand): Handle cache control register names.
3403 * config/nds32/nds32.h (nds32_builtins): New enum values.
3404 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3406 * config/nds32/nds32.md (type): Add mmu.
3407 * config/nds32/pipelines.md (simple_insn): Add mmu.
3409 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3411 * config/nds32/nds32.md (type): Remove call.
3412 * config/nds32/pipelines.md (simple_insn): Likewise.
3414 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3416 * config/nds32/constants.md (unspec_volatile_element): Add
3417 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3418 UNSPEC_VOLATILE_FMFCFG.
3419 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3420 description for fmfcfg and fmfcsr.
3421 (bdesc_1arg): Add fmtcsr.
3422 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3423 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3424 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3425 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3426 unspec_fmfcfg): New patterns.
3427 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3428 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3429 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3430 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3431 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3432 __nds32__fmfcfg): Define.
3434 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3436 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3437 intrinsic register names.
3438 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3439 intrinsic register enum values and macros.
3441 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3443 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3444 for load/store addressing form.
3445 (nds32_print_operand_address): Likewise.
3447 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3450 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3451 based on LABEL_REF. Remove useless assertion.
3452 (pic_address_needs_scratch): Fix formatting.
3453 (sparc_legitimize_pic_address): Minor tweaks.
3454 (sparc_delegitimize_address): Adjust assertion accordingly.
3455 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3456 into symbolic_operand.
3457 (movsi_high_pic_label_ref): Likewise.
3458 (movsi_lo_sum_pic_label_ref): Likewise.
3459 (movdi_pic_label_ref): Likewise.
3460 (movdi_high_pic_label_ref): Likewise.
3461 (movdi_lo_sum_pic_label_ref): Likewise.
3463 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3465 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3466 custom LIB_SPEC setup.
3468 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3469 Kito Cheng <kito.cheng@gmail.com>
3471 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3472 * config/riscv/freebsd.h: New.
3474 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3476 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3477 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3480 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3481 Kito Cheng <kito.cheng@gmail.com>
3483 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3484 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3485 * config/nds32/nds32-protos.h (nds32_output_call,
3486 nds32_output_return): Declare.
3487 * config/nds32/nds32.md: Refine all the call and return patterns.
3489 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3492 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3493 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3495 PR rtl-optimization/84872
3496 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3497 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3500 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3502 * expr.c (copy_blkmode_to_reg): Revert 254862.
3503 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3505 2018-04-06 Richard Biener <rguenther@suse.de>
3508 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3509 after seeing a component reference with an adjacent field. Treat
3510 refs to arrays at struct end of external decls similar to
3511 refs to unconstrained commons.
3513 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3516 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3517 look through SAVE_EXPRs with non-side-effects argument. Adjust
3519 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3522 2018-04-06 Richard Biener <rguenther@suse.de>
3525 * alias.c (find_base_term): New wrapper around find_base_term
3526 unwinding CSELIB_VAL_PTR changes.
3527 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3530 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3532 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3534 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3535 constant definitions.
3536 ("nop"): lr 0,0 -> nopr r0
3537 ("nop_lr0", "nop_lr1"): New insn definitions.
3539 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3541 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3542 NDS32_V3PUSH_AVAILABLE_P macro.
3544 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3545 Chung-Ju Wu <jasonwucj@gmail.com>
3547 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3548 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3549 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3550 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3551 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3552 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3553 * config/nds32/constraints.md: New constraints and checking for hard
3554 float configuration.
3555 * config/nds32/iterators.md: New mode iterator and attribute for hard
3556 float configuration.
3557 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3559 * config/nds32/nds32-fpu.md: New file.
3560 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3561 deal with hard float code generation.
3562 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3564 (abi_type, float_reg_number): New enum type.
3565 * config/nds32/nds32-predicates.c: New predicates for hard float.
3566 * config/nds32/nds32-protos.h: Declare functions for hard float.
3567 * config/nds32/nds32.c: Implementation for hard float configuration.
3568 * config/nds32/nds32.h: Definitions for hard float configuration.
3569 * config/nds32/nds32.md: Include hard float machine description and
3570 modify patterns for hard float configuration.
3571 * config/nds32/nds32.opt: New options for hard float configuration.
3572 * config/nds32/predicates.md: New predicates for hard float
3575 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3577 * common/config/nds32/nds32-common.c
3578 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3580 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3583 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3584 CONSTRUCTOR_ELT (ctor, ...)->value.
3586 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3589 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3591 2018-04-05 Tom de Vries <tom@codesourcery.com>
3594 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3597 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3598 Kito Cheng <kito.cheng@gmail.com>
3600 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3601 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3602 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3603 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3605 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3606 Kito Cheng <kito.cheng@gmail.com>
3608 * config/nds32/constraints.md (Ufe): New memory constraint.
3609 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3610 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3611 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3613 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3614 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3616 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3618 * config/nds32/nds32.md: Use optimize_size in the condition for
3619 alu-shift instructions.
3621 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3623 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3625 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3627 * config/nds32/nds32.md (negsi2): Refine pattern.
3629 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3630 Chung-Ju Wu <jasonwucj@gmail.com>
3632 * config/nds32/iterators.md (shift_rotate): New code iterator.
3633 (shift): New code attribute.
3634 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3635 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3636 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3637 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3638 bit-wise operations.
3639 (andsi3, *andsi3): Ditto.
3640 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3641 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3642 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3643 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3644 nds32_ior_operand, nds32_xor_operand): New predicates.
3646 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3648 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3649 (addsi3, subsi3): ... this.
3651 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3653 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3655 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3657 * config/nds32/nds32.md: Adjust indention.
3659 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3661 * config/nds32/nds32.md (feature): New attribute.
3663 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3665 * config/nds32/nds32.md (subtype): New attribute.
3667 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3670 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3671 expansion to perform a bitwise AND of the argument followed by a
3672 boolean negation of the result.
3674 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3676 PR rtl-optimization/84878
3677 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3678 the basic block. Assert the use reference is not artificial and that
3679 it has an associated insn.
3681 2018-04-04 Michael Matz <matz@suse.de>
3683 * builtins.c (compute_objsize): Pass correct operand
3684 to array_at_struct_end_p.
3686 2018-04-04 Richard Biener <rguenther@suse.de>
3689 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3690 from contexts for DINFO_LEVEL_TERSE and below.
3692 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3694 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3695 resiter_operand condition.
3696 * config/nds32/nds32.md (*move<mode>): Ditto.
3698 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3699 Monk Chiang <sh.chiang04@gmail.com>
3701 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3703 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3705 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3707 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3708 Kito Cheng <kito.cheng@gmail.com>
3710 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3711 nds32_cond_code_str, output_cond_branch,
3712 output_cond_branch_compare_zero, nds32_expand_cbranch,
3713 nds32_expand_cstore, nds32_expand_movcc,
3714 nds32_output_cbranchsi4_equality_zero,
3715 nds32_output_cbranchsi4_equality_reg,
3716 nds32_output_cbranchsi4_equality_reg_or_const_int,
3717 nds32_output_cbranchsi4_greater_less_zero: New functions.
3718 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3719 nds32_expand_cstore, nds32_expand_movcc,
3720 nds32_output_cbranchsi4_equality_zero,
3721 nds32_output_cbranchsi4_equality_reg,
3722 nds32_output_cbranchsi4_equality_reg_or_const_int,
3723 nds32_output_cbranchsi4_greater_less_zero): Declare.
3724 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3725 nds32_rimm11s_operand): New predicates.
3726 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3727 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3730 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3732 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3733 * config/nds32/nds32.md: Ditto.
3734 * config/nds32/pipelines.md: Ditto.
3736 2018-04-04 Richard Biener <rguenther@suse.de>
3738 PR tree-optimization/85168
3739 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3740 propagating abnormals.
3742 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3744 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3746 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3747 Kito Cheng <kito.cheng@gmail.com>
3749 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3750 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3751 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3752 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3753 * config/nds32/nds32.md (sibcall_internal): New.
3754 (sibcall_register): Remove.
3755 (sibcall_immediate): Remove.
3756 (sibcall_value_internal): New.
3757 (sibcall_value_register): Remove.
3758 (sibcall_value_immediate): Remove.
3759 * config/nds32/predicates.md (nds32_general_register_operand): New.
3760 (nds32_call_address_operand): New.
3762 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3764 PR rtl-optimization/85167
3765 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3766 bb_defs if *split_p, instead preinitialize it to NULL.
3768 PR tree-optimization/85156
3769 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3770 evaluating the argument multiple times.
3772 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3774 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3776 (_mm_cvtpd_ps): Likewise.
3777 (_mm_cvttpd_epi32): Likewise.
3778 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3779 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3780 vector, pixel, and bool following altivec.h include.
3782 2018-04-03 Martin Sebor <msebor@redhat.com>
3784 * doc/extend.texi (Common Function Attributes): Clarify.
3785 (const attribute): Likewise.
3786 (pure attribute): Likewise.
3788 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3791 * config/i386/i386.c (ix86_expand_vector_set): Use
3792 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3794 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3796 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3797 instructions when changing rounding bits to preserve precision bits
3798 in the x87 control word.
3800 2018-04-03 Martin Liska <mliska@suse.cz>
3802 PR tree-optimization/82491
3803 * rtl.h (strip_offset_and_add): Replace += suboffset with
3804 poly_uint64 () + suboffset.
3806 2018-03-29 Martin Liska <mliska@suse.cz>
3807 Martin Jambor <mjambor@suse.cz>
3810 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3811 param_type is not an integral or pointer type.
3813 2018-04-03 Richard Biener <rguenther@suse.de>
3815 * sese.h (recompute_all_dominators): Remove.
3817 2018-04-02 Martin Sebor <msebor@redhat.com>
3819 * doc/invoke.texi (-Wrestrict): Fix typos.
3821 2018-04-02 Jim Wilson <jimw@sifive.com>
3823 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3824 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3825 (<optab>di3, <optab>si3_extend): Likewise.
3826 (<optab>si3_mask, <optab>si3_mask_1): New.
3827 (<optab>di3_mask, <optab>di3_mask_1): New.
3828 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3829 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3830 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3832 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3834 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3837 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3839 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3840 (nds32_canonicalize_comparison): New function.
3842 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3843 Kito Cheng <kito.cheng@gmail.com>
3844 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3846 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3847 * config/nds32/constants.md (unspec_volatile_element): Add
3848 UNSPEC_VOLATILE_RELAX_GROUP.
3849 * config/nds32/nds32-relax-opt.c: New file.
3850 * config/nds32/nds32-predicates.c
3851 (nds32_symbol_load_store_p): New function.
3852 * config/nds32/nds32-protos.h
3853 (nds32_symbol_load_store_p): Declare function.
3854 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3855 * config/nds32/nds32.c
3856 (nds32_register_pass): New function to register pass.
3857 (nds32_register_passes): New function to register passes.
3858 * config/nds32/nds32.md (relax_group): New pattern.
3859 * config/nds32/nds32.opt (mrelax-hint): New option.
3860 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3862 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3864 * config/nds32/t-nds32: Modify files dependency.
3866 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3868 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3869 (PROFILE_HOOK): Define its implementation.
3871 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3873 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3874 type and 32-bit size.
3876 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3879 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3880 (V_128_256): New mode iterator.
3881 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3882 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3883 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3885 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3886 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3888 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3891 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3892 NaN inputs correctly.
3894 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3897 * config/rs6000/vsx.md (??r): New mode attribute.
3898 (*vsx_mov<mode>_64bit): Use it.
3899 (*vsx_mov<mode>_32bit): Likewise.
3901 2018-03-30 Martin Sebor <msebor@redhat.com>
3903 PR tree-optimization/84818
3904 * builtins.c (check_access): Use warning_n.
3906 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3909 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3911 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3914 2018-03-30 Julia Koval <julia.koval@intel.com>
3917 * x86-tune.def (movx, partial_reg_dependency): Enable for
3920 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3923 * lra-constraints.c (process_alt_operands): Move setting
3924 this_alternative_matches below.
3926 2018-03-29 Martin Liska <mliska@suse.cz>
3929 * doc/invoke.texi: Document how LTO works with debug info.
3930 Describe auto-load support of binutils. Mention 'x86-64'
3931 as valid option value of -march option.
3933 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3935 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3938 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3939 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3940 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3943 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3946 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3947 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3948 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3949 for RS6000_BTM_POWERPC64.
3950 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3951 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3952 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3957 2018-03-28 Carl Love <cel@us.ibm.com>
3960 2017-09-27 Carl Love <cel@us.ibm.com>
3962 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3963 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3964 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3967 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3969 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3970 instead of __vector bool.
3971 (_mm_max_pu8): Likewise.
3972 (_mm_min_pi16): Likewise.
3974 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3977 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3978 (DIVWEUO): Likewise.
3980 (DIVDEUO): Likewise.
3981 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3982 DIVWEUO and DIVDEUO.
3983 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3984 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3985 (div_extend): Likewise.
3986 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3988 (__builtin_divweuo): Likewise.
3989 (__builtin_divdeo): Likewise.
3990 (__builtin_divdeuo): Likewise.
3992 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3995 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3996 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3998 PR tree-optimization/82004
3999 * gimple-match-head.c (optimize_pow_to_exp): New function.
4000 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4001 Don't fold to exp if optimize_pow_to_exp is false.
4003 2018-03-28 Martin Liska <mliska@suse.cz>
4006 * calls.c (initialize_argument_information): Fix trailing space.
4007 * common.opt: Fix typo and provide better explanation for
4008 -fsanitize-coverage option.
4009 * config/i386/i386.opt: Fix typo.
4011 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4012 Martin Liska <mliska@suse.cz>
4015 * gimplify.c (asan_poison_variable): Don't do the check for
4016 gimplify_omp_ctxp here.
4017 (gimplify_decl_expr): Do it here.
4018 (gimplify_target_expr): Likewise.
4020 2018-03-28 Martin Liska <mliska@suse.cz>
4023 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4024 chkp_type_bounds_count if MPX is not enabled.
4026 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4028 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4030 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4033 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4034 function to create the function decl for complex long double
4035 multiply and divide for -mabi=ieeelongdouble.
4036 (init_float128_ieee): Call it.
4038 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4041 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4042 -fcf-protection=branch -mibt.
4043 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4045 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4048 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4050 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4053 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4054 extern array declarations.
4056 2018-03-27 Richard Biener <rguenther@suse.de>
4059 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4060 explicit single_use checks.
4062 2018-03-27 Richard Biener <rguenther@suse.de>
4064 PR tree-optimization/85082
4065 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4068 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4070 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4071 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4072 Turn on fasynchronous-unwind-tables and funwind-tables.
4074 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4077 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4078 (*bmi_blsr_<mode>_ccz): Ditto.
4080 2018-03-26 Tom de Vries <tom@codesourcery.com>
4082 PR tree-optimization/85063
4083 * omp-general.c (offloading_function_p): New function. Factor out
4085 * omp-offload.c (pass_omp_target_link::gate): ... here.
4086 * omp-general.h (offloading_function_p): Declare.
4087 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4088 with attribute omp declare target for offloading functions.
4090 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4092 PR tree-optimization/84005
4093 * tree-data-ref.h (get_base_for_alignment): Declare.
4094 * tree-data-ref.c (get_base_for_alignment_1): New function.
4095 (get_base_for_alignment): Likewise.
4096 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4097 get_base_for_alignment to find a suitable base object, instead
4098 of always using drb->base_address.
4100 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4103 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4104 known size by default.
4106 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4109 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4110 and non BLKmode operands.
4112 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4115 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4116 Clean up attributes.
4118 2018-03-23 Richard Biener <rguenther@suse.de>
4121 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4122 we are going to emit early debug for LTO.
4124 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4127 * function.c (match_asm_constraints_1): Don't optimize if input
4128 doesn't satisfy general_operand predicate for output's mode.
4131 * alias.c (write_dependence_p): Don't require for x_canonicalized
4132 non-VOIDmode if x has VOIDmode.
4135 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4136 just don't try to optimize it rather than assert it never happens.
4138 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4140 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4141 macro expansions for definition of ST_INTERNAL_<mode> and
4142 LD_INTERNAL_<mode> builtins.
4143 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4145 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4147 (altivec_expand_st_builtin): Likewise.
4148 (altivec_expand_builtin): Remove calls to deleted functions.
4149 (rs6000_address_for_altivec): Delete this function.
4150 * config/rs6000/vector.md: Remove expands for
4151 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4153 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4156 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4157 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4158 re-computing once computed.
4159 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4160 (arm_init_machine_status): Initialize
4161 machine->static_chain_stack_bytes.
4163 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4166 * doc/extend.texi: Add four new prototypes for vec_ld.
4167 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4168 definitions for more logical presentation.
4169 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4170 entries for V1TI variants of __builtin_altivec_ld builtin.
4171 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4172 handling of V1TI variant of LVX icode pattern.
4173 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4174 (rs6000_gimple_fold_builtin): Likewise.
4175 (altivec_init_builtins): Add code to define
4176 __builtin_altivec_lvx_v1ti function.
4178 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4181 * function.c (match_asm_constraints_1): Don't do the optimization
4182 if input isn't a REG, SUBREG, MEM or constant.
4184 2018-03-22 Tom de Vries <tom@codesourcery.com>
4186 PR tree-optimization/84956
4187 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4188 bb_has_abnormal_pred.
4190 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4193 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4194 DECL_INITIAL (decl) to decl at the end.
4195 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4198 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4200 * doc/extend.texi (__builtin_tgmath): Document when complex
4201 integer types are treated as _Complex _Float64.
4203 2018-03-21 Tom de Vries <tom@codesourcery.com>
4205 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4207 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4209 PR tree-optimization/84960
4210 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4211 if it is ENTRY block, move them into single succ of ENTRY in that case.
4213 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4215 PR tree-optimization/84811
4216 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4217 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4218 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4219 (known_subrange_p): Update accordingly. Cast each value involved
4220 in the size comparison, rather than casting the result of the
4223 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4225 PR tree-optimization/84982
4226 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4227 by flipping the least significant bit rather than all bits from
4228 bitpos to bitpos + bitsize - 1.
4230 2018-03-21 Nathan Sidwell <nathan@acm.org>
4232 * doc/extend.texi (Deprecated Features): Remove mention of
4233 long-deleted deprecations.
4235 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4238 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4239 * configure: Regenerate.
4241 2018-03-21 Tom de Vries <tom@codesourcery.com>
4243 PR tree-optimization/83126
4244 * tree-parloops.c (num_phis): New function.
4245 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4247 2018-03-21 Nathan Sidwell <nathan@acm.org>
4249 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4250 mention anon-struct/union members and trailing attributes.
4252 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4254 PR tree-optimization/84969
4255 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4256 builtin memset partitions if they set different rhs values.
4258 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4260 PR rtl-optimization/84989
4261 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4262 VEC_DUPLICATE with scalar result mode.
4264 2018-03-21 Martin Liska <mliska@suse.cz>
4267 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4268 not intended return statement.
4270 2018-03-21 Martin Liska <mliska@suse.cz>
4273 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4274 (chkp_find_bound_slots_1): Limit number of iterations.
4276 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4279 * Minor grammar fixes for x86 options.
4281 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4284 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4285 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4287 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4290 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4291 (altivec_lvx_<mode>_1op): Likewise.
4292 (altivec_stvx_<mode>_2op): Likewise.
4293 (altivec_stvx_<mode>_1op): Likewise.
4294 (altivec_lvx_<VM2:mode>): New define_expand.
4295 (altivec_stvx_<VM2:mode>): Likewise.
4296 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4297 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4298 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4299 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4300 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4301 (rs6000_gen_lvx): Likewise.
4302 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4303 (altivec_expand_stv_builtin): Likewise.
4304 (altivec_expand_builtin): Likewise.
4305 * config/rs6000/vector.md: Likewise.
4307 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4310 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4313 2018-03-20 Richard Biener <rguenther@suse.de>
4316 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4317 sign-conversions as zero, fall back to standard scalar_stmt
4320 2018-03-20 Martin Liska <mliska@suse.cz>
4323 * predict.c (rebuild_frequencies): Handle case when we have
4324 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4326 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4329 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4330 flag_section_anchors.
4331 * varasm.c (use_blocks_for_decl_p): Remove hack for
4332 dw2_force_const_mem.
4335 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4337 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4338 be created, use lowpart_subreg of operands[0] rather than operands[0]
4340 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4341 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4342 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4343 and n constraint instead of aarch64_shift_imm_di and Usd.
4344 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4345 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4347 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4350 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4351 to favor GPR over NEON registers.
4352 (<shift>di3_neon): Likewise.
4354 2018-03-20 Tom de Vries <tom@codesourcery.com>
4357 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4358 (nvptx_process_pars): Emit bar.sync asap and alap.
4360 2018-03-20 Tom de Vries <tom@codesourcery.com>
4363 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4364 seen_label if seen_label is already set.
4366 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4369 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4370 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4371 Use 1U instead of 1. Formatting fixes.
4374 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4375 instead of TREE_TYPE (s1) for the return value.
4377 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4379 PR tree-optimization/84946
4380 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4381 bitsize + bitsize in poly_uint64 rather than poly_int64.
4384 * dwarf2asm.c: Include fold-const.c.
4385 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4386 of decl rather than decl itself.
4388 PR rtl-optimization/84643
4389 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4391 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4394 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4395 calling assemble_variable.
4397 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4400 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4401 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4403 2018-03-19 Jim Wilson <jimw@sifive.com>
4406 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4407 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4408 (riscv_first_stack_step): Likewise.
4409 (riscv_option_override): Use STACK_BOUNDARY instead of
4411 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4413 (BIGGEST_ALIGNMENT): Set to 128.
4414 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4415 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4418 2018-03-19 Richard Biener <rguenther@suse.de>
4420 PR tree-optimization/84933
4421 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4422 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4424 2018-03-19 Richard Biener <rguenther@suse.de>
4426 PR tree-optimization/84859
4427 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4428 (cond_if_else_store_replacement): Perform sinking operation on
4429 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4430 Generalize what a BB with a single eligible store is.
4432 2018-03-19 Richard Biener <rguenther@suse.de>
4434 PR tree-optimization/84929
4435 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4436 chrec_is_positive against non-chrec arg.
4438 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4441 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4443 2018-03-18 Martin Liska <mliska@suse.cz>
4445 PR rtl-optimization/84635
4446 * regrename.c (build_def_use): Use matches_mode only when
4449 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4451 PR tree-optimization/84913
4452 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4453 vectorize chains of COND_EXPRs.
4455 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4457 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4459 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4461 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4463 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4465 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4467 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4468 Kito Cheng <kito.cheng@gmail.com>
4470 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4471 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4472 (nds32_adjust_reg_alloc_order): New function.
4473 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4475 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4477 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4478 nds32_print_operand, nds32_print_operand_address): Use
4479 HOST_WIDE_INT_PRINT_DEC instead.
4481 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4483 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4485 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4488 * config/i386/i386.c (initial_ix86_tune_features,
4489 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4491 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4492 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4493 rather than 1u << ix86_tune. Formatting fix.
4494 (ix86_option_override_internal): Change ix86_arch_mask from
4495 unsigned int to unsigned HOST_WIDE_INT, initialize to
4496 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4497 (ix86_function_specific_restore): Likewise.
4499 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4502 * postreload.c (reload_combine_recognize_pattern): Perform
4503 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4504 truncate_int_for_mode the result for the destination's mode.
4507 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4508 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4510 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4513 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4516 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4517 regno_allocno_class_array and sorted_pseudos.
4518 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4519 insns where regno is used.
4521 2018-03-16 Martin Liska <mliska@suse.cz>
4524 * multiple_target.c (create_dispatcher_calls): Redirect
4525 reference in the symbol table.
4527 2018-03-16 Martin Liska <mliska@suse.cz>
4530 * multiple_target.c (create_dispatcher_calls): Redirect also
4533 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4537 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4538 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4539 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4541 2018-03-16 Julia Koval <julia.koval@intel.com>
4543 * doc/invoke.texi (Skylake Server): Add CLWB.
4544 Cannonlake): Remove CLWB.
4546 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4548 PR tree-optimization/84841
4549 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4551 (FLOAT_ONE_CONST_TYPE): Define.
4552 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4553 (sort_by_operand_rank): Put entries with higher constant_type last
4554 rather than first to match comments.
4556 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4558 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4561 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4564 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4565 check and use address of target always.
4567 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4570 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4571 (indirect_thunk_bnd_needed): Likewise.
4572 (indirect_thunks_used): Likewise.
4573 (indirect_thunks_bnd_used): Likewise.
4574 (indirect_return_needed): New.
4575 (indirect_return_bnd_needed): Likewise.
4576 (output_indirect_thunk_function): Add a bool argument for
4578 (output_indirect_thunk_function): Don't generate alias for
4579 function return thunk.
4580 (ix86_code_end): Call output_indirect_thunk_function to generate
4581 function return thunks.
4582 (ix86_output_function_return): Set indirect_return_bnd_needed
4583 and indirect_return_needed instead of indirect_thunk_bnd_needed
4584 and indirect_thunk_needed.
4586 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4588 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4589 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4590 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4592 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4593 Paul Hua <paul.hua.gm@gmail.com>
4596 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4598 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4600 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4601 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4604 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4607 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4608 instead of GET_MODE_SIZE when comparing Units.
4610 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4613 * varasm.c (hash_section): Return an unchangeble hash value
4614 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4615 Return !aarch64_can_use_per_function_literal_pools_p ().
4617 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4620 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4621 rather than address of cmode as last argument to prepare_cmp_insn.
4623 2018-03-15 Julia Koval <julia.koval@intel.com>
4625 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4626 F_AVX512VNNI, F_AVX512BITALG): New.
4628 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4631 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4632 insn for floating-point loads and stores.
4634 2018-03-14 Carl Love <cel@us.ibm.com>
4636 * config/rs6000/rs6000-c.c: Add macro definitions for
4637 ALTIVEC_BUILTIN_VEC_PERMXOR.
4638 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4639 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4640 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4641 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4643 * config/doc/extend.texi: Add prototypes for vec_permxor.
4645 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4648 * diagnostic-show-locus.c (class layout_point): Convert m_line
4649 from int to linenum_type.
4650 (line_span::comparator): Use linenum "compare" function when
4651 comparing line numbers.
4652 (test_line_span): New function.
4653 (layout_range::contains_point): Convert param "row" from int to
4655 (layout_range::intersects_line_p): Likewise.
4656 (layout::will_show_line_p): Likewise.
4657 (layout::print_source_line): Likewise.
4658 (layout::should_print_annotation_line_p): Likewise.
4659 (layout::print_annotation_line): Likewise.
4660 (layout::print_leading_fixits): Likewise.
4661 (layout::annotation_line_showed_range_p): Likewise.
4662 (struct line_corrections): Likewise for field m_row.
4663 (line_corrections::line_corrections): Likewise for param "row".
4664 (layout::print_trailing_fixits): Likewise.
4665 (layout::get_state_at_point): Likewise.
4666 (layout::get_x_bound_for_row): Likewise.
4667 (layout::print_line): Likewise.
4668 (diagnostic_show_locus): Likewise for locals "last_line" and
4670 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4671 * input.c (selftest::test_linenum_comparisons): New function.
4672 (selftest::input_c_tests): Call it.
4673 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4674 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4675 * selftest.h (ASSERT_GT): New macro.
4676 (ASSERT_GT_AT): New macro.
4677 (ASSERT_LT): New macro.
4678 (ASSERT_LT_AT): New macro.
4680 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4682 PR rtl-optimization/84780
4683 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4685 2018-03-14 Martin Liska <mliska@suse.cz>
4687 * tree.c (record_node_allocation_statistics): Use
4688 get_stats_node_kind.
4689 (get_stats_node_kind): New function extracted from
4690 record_node_allocation_statistics.
4691 (free_node): Use get_stats_node_kind.
4693 2018-03-14 Richard Biener <rguenther@suse.de>
4695 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4696 that the value-set of ANTIC_IN doesn't grow.
4699 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4701 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4702 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4704 2018-03-14 Julia Koval <julia.koval@intel.com>
4706 * config.gcc (icelake-client, icelake-server): New.
4708 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4709 (initial_ix86_arch_features): Ditto.
4710 (PTA_SKYLAKE): Add SGX.
4711 (PTA_ICELAKE): Remove.
4712 (PTA_ICELAKE_CLIENT): New.
4713 (PTA_ICELAKE_SERVER): New.
4714 (ix86_option_override_internal): Split up icelake on icelake client and
4716 (get_builtin_code_for_version): Ditto.
4717 (fold_builtin_cpu): Ditto.
4718 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4719 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4720 * config/i386/i386.h (processor_type): Ditto.
4721 * doc/invoke.texi: Ditto.
4723 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4726 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4727 INTEGER_CST offset, add it together with bitpos / 8 and
4728 sign extend based on POINTER_SIZE.
4732 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4735 * config/i386/constraints.md (Yc): New register constraint.
4736 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4737 Use Yc constraint for alternative 2 of operand 0. Remove
4738 preferred_for_speed attribute.
4740 2018-03-14 Richard Biener <rguenther@suse.de>
4742 PR tree-optimization/84830
4743 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4744 with the old one to avoid oscillations.
4746 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4749 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4751 (assign_by_spills): Return a flag of reload assignment failure.
4752 Do not process the reload assignment failures. Do not spill other
4753 reload pseudos if they has the same reg class. Update n if
4755 (lra_assign): Add a return arg. Set up from the result of
4756 assign_by_spills call.
4757 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4758 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4759 usage_insns if it is not NULL.
4760 (spill_hard_reg_in_range): New function.
4761 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4762 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4763 function prototypes.
4764 (lra_assign): Change prototype.
4765 * lra.c (lra): Add code to deal with fails by splitting hard reg
4768 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4770 * config/riscv/riscv.opt (mrelax): New option.
4771 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4772 "norelax" when riscv_mrelax is disabled.
4773 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4775 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4778 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4779 reassociation for int modes.
4781 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4783 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4784 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4786 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4787 * config/aarch64/aarch64-sve.md
4788 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4789 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4790 (*extend<mode><Vwide>2): Rename to...
4791 (aarch64_sve_extend<mode><Vwide>2): ...this.
4792 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4793 renaming the old pattern to...
4794 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4796 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4797 define_expand, renaming the old pattern to...
4798 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4799 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4800 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4801 account when deciding which SVE instruction the optab should use.
4802 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4804 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4806 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4807 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4808 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4809 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4810 (tlsdesc_small_<mode>): Turn a define_expand and use
4811 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4812 (tlsdesc_small_advsimd_<mode>): ...this.
4813 (tlsdesc_small_sve_<mode>): New pattern.
4815 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4817 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4818 (UNSPEC_UMUL_HIGHPART): New constants.
4819 (MUL_HIGHPART): New int iteraor.
4820 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4821 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4823 (*<su>mul<mode>3_highpart): New define_insn.
4825 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4828 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4831 2018-03-13 Martin Liska <mliska@suse.cz>
4834 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4836 (sem_item_optimizer::~sem_item_optimizer): Release it.
4837 (sem_item_optimizer::merge_classes): Register variable aliases.
4838 (sem_item_optimizer::fixup_pt_set): New function.
4839 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4840 * ipa-icf.h: Declare new variables and functions.
4842 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4845 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4846 integer_pow2p@2 and test integer_pow2p in condition.
4847 (A < 0 ? C : 0): Similarly for @1.
4850 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4851 characters starting at p contain '\0' character, don't look beyond
4855 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4856 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4859 * reg-stack.c (change_stack): Change update_end var from int to
4860 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4861 also call set_block_for_insn on the newly added insns and rescan.
4864 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4865 on the last operand.
4868 * tree.c (stabilize_reference_1): Return save_expr (e) for
4869 STATEMENT_LIST even if it doesn't have side-effects.
4871 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4873 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4875 2018-03-12 Renlin Li <renlin.li@arm.com>
4877 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4878 aarch64_output_scalar_simd_mov_immediate.
4880 2018-03-12 Martin Sebor <msebor@redhat.com>
4882 PR tree-optimization/83456
4883 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4884 for perfectly overlapping calls to memcpy.
4885 (gimple_fold_builtin_memory_chk): Same.
4886 (gimple_fold_builtin_strcpy): Handle no-warning.
4887 (gimple_fold_builtin_stxcpy_chk): Same.
4888 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4890 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4892 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4893 parameter. Use it for SFmode.
4894 (rs6000_function_arg_advance_1): Adjust.
4895 (rs6000_function_arg): Adjust.
4896 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4898 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4900 PR rtl-optimization/84169
4901 PR rtl-optimization/84780
4902 * combine.c (can_combine_p): Check for a 2-insn combination whether
4903 the destination register is used between the two insns, too.
4905 2018-03-12 Richard Biener <rguenther@suse.de>
4907 PR tree-optimization/84803
4908 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4909 for refs DR analysis didn't process.
4911 2018-03-12 Richard Biener <rguenther@suse.de>
4913 PR tree-optimization/84777
4914 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4915 force-vectorize loops ignore whether we are optimizing for size.
4917 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4919 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4920 (TARGET_MD_ASM_ADJUST): Define.
4922 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4923 Kito Cheng <kito.cheng@gmail.com>
4924 Chung-Ju Wu <jasonwucj@gmail.com>
4926 * config/nds32/nds32.c (nds32_compute_stack_frame,
4927 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4928 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4929 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4930 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4931 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4932 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4933 * config/nds32/nds32.md (prologue, epilogue): Use macro
4934 NDS32_V3PUSH_AVAILABLE_P to do checking.
4936 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4939 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4940 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4941 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4942 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4943 addition of most attributes on !orig_type_die or the attribute not
4944 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4946 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4947 Chung-Ju Wu <jasonwucj@gmail.com>
4949 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4951 * config/nds32/nds32.opt (mvh): New option.
4953 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4954 Chung-Ju Wu <jasonwucj@gmail.com>
4956 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4958 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4959 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4962 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4963 Chung-Ju Wu <jasonwucj@gmail.com>
4965 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4967 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4968 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4970 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4971 Kito Cheng <kito.cheng@gmail.com>
4972 Chung-Ju Wu <jasonwucj@gmail.com>
4974 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4975 UNSPEC_FFMISM and UNSPEC_FLMISM.
4976 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4977 for ffb, ffmism and flmism.
4978 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4979 (unspec_ffmism): Ditto.
4980 (unspec_flmism): Ditto.
4981 (nds32_expand_builtin_impl): Check if string extension is available.
4982 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4983 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4985 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4988 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4991 * lra-assigns.c (assign_by_spills): Return a flag of reload
4992 assignment failure. Do not process the reload assignment
4993 failures. Do not spill other reload pseudos if they has the same
4995 (lra_assign): Add a return arg. Set up from the result of
4996 assign_by_spills call.
4997 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4998 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4999 usage_insns if it is not NULL.
5000 (spill_hard_reg_in_range): New function.
5001 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5002 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5003 function prototypes.
5004 (lra_assign): Change prototype.
5005 * lra.c (lra): Add code to deal with fails by splitting hard reg
5008 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5011 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5013 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5016 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5017 (dw_val_node): Add val_symbolic_view.
5018 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5019 (symview_upper_bound): New.
5020 (new_line_info_table): Initialize symviews_since_reset.
5021 (dwarf2out_source_line): Count symviews_since_reset and set
5022 symview_upper_bound.
5023 (dw_val_equal_p): Handle symview.
5024 (add_AT_symview): New.
5025 (print_dw_val): Handle symview.
5026 (attr_checksum, attr_checksum_ordered): Likewise.
5027 (same_dw_val_p, size_of_die): Likewise.
5028 (value_format, output_die): Likewise.
5029 (add_high_low_attributes): Use add_AT_symview for entry_view.
5030 (dwarf2out_finish): Reset symview_upper_bound, clear
5033 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5036 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5037 Add strict argument and use it.
5038 (rs6000_split_multireg_move): Update for new strict argument.
5039 (mem_operand_gpr): Disallow all non-offsettable addresses.
5040 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5042 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5045 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5046 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5047 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5050 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5051 decl, use remap_type if we want to use the type.
5053 2018-03-09 Martin Sebor <msebor@redhat.com>
5055 PR tree-optimization/84526
5056 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5058 (builtin_access::generic_overlap): Be prepared to handle non-array
5061 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5063 PR rtl-optimization/84682
5064 * lra-constraints.c (process_address_1): Check is_address flag
5065 for address constraints.
5066 (process_alt_operands): Likewise.
5067 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5068 preprocess_constraints.
5069 * recog.h (preprocess_constraints): Add oploc parameter.
5071 * recog.c (preprocess_constraints): Test address_operand for
5072 CT_ADDRESS constraints.
5074 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5077 * lra-assigns.c (assign_by_spills): Return a flag of reload
5078 assignment failure. Do not process the reload assignment
5079 failures. Do not spill other reload pseudos if they has the same
5081 (lra_assign): Add a return arg. Set up from the result of
5082 assign_by_spills call.
5083 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5084 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5085 usage_insns if it is not NULL.
5086 (spill_hard_reg_in_range): New function.
5087 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5088 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5089 function prototypes.
5090 (lra_assign): Change prototype.
5091 * lra.c (lra): Add code to deal with fails by splitting hard reg
5094 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5097 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5098 Accept complain bool parameter. Only emit errors if it is true.
5099 (arm_parse_cpu_option_name): Likewise.
5100 (arm_target_thumb_only): Adjust callers of the above.
5101 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5102 prototype to take a default true bool parameter.
5103 (arm_parse_arch_option_name): Likewise.
5105 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5106 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5110 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5111 * configure: Regenerate.
5112 * configure.ac ("linker --version-script option"): New.
5113 ("linker soname option"): New.
5115 2018-03-09 Richard Biener <rguenther@suse.de>
5117 PR tree-optimization/84775
5118 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5119 immediate uses of predicate stmts and mark them modified.
5122 PR tree-optimization/84178
5123 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5125 (version_loop_for_if_conversion): Delay update_ssa call.
5126 (tree_if_conversion): Delay update_ssa until after predicate
5129 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5132 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5133 when the function accesses prior frames.
5135 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5138 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5139 gen_llsym, otherwise call maybe_gen_llsym.
5142 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5143 has ',' character inside of it.
5145 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5148 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5149 as clobbering CC_REGNUM.
5151 2018-03-08 Richard Biener <rguenther@suse.de>
5154 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5155 (follow_copies_to_constant): Do not follow SSA names registered
5158 2018-03-08 Richard Biener <rguenther@suse.de>
5160 PR tree-optimization/84178
5161 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5163 (version_loop_for_if_conversion): Delay update_ssa call.
5164 (tree_if_conversion): Delay update_ssa until after predicate
5167 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5169 PR tree-optimization/84178
5170 * tree-if-conv.c (release_bb_predicate): Remove the
5171 the assertion that the stmts have NULL use_ops.
5172 Discard the statements, asserting that they haven't
5173 yet been added to a BB.
5175 2018-03-08 Richard Biener <rguenther@suse.de>
5177 PR tree-optimization/84746
5178 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5179 (phi_translate): Pass in destination ANTIC_OUT set.
5180 (phi_translate_1): Likewise. For a simplified result lookup
5181 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5182 (phi_translate_set): Adjust.
5183 (do_pre_regular_insertion): Likewise.
5184 (do_pre_partial_partial_insertion): Likewise.
5186 2018-03-08 Martin Liska <mliska@suse.cz>
5188 PR gcov-profile/84735
5189 * doc/gcov.texi: Document usage of profile files.
5190 * gcov-io.h: Document changes in the format.
5192 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5196 * dwarf2out.c (struct dw_line_info_table): Update comments for
5198 (FORCE_RESET_NEXT_VIEW): New.
5199 (FORCE_RESETTING_VIEW_P): New.
5200 (RESETTING_VIEW_P): Check for -1 too.
5201 (ZERO_VIEW_P): Likewise.
5202 (new_line_info_table): Force-reset next view.
5203 (dwarf2out_begin_function): Likewise.
5204 (dwarf2out_source_line): Simplify zero_view_p initialization.
5205 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5206 view directly. Omit view when omitting .loc at line 0.
5208 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5210 PR tree-optimization/84740
5211 * tree-switch-conversion.c (process_switch): Call build_constructors
5212 only if info.phi_count is non-zero.
5214 PR tree-optimization/84739
5215 * tree-tailcall.c (find_tail_calls): Check call arguments against
5216 DECL_ARGUMENTS (current_function_decl) rather than
5217 DECL_ARGUMENTS (func) when checking for tail recursion.
5219 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5221 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5222 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5223 Volker Reichelt's entry and add entries for people that perform
5224 GCC fuzzy testing and report numerous bugs.
5226 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5229 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5230 readonly data in sdata, if that is disabled.
5231 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5232 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5233 -mreadonly-in-sdata option.
5235 2018-03-07 Martin Sebor <msebor@redhat.com>
5237 PR tree-optimization/84468
5238 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5239 basic block when looking for nul assignment.
5241 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5244 * except.h (output_function_exception_table): Adjust prototype.
5245 * except.c (output_function_exception_table): Remove FNNAME parameter
5246 and add SECTION parameter. Ouput one part of the table at a time.
5247 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5248 the first part of the exception table and emit unwind directives.
5249 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5250 (i386_pe_seh_cold_init): Likewise.
5251 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5252 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5253 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5254 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5255 * config/i386/winnt.c: Include except.h.
5256 (struct seh_frame_state): Add reg_offset, after_prologue and
5257 in_cold_section fields.
5258 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5259 (i386_pe_seh_cold_init): New function.
5260 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5261 to seh->in_cold_section.
5262 (seh_emit_push): Record the offset of the push.
5263 (seh_emit_save): Record the offet of the save.
5264 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5265 Test seh->after_prologue to disregard the epilogue.
5266 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5267 (i386_pe_end_cold_function): New function.
5269 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5272 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5273 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5276 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5278 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5279 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5282 * multiple_target.c: Include tree-inline.h and intl.h.
5283 (expand_target_clones): Diagnose and fail if node->definition and
5284 !tree_versionable_function_p (node->decl).
5286 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5288 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5290 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5291 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5292 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5294 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5297 * combine.c (try_combine): Use reg_or_subregno instead of handling
5298 just paradoxical SUBREGs and REGs.
5300 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5302 * config/arc/arc.c (arc_finalize_pic): Remove function.
5303 (arc_must_save_register): We use single base PIC register, remove
5304 checks to save/restore the PIC register.
5305 (arc_expand_prologue): Likewise.
5306 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5308 (arc_verify_short): Likewise.
5309 (arc_attr_type): Likewise.
5310 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5311 (walk_stores): Likewise.
5312 (arc_address_cost): Make it static.
5313 (arc_verify_short): Likewise.
5314 (branch_dest): Likewise.
5315 (arc_attr_type): Likewise.
5316 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5317 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5318 (arc_final_prescan_insn): Remove inserting the nops due to
5319 hardware hazards. It is done in reorg step.
5320 (insn_length_variant_t): Remove.
5321 (insn_length_parameters_t): Likewise.
5322 (arc_insn_length_parameters): Likewise.
5323 (arc_get_insn_variants): Likewise.
5324 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5326 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5329 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5332 PR tree-optimization/84687
5333 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5335 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5337 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5339 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5340 Rename to ppc_speculation_barrier.
5341 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5342 __builtin_ppc_speculation_barrier.
5344 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5347 * combine.c (combine_simplify_rtx): Don't try to simplify if
5348 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5351 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5353 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5354 to 32 bytes when compiling for POWER9.
5356 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5359 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5360 regparm >= 3 with no arg reg available also for calls with
5361 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5364 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5366 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5368 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5371 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5373 2018-03-05 Richard Biener <rguenther@suse.de>
5375 PR tree-optimization/84486
5376 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5377 When inserting a __builtin_assume_aligned call set the LHS
5378 SSA name alignment info accordingly.
5380 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5382 PR tree-optimization/84114
5383 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5384 Avoid reassociation of FLOAT_MODE addition.
5386 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5388 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5389 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5390 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5391 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5392 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5393 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5394 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5396 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5397 __builtin_ia32_wbinvd): New builtins.
5398 (SPECIAL_ARGS2): New.
5399 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5400 (SPECIAL_ARGS2): New.
5401 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5402 (ix86_valid_target_attribute_inner_p): Ditto.
5403 (ix86_init_mmx_sse_builtins): Add special_args2.
5404 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5405 TARGET_WBNOINVD_P): New.
5406 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5407 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5408 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5409 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5410 * config/i386/pconfigintrin.h: New file.
5411 * config/i386/wbnoinvdintrin.h: Ditto.
5412 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5413 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5415 2018-03-05 Richard Biener <rguenther@suse.de>
5417 PR tree-optimization/84670
5418 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5420 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5421 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5422 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5423 doesn't grow if all successors (recursively) were visited at least
5426 2018-03-05 Richard Biener <rguenther@suse.de>
5428 PR tree-optimization/84650
5429 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5430 if executed in the loop pipeline.
5432 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5434 * doc/configfiles.texi (Configuration Files): Move info about
5435 conditionalizing $target-protos.h to...
5436 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5437 differs from $target-protos.h.
5439 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5440 Chung-Ju Wu <jasonwucj@gmail.com>
5442 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5443 * config/nds32/nds32-multiple.md (setmemsi): Define.
5444 * config/nds32/nds32-memory-manipulation.c
5445 (nds32_gen_dup_4_byte_to_word_value): New.
5446 (emit_setmem_word_loop): New.
5447 (emit_setmem_byte_loop): New.
5448 (nds32_expand_setmem_loop): New.
5449 (nds32_expand_setmem_loop_v3m): New.
5450 (nds32_expand_setmem_unroll): New.
5451 (nds32_expand_setmem): New.
5453 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5454 Chung-Ju Wu <jasonwucj@gmail.com>
5456 * config/nds32/nds32-memory-manipulation.c
5457 (nds32_emit_load_store): New.
5458 (nds32_emit_post_inc_load_store): New.
5459 (nds32_emit_mem_move): New.
5460 (nds32_emit_mem_move_block): New.
5461 (nds32_expand_movmemsi_loop_unknown_size): New.
5462 (nds32_expand_movmemsi_loop_known_size): New.
5463 (nds32_expand_movmemsi_loop): New.
5464 (nds32_expand_movmemsi_unroll): New.
5465 (nds32_expand_movmemqi): Rename ...
5466 (nds32_expand_movmemsi): ... to this.
5467 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5468 (movmemsi): ... to this.
5469 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5470 (nds32_expand_movmemsi): ... to this.
5472 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5473 Monk Chiang <sh.chiang04@gmail.com>
5474 Chung-Ju Wu <jasonwucj@gmail.com>
5476 * config/nds32/nds32-protos.h
5477 (nds32_expand_load_multiple): New arguments.
5478 (nds32_expand_store_multiple): Ditto.
5479 (nds32_valid_multiple_load_store): Rename ...
5480 (nds32_valid_multiple_load_store_p): ... to this.
5481 * config/nds32/nds32-memory-manipulation.c
5482 (nds32_expand_load_multiple): Refine implementation.
5483 (nds32_expand_store_multiple): Ditto.
5484 * config/nds32/nds32-multiple.md
5485 (load_multiple): Update nds32_expand_load_multiple interface.
5486 (store_multiple): Update nds32_expand_store_multiple interface.
5487 * config/nds32/nds32-predicates.c
5488 (nds32_valid_multiple_load_store): Rename ...
5489 (nds32_valid_multiple_load_store_p): ... to this and refine
5491 * config/nds32/predicates.md
5492 (nds32_load_multiple_and_update_address_operation): New predicate.
5493 (nds32_store_multiple_and_update_address_operation): New predicate.
5495 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5496 Chung-Ju Wu <jasonwucj@gmail.com>
5498 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5499 (combo): New attribute.
5500 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5502 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5504 * config/nds32/nds32.opt: Change -mcmodel= default value.
5506 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5507 Monk Chiang <sh.chiang04@gmail.com>
5508 Chung-Ju Wu <jasonwucj@gmail.com>
5510 * config/nds32/constants.md (unspec_element): New enum.
5511 * config/nds32/constraints.md (Umw): New constraint.
5512 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5513 * config/nds32/nds32-intrinsic.md: Likewise.
5514 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5515 (nds32_valid_smw_lwm_base_p): New.
5516 (nds32_output_smw_single_word): New.
5517 (nds32_output_lmw_single_word): New.
5518 (nds32_expand_unaligned_load): New.
5519 (nds32_expand_unaligned_store): New.
5520 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5521 (nds32_output_smw_single_word): Declare.
5522 (nds32_output_lmw_single_word): Declare.
5523 (nds32_expand_unaligned_load): Declare.
5524 (nds32_expand_unaligned_store): Declare.
5525 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5526 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5527 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5528 NDS32_BUILTIN_UASTORE_DW.
5529 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5532 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5533 Kito Cheng <kito.cheng@gmail.com>
5534 Chung-Ju Wu <jasonwucj@gmail.com>
5536 * config/nds32/nds32-intrinsic.c
5537 (nds32_expand_builtin_null_ftype_reg): Delete.
5538 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5539 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5540 (nds32_read_argument): New.
5541 (nds32_legitimize_target): Ditto.
5542 (nds32_legitimize_argument): Ditto.
5543 (nds32_check_constant_argument): Ditto.
5544 (nds32_expand_unop_builtin): Ditto.
5545 (nds32_expand_unopimm_builtin): Ditto.
5546 (nds32_expand_binop_builtin): Ditto.
5547 (nds32_builtin_decl_impl): Ditto.
5548 (builtin_description): Ditto.
5549 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5550 (nds32_init_builtins_impl): Ditto.
5551 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5552 (nds32_builtin_decl): New.
5553 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5554 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5556 2018-03-02 Jeff Law <law@redhat.com>
5558 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5559 (redundant_insn, fill_simple_delay_slots): Likewise.
5560 (fill_slots_from_thread): Likewise.
5561 * resource.c (mark_referenced_resources): Likewise.
5562 (mark_set_resources, find_dead_or_set_registers): Likewise.
5564 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5566 * substring-locations.h (format_warning_va): Formatting fix for
5568 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5570 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5571 * substring-locations.c: Include intl.h.
5572 (format_warning_va): Turned into small wrapper around
5573 format_warning_n_va, renamed to ...
5574 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5575 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5577 (format_warning_at_substring_n): New function.
5578 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5579 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5580 format_warning_at_substring with just a shorter name instead of
5581 const function pointer.
5582 (fmtwarn_n): New function.
5583 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5584 appropriate, get rid of all the fmtstr temporaries, move conditionals
5585 with G_() wrapped string literals directly into fmtwarn arguments,
5586 cast dir.len to (int), formatting fixes.
5588 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5590 * doc/invoke.texi: Remove "Cilk Plus" references.
5592 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5593 Richard Biener <rguenther@suse.de>
5596 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5597 for error or warning attributes if CALL_FROM_THUNK_P is set.
5600 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5603 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5604 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5607 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5608 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5610 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5611 instead of -1U in last predictors element's probability member.
5613 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5616 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5617 arguments if they are comparable.
5619 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5621 PR tree-optimization/84634
5622 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5623 masks and masked_loop_p with a single loop_masks, making sure it's
5624 null for bb vectorization.
5626 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5628 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5629 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5630 loop->force_vectorize to check whether there is no alias.
5632 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5635 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5637 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5639 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5640 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5641 instead of a loop around prev_real_insn.
5642 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5646 * config/i386/i386.c (ix86_print_operand): Use conditional
5647 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5650 2018-03-02 Richard Biener <rguenther@suse.de>
5652 PR tree-optimization/84427
5653 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5654 (bitmap_set_subtract_values): Rewrite to handle multiple
5657 (prune_clobbered_mems): Likewise.
5658 (phi_translate): Take edge instead of pred/phiblock.
5659 (phi_translate_1): Likewise.
5660 (phi_translate_set): Likewise. Insert all translated
5661 exprs for a value into the set, keeping possibly multiple
5662 expressions per value.
5663 (compute_antic_aux): Adjust for phi_translate changes.
5664 When intersecting union the expressions and prune those
5665 not in the final value set, keeping possibly multiple
5666 expressions per value. Do not use value-insertion
5667 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5668 all expressions. Add verification that the value-sets
5669 only shrink during iteration.
5670 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5671 (do_pre_regular_insertion): Likewise.
5672 (do_pre_partial_partial_insertion): Likewise.
5674 2018-03-02 Richard Biener <rguenther@suse.de>
5677 * config/darwin.c (saved_debug_info_level): New static global.
5678 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5679 (darwin_asm_lto_end): Restore debug info generation settings.
5681 2018-03-01 Martin Liska <mliska@suse.cz>
5684 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5687 2018-03-01 Richard Biener <rguenther@suse.de>
5690 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5691 type annotation with LTO when debug was disabled at compile-time.
5693 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5695 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5697 (mips_final_postscan_insn): Likewise.
5699 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5701 PR rtl-optimization/84528
5702 * alias.c (init_alias_target): Add commentary.
5703 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5704 a unique base value if the frame pointer is not eliminated
5705 to the stack pointer.
5707 2018-03-01 Tom de Vries <tom@codesourcery.com>
5709 PR rtl-optimization/83327
5710 * lra-int.h (hard_regs_spilled_into): Declare.
5711 * lra.c (hard_regs_spilled_into): Define.
5712 (init_reg_info): Init hard_regs_spilled_into.
5713 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5714 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5715 (process_bb_lives): Handle hard_regs_spilled_into.
5716 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5717 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5719 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5721 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5722 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5723 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5724 * config/rs6000/aix72.h: New file.
5726 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5728 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5729 instead of warning_at with conditional singular and plural messages
5733 * stor-layout.c (update_alignment_for_field): For
5734 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5735 && !DECL_PACKED (field), do the alignment update, just use
5736 only desired_align instead of MAX (type_align, desired_align)
5738 (place_field): Don't do known_align < desired_align handling
5739 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5740 is non-NULL, instead do it after rli->prev_field handling and
5741 only if not within a bitfield word. For DECL_PACKED (field)
5742 use type_align of BITS_PER_UNIT.
5744 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5746 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5747 superfluous parentheses and trailing spaces.
5749 2018-02-28 Richard Biener <rguenther@suse.de>
5751 PR tree-optimization/84584
5752 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5753 SCoPs with fake exit edge.
5755 2018-02-28 Martin Liska <mliska@suse.cz>
5758 * timevar.c (timer::print): Fix format to properly print 100%
5761 2018-02-28 Richard Biener <rguenther@suse.de>
5764 * genmatch.c (capture_info::walk_match): Do not mark
5765 captured expressions without operands as expr_p given
5766 they act more like predicates and should be subject to
5767 "lost tail" side-effect preserving.
5769 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5771 PR rtl-optimization/81611
5772 * auto-inc-dec.c (attempt_change): Move dead note from
5773 mem_insn if it's the next use of regno
5774 (find_address): Take address use of reg holding
5775 non-incremented value. Add parm to limit search to the named
5777 (merge_in_block): Attempt to use a mem insn that is the next
5778 use of the original regno.
5780 2018-02-27 Martin Sebor <msebor@redhat.com>
5783 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5784 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5786 2018-02-27 Martin Sebor <msebor@redhat.com>
5788 PR translation/84207
5789 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5790 n argument to unsigned HOST_WIDE_INT.
5791 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5792 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5793 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5794 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5796 2018-02-27 Richard Biener <rguenther@suse.de>
5798 PR tree-optimization/84512
5799 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5800 Do not use the estimate returned from record_stmt_cost for
5801 the scalar iteration cost but sum properly using add_stmt_cost.
5803 2018-02-27 Richard Biener <rguenther@suse.de>
5805 PR tree-optimization/84466
5806 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5807 Adjust last change to less strictly validate use operands.
5809 2018-02-27 Martin Liska <mliska@suse.cz>
5811 PR gcov-profile/84548
5812 * gcov.c (process_file): Allow partial overlap and consider it
5813 also as group functions.
5814 (output_lines): Properly calculate range of lines for a group.
5816 2018-02-27 Martin Liska <mliska@suse.cz>
5818 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5819 'ggc' suffixes. Change first column width.
5820 (timer::print): Fix formatting of the column.
5822 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5824 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5825 preserve inline entry blocks for the sake of debug inline
5826 entry point markers alone.
5827 (remove_unused_locals): Suggest in comments a better place to
5828 force the preservation of inline entry blocks that are
5829 otherwise unused, but do not preserve them.
5831 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5833 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5835 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5838 * config/i386/constraints.md (Bs): Replace
5839 ix86_indirect_branch_register with
5840 TARGET_INDIRECT_BRANCH_REGISTER.
5842 * config/i386/i386.md (indirect_jump): Likewise.
5843 (tablejump): Likewise.
5844 (*sibcall_memory): Likewise.
5845 (*sibcall_value_memory): Likewise.
5846 Peepholes of indirect call and jump via memory: Likewise.
5847 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5848 (*sibcall_value_GOT_32): Likewise.
5849 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5850 (GOT_memory_operand): Likewise.
5851 (call_insn_operand): Likewise.
5852 (sibcall_insn_operand): Likewise.
5853 (GOT32_symbol_operand): Likewise.
5854 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5856 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5858 PR rtl-optimization/83496
5859 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5860 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5861 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5862 redundant insn, if any.
5863 (relax_delay_slots): Likewise.
5864 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5866 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5868 PR tree-optimization/83965
5869 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5870 that grouped statements are part of a reduction chain. Return
5871 true if the statement is not marked as a reduction itself but
5873 (vect_recog_dot_prod_pattern): Don't check whether the statement
5874 is part of a group here.
5875 (vect_recog_sad_pattern): Likewise.
5876 (vect_recog_widen_sum_pattern): Likewise.
5878 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5881 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5883 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5886 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5888 (ix86_output_indirect_function_return): New prototype.
5889 (ix86_split_simple_return_pop_internal): Likewise.
5890 * config/i386/i386.c (indirect_return_via_cx): New.
5891 (indirect_return_via_cx_bnd): Likewise.
5892 (indirect_thunk_name): Handle return va CX_REG.
5893 (output_indirect_thunk_function): Create alias for
5894 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5895 (ix86_output_indirect_jmp): Remove the bool argument.
5896 (ix86_output_indirect_function_return): New function.
5897 (ix86_split_simple_return_pop_internal): Likewise.
5898 * config/i386/i386.md (*indirect_jump): Don't pass false
5899 to ix86_output_indirect_jmp.
5900 (*tablejump_1): Likewise.
5901 (simple_return_pop_internal): Change it to define_insn_and_split.
5902 Call ix86_split_simple_return_pop_internal to split it for
5904 (simple_return_indirect_internal): Call
5905 ix86_output_indirect_function_return instead of
5906 ix86_output_indirect_jmp.
5908 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5911 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5912 memset and value initialization afterwards.
5914 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5916 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5918 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5921 * common/config/aarch64/aarch64-common.c
5922 (aarch_option_optimization_table[]): Switch
5923 off fomit-frame-pointer
5925 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5926 Chung-Ju Wu <jasonwucj@gmail.com>
5928 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5930 (store_multiple): Ditto.
5932 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5934 * config.gcc: Add --with-cpu support for nds32 target.
5935 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5936 * config/nds32/nds32.opt: Add -mcpu= option.
5938 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5940 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5941 isel=yes): Warn for these deprecated options.
5943 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5945 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5946 ISA_2_5_MASKS_EMBEDDED.
5948 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5950 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5951 p->max as pointers rather than using iterative_hash_expr.
5953 2018-02-23 Carl Love <cel@us.ibm.com>
5955 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5956 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5958 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5959 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5960 P8V_BUILTIN_VEC_VUNSIGNED2.
5962 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5965 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5966 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5968 * lra-constraints.c (curr_insn_transform): Set up
5969 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5971 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5972 * lra-eliminations.c (spill_pseudos): Ditto.
5973 (process_insn_for_elimination): Ditto.
5974 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5975 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5976 LRA_NON_CLOBBERED_ALT.
5978 2018-02-22 Martin Sebor <msebor@redhat.com>
5980 PR tree-optimization/84480
5981 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5982 to maybe_diag_stxncpy_trunc. Call it.
5983 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5984 from gimple_fold_builtin_strcpy. Print inlining stack.
5985 (handle_builtin_stxncpy): Print inlining stack.
5986 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5988 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5991 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5992 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5993 and -fcheck-pointer-bounds are used together.
5994 (indirect_thunk_prefix): New enum.
5995 (indirect_thunk_need_prefix): New function.
5996 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5997 "_nt" instead of "_bnd" for NOTRACK prefix.
5998 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5999 (output_indirect_thunk_function): Likewise.
6001 (ix86_code_end): Update output_indirect_thunk_function calls.
6002 (ix86_output_indirect_branch_via_reg): Replace
6003 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6004 (ix86_output_indirect_branch_via_push): Likewise.
6005 (ix86_output_function_return): Likewise.
6006 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6007 incompatible with -fcf-protection=branch and
6008 -fcheck-pointer-bounds.
6010 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6013 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6014 Change gcc_assert call to output_operand_lossage.
6016 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6018 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6020 2018-02-22 DJ Delorie <dj@redhat.com>
6021 Sebastian Perta <sebastian.perta@renesas.com>
6022 Oleg Endo <olegendo@gcc.gnu.org>
6024 * config/rx/rx.c (rx_rtx_costs): New function.
6025 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6027 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6029 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6031 2018-02-22 Martin Liska <mliska@suse.cz>
6034 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6035 Add "native" as a possible value.
6037 2018-02-22 Martin Liska <mliska@suse.cz>
6040 * config/i386/i386.c (ix86_option_override_internal):
6041 Add "native" as a possible value for -march and -mtune.
6043 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6046 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6047 to all type variants.
6049 PR tree-optimization/84503
6050 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6051 width as info->bitpos + info->bitsize - start.
6052 (merged_store_group::merge_overlapping): Simplify width computation.
6053 (check_no_overlap): New function.
6054 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6055 start + width and last_order of the group, fail if check_no_overlap
6057 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6058 to group if check_no_overlap fails.
6060 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6062 * config/rs6000/altivec.md: Delete contraint arguments to
6063 define_expand, define_split, and define_peephole2, and in
6064 define_insn_and_split if always unused.
6065 * config/rs6000/darwin.md: Ditto.
6066 * config/rs6000/dfp.md: Ditto.
6067 * config/rs6000/rs6000.md: Ditto.
6068 * config/rs6000/sync.md: Ditto.
6069 * config/rs6000/vector.md: Ditto.
6070 * config/rs6000/vsx.md: Ditto.
6072 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6074 * config/rs6000/altivec.md: Write output control strings as braced
6075 blocks instead of double-quoted strings.
6076 * config/rs6000/darwin.md: Ditto.
6077 * config/rs6000/rs6000.md: Ditto.
6078 * config/rs6000/vector.md: Ditto.
6079 * config/rs6000/vsx.md: Ditto.
6081 2018-02-21 Jason Merrill <jason@redhat.com>
6083 PR c++/84314 - ICE with templates and fastcall attribute.
6084 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6086 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6088 * ipa-cp.c (determine_versionability): Fix comment typos.
6090 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6093 * ipa-cp.c (determine_versionability): Do not version functions caling
6096 2018-02-21 Martin Liska <mliska@suse.cz>
6099 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6100 Add "native" as a possible value.
6101 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6102 the macro when native cpu detection is available.
6104 2018-02-21 Martin Liska <mliska@suse.cz>
6107 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6108 Add "native" as a possible value.
6109 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6110 when native cpu detection is available.
6112 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6113 Martin Sebor <msebor@redhat.com>
6115 PR tree-optimization/84478
6116 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6118 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6119 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6120 support which is conservatively correct, for 2 only stay conservative
6121 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6122 argument to the 2 argument get_range_strlen, adjust 6 arg
6123 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6125 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6126 (gimple_fold_builtin_strlen): Pass true as last argument to
6129 2018-02-20 Martin Sebor <msebor@redhat.com>
6132 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6133 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6134 (builtin_memref::builtin_memref): Factor out parts into
6135 set_base_and_offset and call it.
6137 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6140 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6141 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6142 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6143 search at the associated MODE_INT.
6145 2018-02-20 Jeff Law <law@redhat.com>
6148 PR tree-optimization/81592
6150 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6151 for range data rather than using global data.
6152 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6153 range data rather than using global data.
6154 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6155 pass it to children as needed.
6156 (struct directive::fmtresult): Similarly.
6157 (struct directive::set_width): Similarly.
6158 (struct directive::set_precision): Similarly.
6159 (format_integer, format_directive, parse_directive): Similarly.
6160 (format_none): Accept unnamed vr_values parameter.
6161 (format_percent, format_floating, format_character): Similarly.
6162 (format_string, format_plain): Similarly.
6163 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6164 the EVRP range analyzer for range data rather than using global data.
6165 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6166 gimple-ssa-evrp-analyze.h
6167 (class sprintf_dom_walker): Add after_dom_children member function.
6168 Add evrp_range_analyzer member.
6169 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6170 range analyzer as needed.
6171 (sprintf_dom_walker::after_dom_children): New member function.
6172 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6174 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6175 (evrp_range_analyzer::pop_to_marker): Likewise.
6177 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6179 PR tree-optimization/84419
6180 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6181 with the required type if its current type is compatible but
6184 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6187 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6188 after vectorization.
6190 2018-02-20 Martin Liska <mliska@suse.cz>
6193 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6194 possible values if we don't have a hint.
6196 2018-02-20 Martin Liska <mliska@suse.cz>
6200 * final.c (shorten_branches): Build align_tab array with one
6202 * opts.c (finish_options): Add alignment option limit check.
6203 (MAX_CODE_ALIGN): Likewise.
6204 (MAX_CODE_ALIGN_VALUE): Likewise.
6205 * doc/invoke.texi: Document maximum allowed option value for
6206 all -falign-* options.
6208 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6211 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6212 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6213 * var-tracking.c (emit_note_insn_var_location): Remove all references
6214 to NOTE_INSN_CALL_ARG_LOCATION.
6215 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6216 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6217 Use copy_rtx_if_shared.
6218 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6219 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6220 (dwarf2out_var_location): Remove handling of
6221 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6223 * final.c (final_scan_insn): Remove all references to
6224 NOTE_INSN_CALL_ARG_LOCATION.
6225 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6226 before dumping final insns.
6227 * except.c (emit_note_eh_region_end): Remove all references to
6228 NOTE_INSN_CALL_ARG_LOCATION.
6229 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6230 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6231 * config/arc/arc.c (hwloop_optimize): Likewise.
6232 * config/arm/arm.c (create_fix_barrier): Likewise.
6233 * config/s390/s390.c (s390_chunkify_start): Likewise.
6234 * config/sh/sh.c (find_barrier): Likewise.
6235 * config/i386/i386.c (rest_of_insert_endbranch,
6236 ix86_seh_fixup_eh_fallthru): Likewise.
6237 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6238 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6239 * config/frv/frv.c (frv_function_prologue): Likewise.
6240 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6242 (note_outside_basic_block_p): Remove all references to
6243 NOTE_INSN_CALL_ARG_LOCATION.
6244 * gengtype.c (adjust_field_rtx_def): Likewise.
6245 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6247 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6248 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6251 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6254 PR tree-optimization/84452
6255 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6256 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6259 2018-02-19 Martin Liska <mliska@suse.cz>
6262 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6264 2018-02-19 Martin Liska <mliska@suse.cz>
6265 Richard Sandiford <richard.sandiford@linaro.org>
6267 PR tree-optimization/82491
6268 * gimple-fold.c (get_base_constructor): Make earlier bail out
6271 2018-02-19 Carl Love <cel@us.ibm.com>
6273 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6274 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6276 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6277 P8V_BUILTIN_VEC_NEG.
6279 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6281 * config/rl78/rl78.md (movdf): New define expand.
6283 2018-02-19 Martin Liska <mliska@suse.cz>
6286 * doc/invoke.texi: Fix typo.
6287 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6289 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6291 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6292 handle rs6000_single_float and rs6000_double_float specially for
6295 2018-02-16 Jeff Law <law@redhat.com>
6297 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6300 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6303 * ipa-inline.c (inline_small_functions): Fix a typo.
6305 2018-02-16 Nathan Sidwell <nathan@acm.org>
6307 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6309 2018-02-16 Carl Love <cel@us.ibm.com>
6311 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6312 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6313 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6314 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6315 expansion to P8V_BUILTIN_VEC_FLOAT2.
6317 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6319 PR rtl-optimization/70023
6320 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6321 src_regno into account.
6323 2018-02-16 Carl Love <cel@us.ibm.com>
6325 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6326 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6327 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6328 * config/rs6000/rs6000.c: Remove case statements for
6329 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6330 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6331 and P9V_BUILTIN_VEC_VINSERT4B.
6332 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6333 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6334 * config/rs6000/vsx.md:
6335 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6338 2018-02-16 Carl Love <cel@us.ibm.com>
6340 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6342 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6344 * config/rs6000/rs6000-c.c: Add the definitions for
6345 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6346 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6347 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6348 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6349 definition for insert4b and define insn *insert3b_internal.
6350 * doc/extend.texi: Add documentation for vec_extract4b.
6352 2018-02-16 Nathan Sidwell <nathan@acm.org>
6354 * doc/extend.texi (Backwards Compatibility): Mention friend
6355 injection. Note for-scope is deprecated.
6356 * doc/invoke.texi (-ffriend-injection): Deprecate.
6358 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6360 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6361 that moved to I2, also allow destinations that are a paradoxical
6362 subreg (instead of a normal reg).
6364 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6367 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6370 2018-02-16 Richard Biener <rguenther@suse.de>
6372 PR tree-optimization/84037
6373 PR tree-optimization/84016
6375 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6376 Adjust vec_construct for the fact we need additional higher latency
6377 128bit inserts for AVX256 and AVX512 vector builds.
6378 (ix86_add_stmt_cost): Scale vector construction cost for
6381 2018-02-16 Richard Biener <rguenther@suse.de>
6383 PR tree-optimization/84417
6384 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6385 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6386 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6388 2018-02-16 Martin Liska <mliska@suse.cz>
6391 * internal-fn.def (ASAN_CHECK): Set proper flags.
6392 (ASAN_MARK): Likewise.
6394 2018-02-16 Julia Koval <julia.koval@intel.com>
6396 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6397 from PTA_CANNONLAKE.
6399 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6402 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6403 Use ++iter rather than iter++ for std::list iterators.
6404 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6405 defer deleting them until all nodes in the forest are processed. Do
6406 free even leaf nodes. Change to_process into auto_vec.
6409 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6410 * vec.h (vec_default_construct): Use memset instead of placement new
6411 if BROKEN_VALUE_INITIALIZATION is defined.
6412 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6413 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6416 PR rtl-optimization/83723
6417 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6418 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6419 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6421 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6423 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6425 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6427 PR rtl-optimization/81443
6428 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6429 from inner REGs to paradoxical SUBREGs.
6431 2018-02-16 Richard Biener <rguenther@suse.de>
6433 PR tree-optimization/84399
6434 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6435 For operands we can analyze at their definition make sure we can
6436 analyze them at each use as well.
6438 2018-02-16 Richard Biener <rguenther@suse.de>
6440 PR tree-optimization/84190
6441 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6442 volatile accesses if the decl isn't volatile.
6444 2018-02-15 Jason Merrill <jason@redhat.com>
6446 PR c++/84314 - ICE with templates and fastcall attribute.
6447 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6448 TYPE_CANONICAL on an existing type.
6450 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6452 PR tree-optimization/84383
6453 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6454 dstoff nor call operand_equal_p if dstbase is NULL.
6456 PR tree-optimization/84334
6457 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6458 also a CONSTANT_CLASS_P, punt.
6460 2018-02-14 Jim Wilson <jimw@sifive.com>
6462 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6463 first SMALL_OPERAND check. New local min_second_step. Move assert
6464 to where locals are set. Add TARGET_RVC support.
6465 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6467 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6469 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6471 2018-02-14 Martin Sebor <msebor@redhat.com>
6473 PR tree-optimization/83698
6474 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6475 arrays constrain the offset range to their bounds.
6476 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6477 (builtin_access::overlap): Avoid setting the size of overlap if it's
6479 (maybe_diag_overlap): Also consider arrays when deciding what values
6480 of offsets to include in diagnostics.
6482 2018-02-14 Martin Sebor <msebor@redhat.com>
6485 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6486 that correspond to the kind of a declaration.
6488 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6491 * config/pa/pa.md: Load address of PIC label using the linkage table
6492 if the label is nonlocal.
6494 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6496 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6497 warning message if user requests -maltivec=be.
6498 * doc/invoke.texi: Document deprecation of -maltivec=be.
6500 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6503 * config/rs6000/rs6000-c.c: Update definitions for
6504 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6505 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6507 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6510 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6511 add _get_ssp intrinsics. Remove argument from
6512 __builtin_ia32_rdssp[d|q].
6513 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6514 * config/i386/i386-builtin.def: Remove argument from
6515 __builtin_ia32_rdssp[d|q].
6516 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6517 ix86_expand_special_args_builtin for _rdssp[d|q].
6518 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6519 Clear register before usage.
6520 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6521 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6523 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6525 PR tree-optimization/84357
6526 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6527 operand 1 of an ARRAY_REF too.
6529 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6532 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6533 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6535 (set_of_reg): New struct.
6536 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6537 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6538 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6540 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6541 Split into bitclr, bitset, bitinvert patterns if appropriate.
6542 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6543 use rx_fuse_in_memory_bitop.
6544 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6545 to named insn, correct maximum insn length.
6547 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6550 * machmode.def: Define a complex mode for PARTIAL_INT.
6551 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6553 * doc/rtl.texi: Document CSPImode.
6554 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6556 (msp430_hard_regno_nregs_with_padding): Likewise.
6558 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6561 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6563 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6565 PR rtl-optimization/84169
6566 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6567 we generated a parallel as new i3 and we split that to new i2 and i3
6568 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6569 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6570 those to i2, not i1. Partially rewrite this scan code.
6572 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6575 * stor-layout.c (place_field): For variable length fields, adjust
6576 offset_align afterwards not just based on the field's alignment,
6577 but also on the size.
6580 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6581 of exps and logs in the use_exp2 case.
6583 2018-02-13 Jeff Law <law@redhat.com>
6585 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6588 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6591 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6595 * common.opt (gas-loc-support, gas-locview-support): New.
6596 (ginline-points, ginternal-reset-location-views): New.
6597 * doc/invoke.texi: Document them. Use @itemx where intended.
6598 (gvariable-location-views): Adjust.
6599 * target.def (reset_location_view): New.
6600 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6601 (TARGET_RESET_LOCATION_VIEW): New.
6602 * doc/tm.texi: Rebuilt.
6603 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6604 (dwarf2out_default_as_locview_support): New.
6605 (output_asm_line_debug_info): Use option variables.
6606 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6607 (output_loc_list): Likewise.
6608 (add_high_low_attributes): Check option variables.
6609 Don't output entry view attribute in strict mode.
6610 (gen_inlined_subroutine_die): Check option variables.
6611 (dwarf2out_inline_entry): Likewise.
6612 (init_sections_and_labels): Likewise.
6613 (dwarf2out_early_finish): Likewise.
6614 (maybe_reset_location_view): New, from...
6615 (dwarf2out_var_location): ... here. Call it.
6616 * debug.h (dwarf2out_default_as_loc_support): Declare.
6617 (dwarf2out_default_as_locview_support): Declare.
6618 * hooks.c (hook_int_rtx_insn_0): New.
6619 * hooks.h (hook_int_rtx_insn_0): Declare.
6620 * toplev.c (process_options): Take -gas-loc-support and
6621 -gas-locview-support from dwarf2out. Enable
6622 -gvariable-location-views by default only with locview
6623 assembler support. Enable -ginternal-reset-location-views by
6624 default only if the target defines the corresponding hook.
6625 Enable -ginline-points by default if location views are
6626 enabled; force it disabled if statement frontiers are
6628 * tree-inline.c (expand_call_inline): Check option variables.
6629 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6631 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6633 PR tree-optimization/84321
6634 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6635 handling. Also check whether the anti-range contains any values
6636 that satisfy the mask; switch to a VR_RANGE if not.
6638 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6641 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6643 2018-02-13 Martin Jambor <mjambor@suse.cz>
6646 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6647 of call statements, also set location of a load to a temporary.
6649 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6651 * config/rl78/rl78.c (add_vector_labels): New function.
6652 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6653 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6654 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6655 which checks that no arguments are passed.
6656 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6657 * doc/extend.texi: Documentation for the new attribute.
6659 2018-02-13 Andreas Schwab <schwab@suse.de>
6661 * config/riscv/linux.h (CPP_SPEC): Define.
6663 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6666 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6667 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6668 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6669 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6670 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6671 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6672 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6673 temporarily for AES and PCLMUL builtins.
6675 PR tree-optimization/84339
6676 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6677 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6681 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6682 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6683 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6685 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6687 * omp-simd-clone.h: New file.
6688 * omp-simd-clone.c: Include omp-simd-clone.h.
6689 (expand_simd_clones): No longer static.
6690 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6691 cgraph.h and omp-simd-clone.h.
6692 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6693 (vect_recog_widen_shift_pattern): Formatting fix.
6694 (vect_pattern_recog_1): Don't check optab for calls.
6697 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6698 operands[2] into a REG before using gen_lowpart on it.
6700 2018-02-12 Jeff Law <law@redhat.com>
6703 * config/sh/sh.c (find_barrier): Consider a sibling call
6706 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6707 successfully back substituting a reg.
6709 2018-02-12 Richard Biener <rguenther@suse.de>
6711 PR tree-optimization/84037
6712 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6713 parameter, move visited init to caller.
6714 (vect_slp_analyze_operations): Separate cost from validity
6715 check, initialize visited once for all instances.
6716 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6718 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6720 (vect_model_promotion_demotion_cost): Likewise.
6721 (vectorizable_bswap): Guard cost modeling with !slp_node
6722 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6724 (vectorizable_call): Likewise.
6725 (vectorizable_conversion): Likewise.
6726 (vectorizable_assignment): Likewise.
6727 (vectorizable_shift): Likewise.
6728 (vectorizable_operation): Likewise.
6729 (vectorizable_store): Likewise.
6730 (vectorizable_load): Likewise.
6731 (vectorizable_condition): Likewise.
6732 (vectorizable_comparison): Likewise.
6734 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6737 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6738 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6739 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6741 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6744 * params.def (inline-min-speedup): Increase from 8 to 15.
6745 (max-inline-insns-auto): Decrease from 40 to 30.
6746 * ipa-split.c (consider_split): Add some buffer for function to
6747 be considered inlining candidate.
6748 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6751 2018-02-12 Richard Biener <rguenther@suse.de>
6753 PR tree-optimization/84037
6754 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6755 matched stmts if we cannot swap the non-matched ones.
6757 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6759 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6760 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6761 _mm_maskz_scalef_round_ss): New intrinsics.
6762 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6763 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6764 __builtin_ia32_scalefss_round): Remove.
6765 (__builtin_ia32_scalefsd_mask_round,
6766 __builtin_ia32_scalefss_mask_round): New intrinsics.
6767 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6768 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6769 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6770 "<round_constraint>")): Changed to ...
6771 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6772 "<round_scalar_constraint>")): ... this.
6773 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6774 %0, %1, %2<round_op3>}"): Changed to ...
6775 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6776 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6777 %2<round_scalar_mask_op3>}"): ... this.
6778 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6780 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6782 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6783 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6784 (_mm_maskz_sqrt_round_ss): New intrinsics.
6785 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6786 (__builtin_ia32_sqrtsd_mask_round)
6787 (__builtin_ia32_sqrtss_mask_round): New builtins.
6788 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6789 (__builtin_ia32_sqrtss_round): Remove.
6790 (__builtin_ia32_sqrtsd_mask_round)
6791 (__builtin_ia32_sqrtss_mask_round): New builtins.
6792 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6793 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6794 ((match_operand:VF_128 1 "vector_operand"
6795 "xBm,<round_constraint>")): Changed to ...
6796 ((match_operand:VF_128 1 "vector_operand"
6797 "xBm,<round_scalar_constraint>")): ... this.
6798 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6799 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6800 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6801 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6802 %<iptr>1<round_scalar_mask_op3>}): ... this.
6803 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6804 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6806 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6809 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6810 Cast vec_cmpeq result to correct type.
6811 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6812 Cast vec_cmpgt result to correct type.
6814 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6816 * final.c (final_scan_insn_1): Renamed from...
6817 (final_scan_insn): ... this. New wrapper, to recover
6818 seen from the outermost call in recursive ones.
6819 * config/sparc/sparc.c (output_return): Drop seen from call.
6820 (output_sibcall): Likewise.
6821 * config/visium/visium.c (output_branch): Likewise.
6823 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6825 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6828 2018-02-10 Alan Modra <amodra@gmail.com>
6831 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6832 Specify LR as an input.
6834 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6837 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6838 remove_member_access_dummy_vars): New functions.
6839 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6840 lower_omp_1, execute_lower_omp): Use them.
6842 PR rtl-optimization/84308
6843 * shrink-wrap.c (spread_components): Release todo vector.
6845 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6847 PR rtl-optimization/57193
6848 * ira-color.c (struct allocno_color_data): Add member
6849 conflict_allocno_hard_prefs.
6850 (update_conflict_allocno_hard_prefs): New.
6851 (bucket_allocno_compare_func): Add a preference based on
6852 conflict_allocno_hard_prefs.
6853 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6854 (color_allocnos): Remove a dead code. Initiate
6855 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6857 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6860 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6861 constraint from =wa to wa. Avoid a subreg on the output operand,
6862 instead use a pseudo and subreg it in a move.
6863 (p9_xxbrd_<mode>): Changed to ...
6864 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6865 (p9_xxbrd_v2df): New expander.
6866 (p9_xxbrw_<mode>): Changed to ...
6867 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6868 (p9_xxbrw_v4sf): New expander.
6870 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6872 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6874 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6877 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6878 multiply in 32-bit mode.
6879 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6880 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6883 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6885 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6886 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6887 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6888 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6890 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6893 * dwarf2out.c (is_trivial_indirect_ref): New function.
6894 (dwarf2out_late_global_decl): Do not generate a location
6895 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6896 and that are not defined in the current unit.
6898 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6900 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6901 instead of a libcall for UNORDERED.
6903 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6906 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6907 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6909 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6912 * config/s390/s390.c (s390_set_current_function): Invoke
6913 s390_indirect_branch_settings also if fndecl didn't change.
6915 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6917 * config/rs6000/rs6000.md (blockage): Set length to zero.
6919 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6921 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6923 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6926 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6927 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6931 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6932 PARALLEL incoming that failed vt_get_decl_and_offset check.
6935 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6936 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6938 (get_variable_section): For decls in named .bss* sections pass true as
6939 second argument to bss_initializer_p.
6941 2018-02-09 Marek Polacek <polacek@redhat.com>
6942 Jakub Jelinek <jakub@redhat.com>
6945 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6946 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6947 Sync some changes from cxx_fold_indirect_ref.
6949 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6951 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6953 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6954 (BLOCK_INLINE_ENTRY_LABEL): New.
6955 (dwarf2out_var_location): Disregard inline entry markers.
6956 (inline_entry_data): New struct.
6957 (inline_entry_data_hasher): New hashtable type.
6958 (inline_entry_data_hasher::hash): New.
6959 (inline_entry_data_hasher::equal): New.
6960 (inline_entry_data_table): New variable.
6961 (add_high_low_attributes): Add DW_AT_entry_pc and
6962 DW_AT_GNU_entry_view attributes if a pending entry is found
6963 in inline_entry_data_table. Add old entry_pc attribute only
6964 if debug nonbinding markers are disabled.
6965 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6966 markers are enabled.
6967 (block_within_block_p, dwarf2out_inline_entry): New.
6968 (dwarf2out_finish): Check that no entries remained in
6969 inline_entry_data_table.
6970 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6971 (final_scan_insn, notice_source_line): Likewise.
6972 (rest_of_clean_state): Skip inline entry markers.
6973 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6975 * gimple.c (gimple_build_debug_inline_entry): New.
6976 * gimple.h (enum gimple_debug_subcode): Add
6977 GIMPLE_DEBUG_INLINE_ENTRY.
6978 (gimple_build_debug_inline_entry): Declare.
6979 (gimple_debug_inline_entry_p): New.
6980 (gimple_debug_nonbind_marker_p): Adjust.
6981 * insn-notes.def (INLINE_ENTRY): New.
6982 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6983 inline entry marker notes.
6984 (print_insn): Likewise.
6985 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6986 (INSN_DEBUG_MARKER_KIND): Likewise.
6987 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6988 * tree-inline.c (expand_call_inline): Build and insert
6989 debug_inline_entry stmt.
6990 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6991 inline entry blocks early, if nonbind markers are enabled.
6992 (dump_scope_block): Dump fragment info.
6993 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6994 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6995 (gimple_build_debug_inline_entry): New.
6996 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6997 Enable/disable inline entry points too.
6998 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6999 (DEBUG_INSN): Describe inline entry markers.
7001 * common.opt (gvariable-location-views): New.
7002 (gvariable-location-views=incompat5): New.
7003 * config.in: Rebuilt.
7004 * configure: Rebuilt.
7005 * configure.ac: Test assembler for view support.
7006 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7007 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7008 * dwarf2out.c (var_loc_view): New typedef.
7009 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7010 (dwarf2out_locviews_in_attribute): New.
7011 (dwarf2out_locviews_in_loclist): New.
7012 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7013 (enum dw_line_info_opcode): Add LI_adv_address.
7014 (struct dw_line_info_table): Add view.
7015 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7016 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7017 (zero_view_p): New variable.
7018 (ZERO_VIEW_P): New macro.
7019 (output_asm_line_debug_info): New.
7020 (struct var_loc_node): Add view.
7021 (add_AT_view_list, AT_loc_list): New.
7022 (add_var_loc_to_decl): Add view param. Test it against last.
7023 (new_loc_list): Add view params. Record them.
7024 (AT_loc_list_ptr): Handle loc and view lists.
7025 (view_list_to_loc_list_val_node): New.
7026 (print_dw_val): Handle dw_val_class_view_list.
7027 (size_of_die): Likewise.
7028 (value_format): Likewise.
7029 (loc_list_has_views): New.
7030 (gen_llsym): Set vl_symbol too.
7031 (maybe_gen_llsym, skip_loc_list_entry): New.
7032 (dwarf2out_maybe_output_loclist_view_pair): New.
7033 (output_loc_list): Output view list or entries too.
7034 (output_view_list_offset): New.
7035 (output_die): Handle dw_val_class_view_list.
7036 (output_dwarf_version): New.
7037 (output_compilation_unit_header): Use it.
7038 (output_skeleton_debug_sections): Likewise.
7039 (output_rnglists, output_line_info): Likewise.
7040 (output_pubnames, output_aranges): Update version comments.
7041 (output_one_line_info_table): Output view numbers in asm comments.
7042 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7043 Call maybe_gen_llsym.
7044 (loc_list_from_tree_1): Adjust.
7045 (add_AT_location_description): Create view list attribute if
7046 needed, check it's absent otherwise.
7047 (convert_cfa_to_fb_loc_list): Adjust.
7048 (maybe_emit_file): Call output_asm_line_debug_info for test.
7049 (dwarf2out_var_location): Reset views as needed. Precompute
7050 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7051 attribute. Set view.
7052 (new_line_info_table): Reset next view.
7053 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7054 (dwarf2out_source_line): Likewise. Output view resets and labels to
7055 the assembler, or select appropriate line info opcodes.
7056 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7057 (optimize_string_length): Catch it. Adjust.
7058 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7059 dw_val_class_view_list, and remove it if no longer needed.
7060 (hash_loc_list): Hash view numbers.
7061 (loc_list_hasher::equal): Compare them.
7062 (optimize_location_lists): Check whether a view list symbol is
7063 needed, and whether the locview attribute is present, and
7064 whether they match. Remove the locview attribute if no longer
7066 (index_location_lists): Call skip_loc_list_entry for test.
7067 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7068 Use output_dwarf_version.
7069 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7070 (struct dw_val_node): Add val_view_list.
7071 * final.c (SEEN_NEXT_VIEW): New.
7072 (set_next_view_needed): New.
7073 (clear_next_view_needed): New.
7074 (maybe_output_next_view): New.
7075 (final_start_function): Rename to...
7076 (final_start_function_1): ... this. Take pointer to FIRST,
7077 add SEEN parameter. Emit param bindings in the initial view.
7078 (final_start_function): Reintroduce SEEN-less interface.
7079 (final): Rename to...
7080 (final_1): ... this. Take SEEN parameter. Output final pending
7081 next view at the end.
7082 (final): Reintroduce seen-less interface.
7083 (final_scan_insn): Output pending next view before switching
7084 sections or ending a block. Mark the next view as needed when
7085 outputting variable locations. Notify debug backend of section
7086 changes, and of location view changes.
7087 (rest_of_handle_final): Adjust.
7088 * toplev.c (process_options): Autodetect value for debug variable
7089 location views option. Warn on incompat5 without -gdwarf-5.
7090 * doc/invoke.texi (gvariable-location-views): New.
7091 (gvariable-location-views=incompat5): New.
7092 (gno-variable-location-views): New.
7094 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7096 PR tree-optimization/84136
7097 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7098 that the result of find_edge is non-NULL.
7100 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7103 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7104 storing integer register in SImode. Fix cost of 256 and 512
7105 byte aligned SSE register store.
7107 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7109 * config/i386/i386.c (ix86_multiplication_cost): Fix
7110 multiplication cost for TARGET_AVX512DQ.
7112 2018-02-08 Marek Polacek <polacek@redhat.com>
7114 PR tree-optimization/84238
7115 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7118 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7120 PR tree-optimization/84265
7121 * tree-vect-stmts.c (vectorizable_store): Don't treat
7122 VMAT_CONTIGUOUS accesses as grouped.
7123 (vectorizable_load): Likewise.
7125 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7127 PR tree-optimization/81635
7128 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7129 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7130 (test_round_for_mask): New functions.
7131 (wide_int_cc_tests): Call test_round_for_mask.
7132 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7133 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7134 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7135 range returned by get_range_info.
7137 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7140 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7141 * symtab.c: Include builtins.h
7142 (symtab_node::output_to_lto_symbol_table_p): Move here
7143 from lto-streamer-out.c:output_symbol_p.
7144 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7145 (output_symbol_p): Move all logic to symtab.c
7146 (produce_symtab): Update.
7148 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7150 * config/s390/s390-opts.h (enum indirect_branch): Define.
7151 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7152 (s390_indirect_branch_via_thunk)
7153 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7154 (enum s390_indirect_branch_type): Define.
7155 * config/s390/s390.c (struct s390_frame_layout, struct
7156 machine_function): Remove.
7157 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7158 (indirect_branch_table_label_no, indirect_branch_table_name):
7160 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7161 (enum s390_indirect_branch_option): Define.
7162 (s390_return_addr_from_memory): New function.
7163 (s390_handle_string_attribute): New function.
7164 (s390_attribute_table): Add new attribute handler.
7165 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7166 (s390_indirect_branch_via_thunk): New function.
7167 (s390_indirect_branch_via_inline_thunk): New function.
7168 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7169 sibling call optimization for non z10 compiles.
7170 (s390_emit_call): Force indirect branch target to be a single
7171 register. Add r1 clobber for non-z10 compiles.
7172 (s390_emit_epilogue): Emit return jump via return_use expander.
7173 (s390_reorg): Handle JUMP_INSNs as execute targets.
7174 (s390_option_override_internal): Perform validity checks for the
7175 new command line options.
7176 (s390_indirect_branch_attrvalue): New function.
7177 (s390_indirect_branch_settings): New function.
7178 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7179 (s390_output_indirect_thunk_function): New function.
7180 (s390_code_end): Implement target hook.
7181 (s390_case_values_threshold): Implement target hook.
7182 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7184 * config/s390/s390.h (struct s390_frame_layout)
7185 (struct machine_function): Move here from s390.c.
7186 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7187 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7188 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7189 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7190 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7191 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7192 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7193 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7194 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7195 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7196 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7197 (mnemonic attribute): Add values which aren't recognized
7199 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7200 pattern for branch conversion. Fix mnemonic attribute.
7201 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7202 indirect branch via thunk if requested.
7203 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7204 ("*indirect_jump"): Disable for branch conversion using out of
7206 ("indirect_jump_via_thunk<mode>_z10")
7207 ("indirect_jump_via_thunk<mode>")
7208 ("indirect_jump_via_inlinethunk<mode>_z10")
7209 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7210 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7211 ("casesi_jump_via_inlinethunk<mode>_z10")
7212 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7213 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7214 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7215 ("*indirect2_jump"): Disable for branch conversion.
7216 ("casesi_jump"): Turn into expander and expand patterns for branch
7218 ("return_use"): New expander.
7219 ("*return"): Emit return via thunk and rename it to ...
7220 ("*return<mode>"): ... this one.
7221 * config/s390/s390.opt: Add new options and and enum for the
7224 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7226 * lra-constraints.c (match_reload): Unconditionally use
7227 gen_lowpart_SUBREG, rather than selecting between that
7228 and equivalent gen_rtx_SUBREG code.
7230 2018-02-08 Richard Biener <rguenther@suse.de>
7232 PR tree-optimization/84233
7233 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7234 changed flag instead of boguously re-using phi_inserted.
7236 2018-02-08 Martin Jambor <mjambor@suse.cz>
7238 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7239 static local variables.
7241 2018-02-08 Richard Biener <rguenther@suse.de>
7243 PR tree-optimization/84278
7244 * tree-vect-stmts.c (vectorizable_store): When looking for
7245 smaller vector types to perform grouped strided loads/stores
7246 make sure the mode is supported by the target.
7247 (vectorizable_load): Likewise.
7249 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7251 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7252 Increase LDP/STP opportunities by adding adjacent callee-saves.
7254 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7256 PR rtl-optimization/84068
7257 PR rtl-optimization/83459
7258 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7260 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7262 PR tree-optimization/84224
7263 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7264 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7267 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7270 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7271 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7272 count, remove one USE.
7274 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7276 * doc/install.texi (Configuration): Document the
7277 --with-long-double-format={ibm,ieee} PowerPC configuration
7281 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7282 Convert from define_expand to be define_insn_and_split. Rework
7283 float/double/_Float128 conversions to QI/HI/SImode to work with
7284 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7285 conversions to QI/HImode types did a store and then a load to
7286 truncate the value. For conversions to VSX registers, don't split
7287 the insn, instead emit the code directly. Use the code iterator
7288 any_fix to combine signed and unsigned conversions.
7289 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7290 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7291 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7292 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7293 (fix_<mode>di2_hw): Likewise.
7294 (fixuns_<mode>di2_hw): Likewise.
7295 (fix_<mode>si2_hw): Likewise.
7296 (fixuns_<mode>si2_hw): Likewise.
7297 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7298 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7299 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7300 fix<uns>_trunc<SFDF:mode>si2_p8.
7301 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7303 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7304 (fix<uns>_<mode>_mem): Likewise.
7305 (fctiw<u>z_<mode>_mem): Likewise.
7306 (fix<uns>_<mode>_mem): Likewise.
7307 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7308 the register allocator from doing a direct move to the GPRs to do
7309 a store, and instead use the ISA 3.0 store byte/half-word from
7310 vector register instruction. For IEEE 128-bit floating point,
7311 also optimize stores of 32-bit ints.
7312 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7314 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7316 * genextract.c (push_pathstr_operand): New function to support
7318 (walk_rtx): Call push_pathstr_operand.
7319 (print_path): Support [a-zA-Z].
7321 2018-02-07 Richard Biener <rguenther@suse.de>
7323 PR tree-optimization/84037
7324 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7325 (cse_and_gimplify_to_preheader): Declare.
7326 (vect_get_place_in_interleaving_chain): Likewise.
7327 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7329 (_loop_vec_info::~_loop_vec_info): Delete it.
7330 (cse_and_gimplify_to_preheader): New function.
7331 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7332 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7333 (vectorizable_load): Likewise. For grouped stores always base
7334 the IV on the first element.
7335 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7336 condition before gimplifying.
7338 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7340 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7341 *DIV_EXPR and *MOD_EXPR.
7343 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7346 * config/i386/i386.c (ix86_option_override_internal): Mask out
7347 the CF_SET bit when checking -fcf-protection.
7349 2018-02-07 Tom de Vries <tom@codesourcery.com>
7352 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7355 2018-02-07 Richard Biener <rguenther@suse.de>
7357 PR tree-optimization/84204
7358 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7361 PR tree-optimization/84205
7362 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7363 special-case isl_ast_op_zdiv_r.
7365 PR tree-optimization/84223
7366 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7367 Only add conditions from within the region.
7368 (gather_bbs::after_dom_children): Adjust.
7370 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7373 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7374 * config/avr/avr.md: Only post-reload split REG-REG moves if
7375 either register is GENERAL_REG_P.
7377 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7379 PR tree-optimization/84235
7380 * tree-ssa-scopedtables.c
7381 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7382 if the subtraction is performed in floating point type where NaNs are
7383 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7384 build 1. Formatting fix.
7386 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7389 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7390 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7391 and skip it regardless of bb boundaries. Use CALL_P macro,
7392 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7393 unnecessarily, formatting fix.
7395 2018-02-06 Michael Collison <michael.collison@arm.com>
7397 * config/arm/thumb2.md:
7398 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7399 (*thumb_mov_notscc): Ditto.
7401 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7404 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7405 unsigned_fix, not "s".
7407 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7409 * configure.ac (gcc_fn_eh_frame_ro): New function.
7410 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7411 correct .eh_frame permissions.
7412 * configure: Regenerate.
7414 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7416 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7419 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7421 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7422 Display warning message for -mno-speculate-indirect-jumps.
7424 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7426 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7428 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7430 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7432 PR tree-optimization/84225
7433 * tree-eh.c (find_trapping_overflow): Only call
7434 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7436 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7439 * config/i386/i386.c: Reimplement the check of possible options
7440 -mibt/-mshstk conbination. Change error messages.
7441 * doc/invoke.texi: Fix a typo: remove extra '='.
7443 2018-02-06 Marek Polacek <polacek@redhat.com>
7445 PR tree-optimization/84228
7446 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7448 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7451 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7452 emitted arch directives.
7453 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7454 __ARM_FEATURE_COPROC before changing architectures.
7456 2018-02-06 Richard Biener <rguenther@suse.de>
7458 * config/i386/i386.c (print_reg): Fix typo.
7459 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7461 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7463 * configure: Regenerate.
7465 2018-02-05 Martin Sebor <msebor@redhat.com>
7467 PR tree-optimization/83369
7468 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7471 2018-02-05 Martin Liska <mliska@suse.cz>
7473 * doc/invoke.texi: Cherry-pick upstream r323995.
7475 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7477 * ira.c (ira_init_register_move_cost): Adjust comment.
7479 2018-02-05 Martin Liska <mliska@suse.cz>
7481 PR gcov-profile/84137
7482 * doc/gcov.texi: Fix typo in documentation.
7484 2018-02-05 Martin Liska <mliska@suse.cz>
7486 PR gcov-profile/83879
7487 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7488 using dlopen functionality.
7490 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7492 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7493 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7494 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7495 _mm_maskz_range_round_ss): New intrinsics.
7496 (__builtin_ia32_rangesd128_round)
7497 (__builtin_ia32_rangess128_round): Remove.
7498 (__builtin_ia32_rangesd128_mask_round,
7499 __builtin_ia32_rangess128_mask_round): New builtins.
7500 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7501 __builtin_ia32_rangess128_round): Remove.
7502 (__builtin_ia32_rangesd128_mask_round,
7503 __builtin_ia32_rangess128_mask_round): New builtins.
7504 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7505 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7506 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7507 "<round_saeonly_constraint>")): Changed to ...
7508 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7509 "<round_saeonly_scalar_constraint>")): ... this.
7510 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7511 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7512 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7513 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7514 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7516 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7518 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7520 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7521 Remove all values except native, 8540 and 8548.
7523 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7525 * config/i386/i386.c (ix86_output_function_return): Pass
7526 INVALID_REGNUM, instead of -1, as invalid register number to
7527 indirect_thunk_name and output_indirect_thunk.
7529 2018-02-02 Julia Koval <julia.koval@intel.com>
7531 * config.gcc: Add -march=icelake.
7532 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7533 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7534 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7535 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7536 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7537 (processor_target_table): Add icelake.
7538 (ix86_option_override_internal): Handle new PTAs.
7539 (get_builtin_code_for_version): Handle icelake.
7540 (M_INTEL_COREI7_ICELAKE): New.
7541 (fold_builtin_cpu): Handle icelake.
7542 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7543 * doc/invoke.texi: Add -march=icelake.
7545 2018-02-02 Julia Koval <julia.koval@intel.com>
7547 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7548 to wide_int_bitmask.
7549 * wide-int-bitmask.h: New.
7551 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7554 * config/i386/i386.md: Replace Pmode with word_mode in
7555 builtin_setjmp_setup and builtin_longjmp to support x32.
7557 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7561 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7563 (rs6000_supported_cpu_names): New static variable.
7564 (linux_cpu_translation_table): Likewise.
7565 (elf_platform) <cpu>: Define new static variable and use it.
7566 Translate kernel AT_PLATFORM name to canonical name if needed.
7567 Error if platform name is unknown.
7569 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7572 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7574 2018-02-01 Jeff Law <law@redhat.com>
7577 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7578 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7579 the scratch if RELEASE_VIA_POP is false.
7580 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7581 If we have to save a temporary register, decrement SIZE appropriately.
7582 Pass new arguments to release_scratch_register_on_entry.
7583 (ix86_adjust_stack_and_probe): Likewise.
7584 (ix86_emit_probe_stack_range): Pass new arguments to
7585 release_scratch_register_on_entry.
7587 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7589 PR rtl-optimization/84157
7590 * combine.c (change_zero_ext): Use REG_P predicate in
7591 front of HARD_REGISTER_P predicate.
7593 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7595 * config/avr/avr.c (avr_option_override): Move disabling of
7596 -fdelete-null-pointer-checks to...
7597 * common/config/avr/avr-common.c (avr_option_optimization_table):
7600 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7602 PR tree-optimization/81635
7603 * tree-data-ref.c (split_constant_offset_1): For types that
7604 wrap on overflow, try to use range info to prove that wrapping
7607 2018-02-01 Renlin Li <renlin.li@arm.com>
7610 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7612 (aarch64_register_move_cost): Likewise.
7613 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7615 (REG_CLASS_NAMES): Likewise.
7616 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7617 TAILCALL_ADDR_REGS. Remove IP registers.
7618 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7620 2018-02-01 Richard Biener <rguenther@suse.de>
7622 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7623 for specifying RPO order and allow NULL for that.
7624 * domwalk.c (dom_walker::dom_walker): Likewise.
7625 (dom_walker::walk): Handle NULL RPO order.
7626 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7628 (rewrite_update_dom_walker): Likewise.
7629 (mark_def_dom_walker): Likewise.
7631 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7633 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7634 (aarch64_maybe_expand_sve_subreg_move): Declare.
7635 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7636 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7638 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7639 that are semantically a reverse operation.
7640 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7641 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7642 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7644 (aarch64_can_change_mode_class): For big-endian, forbid changes
7645 between two SVE modes if they have different element sizes.
7647 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7649 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7650 the TImode handling for big-endian targets.
7652 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7654 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7655 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7657 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7658 Remove BSWAP handing for big-endian targets and use the form of
7659 LD1RQ appropariate for the mode.
7661 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7663 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7664 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7667 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7670 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7671 check for operands that need to go through aarch64_sve_reload_be.
7673 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7675 PR tree-optimization/81661
7676 PR tree-optimization/84117
7677 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7678 * tree-eh.c: Include gimplify.h.
7679 (find_trapping_overflow, replace_trapping_overflow,
7680 rewrite_to_non_trapping_overflow): New functions.
7681 * tree-vect-loop.c: Include tree-eh.h.
7682 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7683 * tree-data-ref.c: Include tree-eh.h.
7684 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7686 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7688 PR rtl-optimization/84123
7689 * combine.c (change_zero_ext): Check if hard register satisfies
7690 can_change_dest_mode before calling gen_lowpart_SUBREG.
7692 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7695 * ira.c (ira_init_register_move_cost): Remove assert.
7697 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7699 PR rtl-optimization/84071
7700 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7701 * doc/tm.texi: Regenerate.
7703 2018-01-31 Richard Biener <rguenther@suse.de>
7705 PR tree-optimization/84132
7706 * tree-data-ref.c (analyze_miv_subscript): Properly
7707 check whether evolution_function_is_affine_multivariate_p
7708 before calling gcd_of_steps_may_divide_p.
7710 2018-01-31 Julia Koval <julia.koval@intel.com>
7713 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7714 * config/i386/i386.md (rdpid_rex64) New.
7715 (rdpid): Make 32bit only.
7717 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7720 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7721 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7723 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7726 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7728 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7730 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7732 PR rtl-optimization/84071
7733 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7734 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7736 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7738 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7739 (arc_attribute_table): Add 'aux' attribute.
7740 (arc_in_small_data_p): Consider aux like variables.
7741 (arc_is_aux_reg_p): New function.
7742 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7743 (arc_get_aux_arg): New function.
7744 (prepare_move_operands): Handle aux-register access.
7745 (arc_handle_aux_attribute): New function.
7746 * doc/extend.texi (ARC Variable attributes): Add subsection.
7748 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7750 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7751 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7752 (arc_attribute_table): Add 'uncached' attribute.
7753 (arc_print_operand): Print '.di' flag for uncached memory
7755 (arc_in_small_data_p): Do not consider for small data the uncached
7757 (arc_is_uncached_mem_p): New function.
7758 * config/arc/predicates.md (compact_store_memory_operand): Check
7759 for uncached memory accesses.
7760 (nonvol_nonimm_operand): Likewise.
7761 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7763 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7766 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7767 falign-loops=): Add Optimization flag.
7769 2018-01-30 Jeff Law <law@redhat.com>
7772 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7773 INT_REGISTERS_SAVED. Check it prior to calling
7774 get_scratch_register_on_entry.
7775 (ix86_adjust_stack_and_probe): Similarly.
7776 (ix86_emit_probe_stack_range): Similarly.
7777 (ix86_expand_prologue): Corresponding changes.
7779 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7782 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7783 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7785 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7788 * lra-constraints.c (curr_insn_transform): Process AND in the
7791 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7793 PR rtl-optimization/83986
7794 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7795 dependence against last_pending_memory_flush in addition to
7798 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7800 PR tree-optimization/81611
7801 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7804 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7807 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7810 2018-01-30 Richard Biener <rguenther@suse.de>
7811 Jakub Jelinek <jakub@redhat.com>
7813 PR tree-optimization/84111
7814 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7815 inner loops added during recursion, as they don't have up-to-date
7818 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7821 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7822 (can_inline_edge_by_limits_p): ... here.
7823 (can_early_inline_edge_p, check_callers,
7824 update_caller_keys, update_callee_keys, recursive_inlining,
7825 add_new_edges_to_heap, speculation_useful_p,
7826 inline_small_functions,
7827 inline_small_functions, flatten_function,
7828 inline_to_all_callers_1): Update.
7830 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7832 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7835 2018-01-30 Richard Biener <rguenther@suse.de>
7837 PR tree-optimization/83008
7838 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7839 invariant and constant vector uses in stmts when they need
7842 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7845 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7846 * configure: Regenerate.
7848 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7850 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7852 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7853 Use gen_rtx_REG rather than gen_lowpart.
7855 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7857 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7858 rather than 0 when creating partial subregs.
7860 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7862 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7865 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7868 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7869 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7870 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7871 flags. This restores the settings used before the 2017-07-24.
7872 Turning off pre increment/decrement/modify allows IVOPTS to
7873 optimize DF/SF loops where the index is an int.
7875 2018-01-29 Richard Biener <rguenther@suse.de>
7876 Kelvin Nilsen <kelvin@gcc.gnu.org>
7879 * tree-vect-stmts.c (vectorizable_call): Don't call
7880 targetm.vectorize_builtin_md_vectorized_function if callee is
7883 2018-01-22 Carl Love <cel@us.ibm.com>
7885 * doc/extend.tex: Fix typo in second arg in
7886 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7888 2018-01-29 Richard Biener <rguenther@suse.de>
7890 PR tree-optimization/84086
7891 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7892 (flush_ssaname_freelist): When SSA names were released reset
7893 the SCEV hash table.
7895 2018-01-29 Richard Biener <rguenther@suse.de>
7897 PR tree-optimization/84057
7898 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7899 removed paths when removing edges.
7901 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7903 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7904 -mfunction-return=@var{choice}.
7906 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7909 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7911 (layout::print_source_line): Likewise.
7912 (test_get_line_width_without_trailing_whitespace): Add test cases.
7914 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7917 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7920 2018-01-26 Jim Wilson <jimw@sifive.com>
7922 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7924 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7927 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7929 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7930 and CMP + SUB-immediate -> SUBS.
7932 2018-01-26 Martin Sebor <msebor@redhat.com>
7934 PR tree-optimization/83896
7935 * tree-ssa-strlen.c (get_string_len): Rename...
7936 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7937 Avoid assuming length is constant.
7938 (handle_char_store): Use HOST_WIDE_INT for string length.
7940 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7943 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7944 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7946 2018-01-26 Richard Biener <rguenther@suse.de>
7948 PR rtl-optimization/84003
7949 * dse.c (record_store): Only record redundant stores when
7950 the earlier store aliases at least all accesses the later one does.
7952 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7954 PR rtl-optimization/83985
7955 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7956 REG_CFA_RESTORE insns.
7957 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7958 REG_CFA_RESTORE insns here.
7961 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7962 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7964 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7966 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7967 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7968 (arc_init): Likewise.
7969 (arc_override_options): Likewise.
7970 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7972 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7974 * config/arc/arc.h (TARGET_DBNZ): Define.
7975 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7976 properly set the tune attribute.
7977 (dbnz): Use TARGET_DBNZ guard.
7978 * config/arc/arc.opt (mtune): Add core3 option.
7980 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7982 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7983 recognize new pic like addresses.
7984 (arc_delegitimize_address): Clean up.
7986 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7988 * config/arc/arc-arches.def: Option mrf16 valid for all
7990 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7991 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7992 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7993 * config/arc/arc-tables.opt: Regenerate.
7994 * config/arc/arc.c (arc_conditional_register_usage): Handle
7995 reduced register file case.
7996 (arc_file_start): Set must have build attributes.
7997 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7999 * config/arc/arc.opt (mrf16): Add new option.
8000 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8001 * config/arc/genmultilib.awk: Handle new mrf16 option.
8002 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8003 * config/arc/t-multilib: Regenerate.
8004 * doc/invoke.texi (ARC Options): Document mrf16 option.
8006 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8008 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8009 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8010 (arc_attribute_table): Add 'secure_call' attribute.
8011 (arc_print_operand): Print secure call operand.
8012 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8014 (arc_is_secure_call_p): New function. * config/arc/arc.md
8015 (call_i): Add support for sjli instruction.
8016 (call_value_i): Likewise.
8017 * config/arc/constraints.md (Csc): New constraint.
8019 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8020 John Eric Martin <John.Martin@emmicro-us.com>
8022 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8023 * config/arc/arc.c (_arc_jli_section): New struct.
8024 (arc_jli_section): New type.
8025 (rc_jli_sections): New static variable.
8026 (arc_handle_jli_attribute): New function.
8027 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8028 (arc_file_end): New function.
8029 (TARGET_ASM_FILE_END): Define.
8030 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8031 (arc_add_jli_section): New function.
8032 (jli_call_scan): Likewise.
8033 (arc_reorg): Call jli_call_scan.
8034 (arc_output_addsi): Remove 'S' from printing asm operand.
8035 (arc_is_jli_call_p): New function.
8036 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8038 (movhi_insn): Likewise.
8039 (movsi_insn): Likewise.
8040 (movsi_set_cc_insn): Likewise.
8041 (loadqi_update): Likewise.
8042 (load_zeroextendqisi_update): Likewise.
8043 (load_signextendqisi_update): Likewise.
8044 (loadhi_update): Likewise.
8045 (load_zeroextendhisi_update): Likewise.
8046 (load_signextendhisi_update): Likewise.
8047 (loadsi_update): Likewise.
8048 (loadsf_update): Likewise.
8049 (movsicc_insn): Likewise.
8050 (bset_insn): Likewise.
8051 (bxor_insn): Likewise.
8052 (bclr_insn): Likewise.
8053 (bmsk_insn): Likewise.
8054 (bicsi3_insn): Likewise.
8055 (cmpsi_cc_c_insn): Likewise.
8056 (movsi_ne): Likewise.
8057 (movsi_cond_exec): Likewise.
8058 (clrsbsi2): Likewise.
8069 (call_i): Remove 'S' asm letter, add jli instruction.
8070 (call_value_i): Likewise.
8071 * config/arc/arc.op (mjli-always): New option.
8072 * config/arc/constraints.md (Cji): New constraint.
8073 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8075 (subsf3_fpx): Likewise.
8076 (mulsf3_fpx): Likewise.
8077 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8079 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8081 * doc/invoke.texi (ARC): Document mjli-always option.
8083 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8085 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8086 avoid addition with 0 and use incw and decw where possible.
8088 2018-01-26 Richard Biener <rguenther@suse.de>
8090 PR tree-optimization/81082
8091 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8092 association if it requires casting to unsigned.
8093 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8094 from fold_plusminus_mult_expr to catch important cases late when
8095 range info is available.
8097 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8099 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8100 * configure.ac (hidden_linkonce): New test.
8101 * configure: Regenerate.
8102 * config.in: Regenerate.
8104 2018-01-26 Julia Koval <julia.koval@intel.com>
8106 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8107 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8108 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8109 _mm_mask_bitshuffle_epi64_mask): Fix type.
8110 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8111 USI_FTYPE_V4DI_V4DI_USI): Remove.
8112 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8113 __builtin_ia32_vpshufbitqmb256_mask,
8114 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8115 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8116 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8118 2018-01-26 Alan Modra <amodra@gmail.com>
8121 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8122 UNSPEC_VBPERMQ. Sort other unspecs.
8124 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8126 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8128 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8131 * predict.c (drop_profile): Do not push/pop cfun; update also
8133 (handle_missing_profiles): Fix logic looking for zero profiles.
8135 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8138 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8139 on functions with #pragma omp declare simd or functions with simd
8141 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8142 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8143 Remove trailing \n from warning_at calls.
8145 2018-01-25 Tom de Vries <tom@codesourcery.com>
8148 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8149 for neutered workers.
8151 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8154 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8155 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8157 2018-01-24 Jeff Law <law@redhat.com>
8160 * i386.c (get_probe_interval): Move to earlier point.
8161 (ix86_compute_frame_layout): If -fstack-clash-protection and
8162 the frame is larger than the probe interval, then use pushes
8163 to save registers rather than reg->mem moves.
8164 (ix86_expand_prologue): Remove conditional for int_registers_saved
8167 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8170 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8171 min/max for never referenced object.
8173 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8176 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8178 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8179 attributes from DECL_ATTRIBUTES (decl) without affecting
8180 DECL_ATTRIBUTES (current_function_decl).
8181 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8182 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8184 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8186 PR tree-optimization/83979
8187 * fold-const.c (fold_comparison): Use constant_boolean_node
8188 instead of boolean_{true,false}_node.
8190 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8192 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8195 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8197 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8198 Simplify the clause that sets the length attribute.
8199 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8200 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8201 clause that sets the length attribute.
8202 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8204 2018-01-24 Tom de Vries <tom@codesourcery.com>
8207 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8208 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8209 Add strict parameter.
8210 (prevent_branch_around_nothing): Insert dummy insn between branch to
8211 label and label with no ptx insn inbetween.
8212 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8214 2018-01-24 Tom de Vries <tom@codesourcery.com>
8217 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8218 for neutered threads in warp.
8219 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8221 2018-01-24 Richard Biener <rguenther@suse.de>
8223 PR tree-optimization/83176
8224 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8227 2018-01-24 Richard Biener <rguenther@suse.de>
8229 PR tree-optimization/82819
8230 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8231 code generating pluses that are no-ops in the target precision.
8233 2018-01-24 Richard Biener <rguenther@suse.de>
8236 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8238 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8240 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8241 to merge probabilities.
8242 * predict.c (probably_never_executed): Also mark as cold functions
8243 with global 0 profile and guessed local profile.
8244 * profile-count.c (profile_probability::combine_with_count): New
8246 * profile-count.h (profile_probability::operator*,
8247 profile_probability::operator*=, profile_probability::operator/,
8248 profile_probability::operator/=): Reduce precision to adjusted
8249 and set value to guessed on contradictory divisions.
8250 (profile_probability::combine_with_freq): Remove.
8251 (profile_probability::combine_wiht_count): Declare.
8252 (profile_count::force_nonzero):: Set to adjusted.
8253 (profile_count::probability_in):: Set quality to adjusted.
8254 * tree-ssa-tail-merge.c (replace_block_by): Use
8257 2018-01-23 Andrew Waterman <andrew@sifive.com>
8258 Jim Wilson <jimw@sifive.com>
8260 * config/riscv/riscv.c (riscv_stack_boundary): New.
8261 (riscv_option_override): Set riscv_stack_boundary. Handle
8262 riscv_preferred_stack_boundary_arg.
8263 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8264 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8265 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8266 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8267 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8268 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8270 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8273 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8274 of struct ix86_frame.
8275 (ix86_expand_epilogue): Likewise. Add a local variable for
8276 the reg_save_offset field in struct ix86_frame.
8278 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8280 PR tree-optimization/82604
8281 * tree-loop-distribution.c (enum partition_kind): New enum item
8282 PKIND_PARTIAL_MEMSET.
8283 (partition_builtin_p): Support above new enum item.
8284 (generate_code_for_partition): Ditto.
8285 (compute_access_range): Differentiate cases that equality can be
8286 proven at all loops, the innermost loops or no loops.
8287 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8288 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8289 (finalize_partitions, distribute_loop): Don't fuse partition of
8290 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8291 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8294 2018-01-23 Martin Liska <mliska@suse.cz>
8296 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8297 order to ignore the predictor.
8298 (PRED_POLYMORPHIC_CALL): Likewise.
8299 (PRED_RECURSIVE_CALL): Likewise.
8301 2018-01-23 Martin Liska <mliska@suse.cz>
8303 * tree-profile.c (tree_profiling): Print function header to
8304 aware reader which function we are working on.
8305 * value-prof.c (gimple_find_values_to_profile): Do not print
8306 not interesting value histograms.
8308 2018-01-23 Martin Liska <mliska@suse.cz>
8310 * profile-count.h (enum profile_quality): Add
8311 profile_uninitialized as the first value. Do not number values
8312 as they are zero based.
8313 (profile_count::verify): Update sanity check.
8314 (profile_probability::verify): Likewise.
8316 2018-01-23 Nathan Sidwell <nathan@acm.org>
8318 * doc/invoke.texi (ffor-scope): Deprecate.
8320 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8322 PR tree-optimization/83510
8323 * domwalk.c (set_all_edges_as_executable): New function.
8324 (dom_walker::dom_walker): Convert bool param
8325 "skip_unreachable_blocks" to enum reachability. Move setup of
8326 edge flags to set_all_edges_as_executable and only do it when
8327 reachability is REACHABLE_BLOCKS.
8328 * domwalk.h (enum dom_walker::reachability): New enum.
8329 (dom_walker::dom_walker): Convert bool param
8330 "skip_unreachable_blocks" to enum reachability.
8331 (set_all_edges_as_executable): New decl.
8332 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8333 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8335 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8336 but converting true to REACHABLE_BLOCKS.
8337 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8339 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8340 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8341 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8343 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8344 if check_all_array_refs will be called.
8346 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8348 * tree.c (selftest::test_location_wrappers): Add more test
8351 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8353 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8354 (selftest::test_bit_in_range): Likewise.
8356 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8359 * doc/sourcebuild.texi (vect_float): Say that the selector
8360 only describes the situation when -funsafe-math-optimizations is on.
8361 (vect_float_strict): Document.
8363 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8365 PR tree-optimization/83965
8366 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8367 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8368 instead of checking only for a reduction.
8369 (vect_recog_widen_sum_pattern): Likewise.
8371 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8373 * predict.c (probably_never_executed): Only use precise profile info.
8374 (compute_function_frequency): Skip after inlining hack since we now
8375 have quality checking.
8377 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8379 * profile-count.h (profile_probability::very_unlikely,
8380 profile_probability::unlikely, profile_probability::even): Set
8381 precision to guessed.
8383 2018-01-23 Richard Biener <rguenther@suse.de>
8385 PR tree-optimization/83963
8386 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8387 Properly terminate dominator walk when crossing the exit edge not
8388 when visiting its source block.
8390 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8393 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8394 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8396 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8398 PR tree-optimization/83957
8399 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8400 semicolon after for body surrounded by braces.
8402 PR tree-optimization/83081
8403 * profile-count.h (profile_probability::split): New method.
8404 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8405 Use profile_probability::split.
8406 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8407 when splitting a single conditional jump into 2.
8409 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8411 PR tree-optimization/69452
8412 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8415 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8417 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8418 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8419 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8421 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8423 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8424 * config/rl78/rl78.md (movdi): New define_expand.
8425 * config/rl78/rl78.c (rl78_split_movdi): New function.
8427 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8430 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8432 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8433 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8434 128-bit to produce an UNSPEC move to get the double word with the
8435 signbit and then a shift directly to do signbit.
8436 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8437 implementation with a new version that just does either a direct
8438 move or a regular move. Move memory interface to separate insns.
8439 Move insns so they are next to the expander.
8440 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8441 with signbit move. Split big and little endian case.
8442 (signbit<mode>2_dm_mem_le): Likewise.
8443 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8444 (signbit<mode>2_dm2): Likewise.
8446 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8448 * config/rl78/rl78.md (anddi3): New define_expand.
8450 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8452 * config/rl78/rl78.md (umindi3): New define_expand.
8454 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8456 * config/rl78/rl78.md (smindi3): New define_expand.
8458 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8460 * config/rl78/rl78.md (smaxdi3): New define_expand.
8462 2018-01-22 Carl Love <cel@us.ibm.com>
8464 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8465 LVX_V1TI): Add macro expansion.
8466 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8467 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8468 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8469 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8470 Change check to determine if the instruction is a byte reversing
8471 entry. Fix typo in comment.
8472 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8473 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8474 Add def_builtin calls for new builtins.
8475 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8476 Add define_insn expansion.
8478 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8480 * config/rl78/rl78.md (umaxdi3): New define_expand.
8482 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8484 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8485 for non-QImode registers.
8487 2018-01-22 Richard Biener <rguenther@suse.de>
8489 PR tree-optimization/83963
8490 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8491 including the loop exit block.
8492 (scop_detection::merge_sese): Likewise.
8493 (scop_detection::add_scop): Do it here instead.
8495 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8497 * doc/sourcebuild.texi (arm_softfloat): Document.
8499 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8502 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8503 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8504 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8506 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8507 David Edelsohn <dje.gcc@gmail.com>
8510 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8511 Change "crset eq" to "crset 2".
8512 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8513 (*call_indirect_aix<mode>_nospec): Likewise.
8514 (*call_value_indirect_aix<mode>_nospec): Likewise.
8515 (*call_indirect_elfv2<mode>_nospec): Likewise.
8516 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8517 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8518 change assembly output from . to $.
8519 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8520 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8521 (*tablejump<mode>_internal1_nospec): Likewise.
8523 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8526 * config/sh/sh_optimize_sett_clrt.cc:
8527 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8529 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8531 PR tree-optimization/83940
8532 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8533 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8534 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8535 use it to pass back the definition type.
8536 (vect_check_store_rhs): Likewise rhs_dt_out.
8537 (vect_build_gather_load_calls): Add a mask_dt argument and use
8538 it instead of a call to vect_is_simple_use.
8539 (vectorizable_store): Update calls to vect_check_load_store_mask
8540 and vect_check_store_rhs. Use the dt returned by the latter instead
8541 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8542 instead of calls to vect_is_simple_use. Pass the scalar rather
8543 than the vector operand to vect_is_simple_use when handling
8544 second and subsequent copies of an rhs value.
8545 (vectorizable_load): Update calls to vect_check_load_store_mask
8546 and vect_build_gather_load_calls. Use the cached mask_dt and
8547 gs_info.offset_dt instead of calls to vect_is_simple_use.
8549 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8552 * tree-emutls.c: Include gimplify.h.
8553 (lower_emutls_2): New function.
8554 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8555 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8556 it before further processing.
8559 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8560 UINTVAL (trueop1) instead of INTVAL (op1).
8562 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8566 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8567 INCOMING_FRAME_SP_OFFSET if not defined.
8568 (scan_trace): Add ENTRY argument. If true and
8569 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8570 emit a note to adjust the CFA offset.
8571 (create_cfi_notes): Adjust scan_trace callers.
8572 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8573 INCOMING_FRAME_SP_OFFSET in the CIE.
8574 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8575 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8577 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8578 * doc/tm.texi: Regenerated.
8580 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8582 PR rtl-optimization/83147
8583 * lra-constraints.c (remove_inheritance_pseudos): Use
8584 lra_substitute_pseudo_within_insn.
8586 2018-01-19 Tom de Vries <tom@codesourcery.com>
8587 Cesar Philippidis <cesar@codesourcery.com>
8590 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8592 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8595 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8596 spaces for function labels.
8598 2018-01-19 Martin Liska <mliska@suse.cz>
8600 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8601 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8602 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8603 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8604 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8605 (PRED_CONST_RETURN): Change from 69 to 65.
8606 (PRED_NULL_RETURN): Change from 91 to 71.
8607 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8608 (PRED_LOOP_GUARD): Change from 66 to 73.
8610 2018-01-19 Martin Liska <mliska@suse.cz>
8612 * predict.c (predict_insn_def): Add new assert.
8613 (struct branch_predictor): Change type to signed integer.
8614 (test_prediction_value_range): Amend test to cover
8616 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8617 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8618 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8619 (PRED_LOOP_IV_COMPARE): Likewise.
8620 * predict.h (PROB_UNINITIALIZED): Define new constant.
8622 2018-01-19 Martin Liska <mliska@suse.cz>
8624 * predict.c (dump_prediction): Add new format for
8625 analyze_brprob.py script which is enabled with -details
8627 * profile-count.h (precise_p): New function.
8629 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8631 PR tree-optimization/83922
8632 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8633 there are no statements that need masking.
8634 (vect_active_double_reduction_p): New function.
8635 (vect_analyze_loop_operations): Use it when handling phis that
8636 are not in the loop header.
8638 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8640 PR tree-optimization/83914
8641 * tree-vect-loop.c (vectorizable_induction): Don't convert
8642 init_expr or apply the peeling adjustment for inductions
8643 that are nested within the vectorized loop.
8645 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8647 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8650 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8654 * function.h (gimplify_parameters): Add gimple_seq * argument.
8655 * function.c: Include gimple.h and options.h.
8656 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8657 for the added local temporaries if needed.
8658 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8659 if there are any parameter cleanups, wrap whole body into a
8660 try/finally with the cleanups.
8662 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8665 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8666 Use GET_MODE_CLASS for scalar floating point.
8668 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8672 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8673 Fix call of call_cgraph_insertion_hooks.
8675 2018-01-18 Martin Sebor <msebor@redhat.com>
8677 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8679 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8682 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8685 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8688 * common.opt: (-ffile-prefix-map): New option.
8689 * opts.c (common_handle_option): Defer it.
8690 * opts-global.c (handle_common_deferred_options): Handle it.
8691 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8692 * file-prefix-map.h: New file.
8693 (remap_debug_filename, add_debug_prefix_map): ...here.
8694 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8695 * final.c (debug_prefix_map, add_debug_prefix_map
8696 remap_debug_filename): Move to...
8697 * file-prefix-map.c: New file.
8698 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8699 generalize, get rid of alloca(), use strrchr() instead of strchr().
8700 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8701 Implement in terms of add_prefix_map().
8702 (remap_macro_filename, remap_debug_filename): Implement in term of
8704 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8705 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8706 * dbxout.c: Include file-prefix-map.h.
8707 * varasm.c: Likewise.
8708 * vmsdbgout.c: Likewise.
8709 * xcoffout.c: Likewise.
8710 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8711 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8712 * doc/invoke.texi (-ffile-prefix-map): Document.
8713 (-fdebug-prefix-map): Update description.
8715 2018-01-18 Martin Liska <mliska@suse.cz>
8717 * config/i386/i386.c (indirect_thunk_name): Document that also
8719 (output_indirect_thunk): Document why both instructions
8720 (pause and lfence) are generated.
8722 2018-01-18 Richard Biener <rguenther@suse.de>
8724 PR tree-optimization/83887
8725 * graphite-scop-detection.c
8726 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8727 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8728 (scop_detection::merge_sese): Re-implement with a flood-fill
8729 algorithm that properly finds a SESE region if it exists.
8731 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8734 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8735 pointer_diff optimizations use view_convert instead of convert.
8737 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8739 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8740 Generate different code for -mno-speculate-indirect-jumps.
8741 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8742 (*call_indirect_aix<mode>): Disable for
8743 -mno-speculate-indirect-jumps.
8744 (*call_indirect_aix<mode>_nospec): New define_insn.
8745 (*call_value_indirect_aix<mode>): Disable for
8746 -mno-speculate-indirect-jumps.
8747 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8748 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8749 -mno-speculate-indirect-jumps.
8750 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8752 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8754 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8755 long double type, set the flags for noting the default long double
8756 type, even if we don't pass or return a long double type.
8758 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8761 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8764 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8766 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8767 support for merge[hl].
8768 (fold_mergehl_helper): New helper function.
8769 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8770 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8771 (altivec_vmrglw_direct): Add xxmrglw insn.
8773 2018-01-17 Andrew Waterman <andrew@sifive.com>
8775 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8776 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8778 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8781 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8782 call the lto_location_cache before reading the
8783 DECL_SOURCE_LOCATION of the types.
8785 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8786 Richard Sandiford <richard.sandiford@linaro.org>
8788 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8789 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8790 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8791 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8792 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8794 * config/aarch64/constraints.md (aarch64_movti_operand):
8796 * config/aarch64/predicates.md (Uti): Add new constraint.
8798 2018-01-17 Carl Love <cel@us.ibm.com>
8800 * config/rs6000/vsx.md (define_expand xl_len_r,
8801 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8802 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8804 (define_expand, define_insn): Move the shift left from the
8805 define_insn to the define_expand for lxvl and stxvl instructions.
8806 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8807 and XL_LEN_R definitions to PURE.
8809 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8811 * config/i386/i386.c (indirect_thunk_name): Declare regno
8812 as unsigned int. Compare regno with INVALID_REGNUM.
8813 (output_indirect_thunk): Ditto.
8814 (output_indirect_thunk_function): Ditto.
8815 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8816 in the call to output_indirect_thunk_function.
8818 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8821 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8822 rather than the size of inner_type to determine the stack slot size
8823 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8825 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8828 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8831 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8833 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8834 endian Linux systems to optionally enable multilibs for selecting
8835 the long double type if the user configured an explicit type.
8836 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8837 have no long double multilibs if not defined.
8838 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8839 warn if the user used -mabi={ieee,ibm}longdouble and we built
8840 multilibs for long double.
8841 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8842 appropriate multilib option.
8843 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8845 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8846 for building long double multilibs.
8847 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8849 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8851 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8854 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8856 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8859 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8862 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8865 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8867 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8868 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8870 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8872 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8873 different rtl trees depending on TARGET_64BIT.
8874 (rs6000_gen_lvx): Likewise.
8876 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8878 * config/visium/visium.md (nop): Tweak comment.
8879 (hazard_nop): Likewise.
8881 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8883 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8884 -mspeculate-indirect-jumps.
8885 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8886 for -mno-speculate-indirect-jumps.
8887 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8888 (*call_value_indirect_elfv2<mode>): Disable for
8889 -mno-speculate-indirect-jumps.
8890 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8891 (indirect_jump): Emit different RTL for
8892 -mno-speculate-indirect-jumps.
8893 (*indirect_jump<mode>): Disable for
8894 -mno-speculate-indirect-jumps.
8895 (*indirect_jump<mode>_nospec): New define_insn.
8896 (tablejump): Emit different RTL for
8897 -mno-speculate-indirect-jumps.
8898 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8899 (tablejumpsi_nospec): New define_expand.
8900 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8901 (tablejumpdi_nospec): New define_expand.
8902 (*tablejump<mode>_internal1): Disable for
8903 -mno-speculate-indirect-jumps.
8904 (*tablejump<mode>_internal1_nospec): New define_insn.
8905 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8908 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8910 * caller-save.c (insert_save): Drop unnecessary parameter. All
8913 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8914 Richard Biener <rguenth@suse.de>
8917 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8918 return early, inline manually is_gimple_sizepos. Make sure if we
8919 call gimplify_expr we don't end up with a gimple constant.
8920 * tree.c (variably_modified_type_p): Don't return true for
8921 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8922 * gimplify.h (is_gimple_sizepos): Remove.
8924 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8926 PR tree-optimization/83857
8927 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8928 vectorizable_live_operation for pure SLP statements.
8929 (vectorizable_live_operation): Handle PHIs.
8931 2018-01-16 Richard Biener <rguenther@suse.de>
8933 PR tree-optimization/83867
8934 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8935 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8937 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8940 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8941 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8942 If off is not INTEGER_CST, issue a may not be aligned warning
8943 rather than isn't aligned. Use isn%'t rather than isn't.
8944 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8946 <case MULT_EXPR>: Improve the case when bottom and one of the
8947 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8948 operand, in that case check if the other operand is multiple of
8949 bottom divided by the INTEGER_CST operand.
8951 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8954 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8955 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8956 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8957 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8958 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8959 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8960 (pa_function_arg_size): New function.
8962 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8964 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8965 in a separate statement.
8967 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8969 PR tree-optimization/83847
8970 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8971 group gathers and scatters.
8973 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8975 PR rtl-optimization/86620
8976 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8978 PR rtl-optimization/83213
8979 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8980 to last if both are JUMP_INSNs.
8982 PR tree-optimization/83843
8983 * gimple-ssa-store-merging.c
8984 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8985 store_immediate_info for bswap/nop orig_stores.
8987 2018-01-15 Andrew Waterman <andrew@sifive.com>
8989 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8991 <UDIV>: Increase cost if !TARGET_DIV.
8993 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8995 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8996 (define_attr "cr_logical_3op"): New.
8997 (cceq_ior_compare): Adjust.
8998 (cceq_ior_compare_complement): Adjust.
8999 (*cceq_rev_compare): Adjust.
9000 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9001 (is_cracked_insn): Adjust.
9002 (insn_must_be_first_in_group): Adjust.
9003 * config/rs6000/40x.md: Adjust.
9004 * config/rs6000/440.md: Adjust.
9005 * config/rs6000/476.md: Adjust.
9006 * config/rs6000/601.md: Adjust.
9007 * config/rs6000/603.md: Adjust.
9008 * config/rs6000/6xx.md: Adjust.
9009 * config/rs6000/7450.md: Adjust.
9010 * config/rs6000/7xx.md: Adjust.
9011 * config/rs6000/8540.md: Adjust.
9012 * config/rs6000/cell.md: Adjust.
9013 * config/rs6000/e300c2c3.md: Adjust.
9014 * config/rs6000/e500mc.md: Adjust.
9015 * config/rs6000/e500mc64.md: Adjust.
9016 * config/rs6000/e5500.md: Adjust.
9017 * config/rs6000/e6500.md: Adjust.
9018 * config/rs6000/mpc.md: Adjust.
9019 * config/rs6000/power4.md: Adjust.
9020 * config/rs6000/power5.md: Adjust.
9021 * config/rs6000/power6.md: Adjust.
9022 * config/rs6000/power7.md: Adjust.
9023 * config/rs6000/power8.md: Adjust.
9024 * config/rs6000/power9.md: Adjust.
9025 * config/rs6000/rs64.md: Adjust.
9026 * config/rs6000/titan.md: Adjust.
9028 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9030 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9031 ix86_indirect_branch_register logic.
9033 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9035 * config/i386/constraints.md (Bs): Update
9036 ix86_indirect_branch_register check. Don't check
9037 ix86_indirect_branch_register with GOT_memory_operand.
9039 * config/i386/predicates.md (GOT_memory_operand): Don't check
9040 ix86_indirect_branch_register here.
9041 (GOT32_symbol_operand): Likewise.
9043 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9045 * config/i386/predicates.md (constant_call_address_operand):
9046 Rewrite ix86_indirect_branch_register logic.
9047 (sibcall_insn_operand): Likewise.
9049 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9051 * config/i386/constraints.md (Bs): Replace
9052 ix86_indirect_branch_thunk_register with
9053 ix86_indirect_branch_register.
9055 * config/i386/i386.md (indirect_jump): Likewise.
9056 (tablejump): Likewise.
9057 (*sibcall_memory): Likewise.
9058 (*sibcall_value_memory): Likewise.
9059 Peepholes of indirect call and jump via memory: Likewise.
9060 * config/i386/i386.opt: Likewise.
9061 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9062 (GOT_memory_operand): Likewise.
9063 (call_insn_operand): Likewise.
9064 (sibcall_insn_operand): Likewise.
9065 (GOT32_symbol_operand): Likewise.
9067 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9070 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9071 type rather than type addr's type points to.
9072 (expand_omp_atomic_mutex): Likewise.
9073 (expand_omp_atomic): Likewise.
9075 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9078 * config/i386/i386.c (output_indirect_thunk_function): Use
9079 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9080 for __x86_return_thunk.
9082 2018-01-15 Richard Biener <rguenther@suse.de>
9085 * expmed.c (extract_bit_field_1): Fix typo.
9087 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9090 * config/arm/iterators.md (VF): New mode iterator.
9091 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9092 Remove integer-related logic from pattern.
9093 (neon_vabd<mode>_3): Likewise.
9095 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9098 * common.opt (fstrict-overflow): No longer an alias.
9099 (fwrapv-pointer): New option.
9100 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9101 also for pointer types based on flag_wrapv_pointer.
9102 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9103 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9104 opts->x_flag_wrapv got set.
9105 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9106 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9107 POINTER_TYPE_OVERFLOW_UNDEFINED.
9108 * match.pd: Likewise in address comparison pattern.
9109 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9111 2018-01-15 Richard Biener <rguenther@suse.de>
9114 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9115 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9116 Reset type names to their identifier if their TYPE_DECL doesn't
9117 have linkage (and thus is used for ODR and devirt).
9118 (save_debug_info_for_decl): Remove.
9119 (save_debug_info_for_type): Likewise.
9120 (add_tree_to_fld_list): Adjust.
9121 * tree-pretty-print.c (dump_generic_node): Make dumping of
9122 type names more robust.
9124 2018-01-15 Richard Biener <rguenther@suse.de>
9126 * BASE-VER: Bump to 8.0.1.
9128 2018-01-14 Martin Sebor <msebor@redhat.com>
9131 * builtins.c (check_access): Avoid warning when the no-warning bit
9134 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9136 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9137 * ira-color (allocno_hard_regs_compare): Likewise.
9139 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9142 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9143 Use .pushsection/.popsection.
9145 2018-01-14 Martin Sebor <msebor@redhat.com>
9148 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9150 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9152 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9153 entry from extra_headers.
9154 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9155 extra_headers, make the list bitwise identical to the i?86-*-* one.
9157 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9159 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9160 -mcmodel=large with -mindirect-branch=thunk,
9161 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9162 -mfunction-return=thunk-extern.
9163 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9164 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9165 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9167 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9169 * config/i386/i386.c (print_reg): Print the name of the full
9170 integer register without '%'.
9171 (ix86_print_operand): Handle 'V'.
9172 * doc/extend.texi: Document 'V' modifier.
9174 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9176 * config/i386/constraints.md (Bs): Disallow memory operand for
9177 -mindirect-branch-register.
9179 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9180 (GOT_memory_operand): Likewise.
9181 (call_insn_operand): Likewise.
9182 (sibcall_insn_operand): Likewise.
9183 (GOT32_symbol_operand): Likewise.
9184 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9185 for -mindirect-branch-register.
9186 (tablejump): Likewise.
9187 (*sibcall_memory): Likewise.
9188 (*sibcall_value_memory): Likewise.
9189 Disallow peepholes of indirect call and jump via memory for
9190 -mindirect-branch-register.
9191 (*call_pop): Replace m with Bw.
9192 (*call_value_pop): Likewise.
9193 (*sibcall_pop_memory): Replace m with Bs.
9194 * config/i386/i386.opt (mindirect-branch-register): New option.
9195 * doc/invoke.texi: Document -mindirect-branch-register option.
9197 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9199 * config/i386/i386-protos.h (ix86_output_function_return): New.
9200 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9201 set function_return_type.
9202 (indirect_thunk_name): Add ret_p to indicate thunk for function
9204 (output_indirect_thunk_function): Pass false to
9205 indirect_thunk_name.
9206 (ix86_output_indirect_branch_via_reg): Likewise.
9207 (ix86_output_indirect_branch_via_push): Likewise.
9208 (output_indirect_thunk_function): Create alias for function
9209 return thunk if regno < 0.
9210 (ix86_output_function_return): New function.
9211 (ix86_handle_fndecl_attribute): Handle function_return.
9212 (ix86_attribute_table): Add function_return.
9213 * config/i386/i386.h (machine_function): Add
9214 function_return_type.
9215 * config/i386/i386.md (simple_return_internal): Use
9216 ix86_output_function_return.
9217 (simple_return_internal_long): Likewise.
9218 * config/i386/i386.opt (mfunction-return=): New option.
9219 (indirect_branch): Mention -mfunction-return=.
9220 * doc/extend.texi: Document function_return function attribute.
9221 * doc/invoke.texi: Document -mfunction-return= option.
9223 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9225 * config/i386/i386-opts.h (indirect_branch): New.
9226 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9227 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9228 with local indirect jump when converting indirect call and jump.
9229 (ix86_set_indirect_branch_type): New.
9230 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9231 (indirectlabelno): New.
9232 (indirect_thunk_needed): Likewise.
9233 (indirect_thunk_bnd_needed): Likewise.
9234 (indirect_thunks_used): Likewise.
9235 (indirect_thunks_bnd_used): Likewise.
9236 (INDIRECT_LABEL): Likewise.
9237 (indirect_thunk_name): Likewise.
9238 (output_indirect_thunk): Likewise.
9239 (output_indirect_thunk_function): Likewise.
9240 (ix86_output_indirect_branch_via_reg): Likewise.
9241 (ix86_output_indirect_branch_via_push): Likewise.
9242 (ix86_output_indirect_branch): Likewise.
9243 (ix86_output_indirect_jmp): Likewise.
9244 (ix86_code_end): Call output_indirect_thunk_function if needed.
9245 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9247 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9248 (ix86_attribute_table): Add indirect_branch.
9249 * config/i386/i386.h (machine_function): Add indirect_branch_type
9250 and has_local_indirect_jump.
9251 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9253 (tablejump): Likewise.
9254 (*indirect_jump): Use ix86_output_indirect_jmp.
9255 (*tablejump_1): Likewise.
9256 (simple_return_indirect_internal): Likewise.
9257 * config/i386/i386.opt (mindirect-branch=): New option.
9258 (indirect_branch): New.
9261 (thunk-inline): Likewise.
9262 (thunk-extern): Likewise.
9263 * doc/extend.texi: Document indirect_branch function attribute.
9264 * doc/invoke.texi: Document -mindirect-branch= option.
9266 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9269 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9271 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9273 * ipa-inline.c (want_inline_small_function_p): Return false if
9274 inlining has already failed with CIF_FINAL_ERROR.
9275 (update_caller_keys): Call want_inline_small_function_p before
9277 (update_callee_keys): Likewise.
9279 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9281 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9283 (rs6000_quadword_masked_address_p): Likewise.
9284 (quad_aligned_load_p): Likewise.
9285 (quad_aligned_store_p): Likewise.
9286 (const_load_sequence_p): Add comment to describe the outer-most loop.
9287 (mimic_memory_attributes_and_flags): New function.
9288 (rs6000_gen_stvx): Likewise.
9289 (replace_swapped_aligned_store): Likewise.
9290 (rs6000_gen_lvx): Likewise.
9291 (replace_swapped_aligned_load): Likewise.
9292 (replace_swapped_load_constant): Capitalize argument name in
9293 comment describing this function.
9294 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9295 and stores that access quad-word aligned addresses and replace
9296 with stvx or lvx instructions when appropriate.
9297 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9298 New function prototype.
9299 (rs6000_quadword_masked_address_p): Likewise.
9300 (rs6000_gen_lvx): Likewise.
9301 (rs6000_gen_stvx): Likewise.
9302 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9303 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9304 when memory address is aligned.
9305 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9306 this split to select lvx instruction when memory address is aligned.
9307 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9308 instruction when memory address is aligned.
9309 (*vsx_le_perm_load_v16qi): Likewise.
9310 (four unnamed splitters): Modify to select the stvx instruction
9311 when memory is aligned.
9313 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9315 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9316 which appears in the queue multiple times.
9318 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9319 Alan Hayward <alan.hayward@arm.com>
9320 David Sherwood <david.sherwood@arm.com>
9322 * tree-vectorizer.h (vec_lower_bound): New structure.
9323 (_loop_vec_info): Add check_nonzero and lower_bounds.
9324 (LOOP_VINFO_CHECK_NONZERO): New macro.
9325 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9326 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9327 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9328 fields. Make seg_len the distance travelled, not including the
9330 (dr_direction_indicator): Declare.
9331 (dr_zero_step_indicator): Likewise.
9332 (dr_known_forward_stride_p): Likewise.
9333 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9335 (runtime_alias_check_p): Allow runtime alias checks with
9337 (operator ==): Compare access_size and align.
9338 (prune_runtime_alias_test_list): Rework for new distinction between
9339 the access_size and seg_len.
9340 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9342 (get_segment_min_max): New function.
9343 (create_intersect_range_checks): Use it.
9344 (dr_step_indicator): New function.
9345 (dr_direction_indicator): Likewise.
9346 (dr_zero_step_indicator): Likewise.
9347 (dr_known_forward_stride_p): Likewise.
9348 * tree-loop-distribution.c (data_ref_segment_size): Return
9349 DR_STEP * (niters - 1).
9350 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9352 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9353 (vect_preserves_scalar_order_p): New function, split out from...
9354 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9355 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9356 (vect_vfa_access_size): New function.
9357 (vect_vfa_align): Likewise.
9358 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9359 (dump_lower_bound): New function.
9360 (vect_check_lower_bound): Likewise.
9361 (vect_small_gap_p): Likewise.
9362 (vectorizable_with_step_bound_p): Likewise.
9363 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9364 depencies if the vectorization factor is 1. Convert the checks
9365 for nonzero steps into checks on the bounds of DR_STEP. Try using
9366 a bunds check for variable steps if the minimum required step is
9367 relatively small. Update calls to the dr_with_seg_len
9368 constructor and to vect_compile_time_alias.
9369 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9371 (vect_loop_versioning): Call it.
9372 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9374 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9376 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9377 Alan Hayward <alan.hayward@arm.com>
9378 David Sherwood <david.sherwood@arm.com>
9380 * doc/sourcebuild.texi (vect_scatter_store): Document.
9381 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9383 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9385 * genopinit.c (main): Add supports_vec_scatter_store and
9386 supports_vec_scatter_store_cached to target_optabs.
9387 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9388 IFN_MASK_SCATTER_STORE.
9389 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9391 * internal-fn.h (internal_store_fn_p): Declare.
9392 (internal_fn_stored_value_index): Likewise.
9393 * internal-fn.c (scatter_store_direct): New macro.
9394 (expand_scatter_store_optab_fn): New function.
9395 (direct_scatter_store_optab_supported_p): New macro.
9396 (internal_store_fn_p): New function.
9397 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9398 IFN_MASK_SCATTER_STORE.
9399 (internal_fn_mask_index): Likewise.
9400 (internal_fn_stored_value_index): New function.
9401 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9403 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9404 * optabs-query.c (supports_vec_scatter_store_p): New function.
9405 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9406 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9407 true for scatter stores.
9408 (vect_gather_scatter_fn_p): Handle scatter stores too.
9409 (vect_check_gather_scatter): Consider using scatter stores if
9410 supports_vec_scatter_store_p.
9411 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9413 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9414 internal_fn_stored_value_index.
9415 (check_load_store_masking): Handle scatter stores too.
9416 (vect_get_store_rhs): Make public.
9417 (vectorizable_call): Use internal_store_fn_p.
9418 (vectorizable_store): Handle scatter store internal functions.
9419 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9420 when deciding whether the end of the group has been reached.
9421 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9422 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9423 (mask_scatter_store<mode>): New insns.
9425 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9426 Alan Hayward <alan.hayward@arm.com>
9427 David Sherwood <david.sherwood@arm.com>
9429 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9430 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9431 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9433 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9434 Use vect_truncate_gather_scatter_offset if we can't treat the
9435 operation as a normal gather load or scatter store.
9436 (get_group_load_store_type): Take the gather_scatter_info
9437 as argument. Try using a gather load or scatter store for
9438 single-element groups.
9439 (get_load_store_type): Update calls to get_group_load_store_type
9440 and vect_use_strided_gather_scatters_p.
9442 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9443 Alan Hayward <alan.hayward@arm.com>
9444 David Sherwood <david.sherwood@arm.com>
9446 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9447 optional tree argument.
9448 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9450 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9451 but continue to use the current value as a fallback.
9452 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9453 to compare the updates.
9454 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9455 (get_load_store_type): Use it when handling a strided access.
9456 (vect_get_strided_load_store_ops): New function.
9457 (vect_get_data_ptr_increment): Likewise.
9458 (vectorizable_load): Handle strided gather loads. Always pass
9459 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9461 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9462 Alan Hayward <alan.hayward@arm.com>
9463 David Sherwood <david.sherwood@arm.com>
9465 * doc/md.texi (gather_load@var{m}): Document.
9466 (mask_gather_load@var{m}): Likewise.
9467 * genopinit.c (main): Add supports_vec_gather_load and
9468 supports_vec_gather_load_cached to target_optabs.
9469 * optabs-tree.c (init_tree_optimization_optabs): Use
9470 ggc_cleared_alloc to allocate target_optabs.
9471 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9472 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9474 * internal-fn.h (internal_load_fn_p): Declare.
9475 (internal_gather_scatter_fn_p): Likewise.
9476 (internal_fn_mask_index): Likewise.
9477 (internal_gather_scatter_fn_supported_p): Likewise.
9478 * internal-fn.c (gather_load_direct): New macro.
9479 (expand_gather_load_optab_fn): New function.
9480 (direct_gather_load_optab_supported_p): New macro.
9481 (direct_internal_fn_optab): New function.
9482 (internal_load_fn_p): Likewise.
9483 (internal_gather_scatter_fn_p): Likewise.
9484 (internal_fn_mask_index): Likewise.
9485 (internal_gather_scatter_fn_supported_p): Likewise.
9486 * optabs-query.c (supports_at_least_one_mode_p): New function.
9487 (supports_vec_gather_load_p): Likewise.
9488 * optabs-query.h (supports_vec_gather_load_p): Declare.
9489 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9490 and memory_type field.
9491 (NUM_PATTERNS): Bump to 15.
9492 * tree-vect-data-refs.c: Include internal-fn.h.
9493 (vect_gather_scatter_fn_p): New function.
9494 (vect_describe_gather_scatter_call): Likewise.
9495 (vect_check_gather_scatter): Try using internal functions for
9496 gather loads. Recognize existing calls to a gather load function.
9497 (vect_analyze_data_refs): Consider using gather loads if
9498 supports_vec_gather_load_p.
9499 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9500 (vect_get_gather_scatter_offset_type): Likewise.
9501 (vect_convert_mask_for_vectype): Likewise.
9502 (vect_add_conversion_to_patterm): Likewise.
9503 (vect_try_gather_scatter_pattern): Likewise.
9504 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9505 (vect_vect_recog_func_ptrs): Add it.
9506 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9507 internal_fn_mask_index and internal_gather_scatter_fn_p.
9508 (check_load_store_masking): Take the gather_scatter_info as an
9509 argument and handle gather loads.
9510 (vect_get_gather_scatter_ops): New function.
9511 (vectorizable_call): Check internal_load_fn_p.
9512 (vectorizable_load): Likewise. Handle gather load internal
9514 (vectorizable_store): Update call to check_load_store_masking.
9515 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9516 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9517 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9518 (aarch64_gather_scale_operand_d): New predicates.
9519 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9520 (mask_gather_load<mode>): New insns.
9522 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9523 Alan Hayward <alan.hayward@arm.com>
9524 David Sherwood <david.sherwood@arm.com>
9526 * optabs.def (fold_left_plus_optab): New optab.
9527 * doc/md.texi (fold_left_plus_@var{m}): Document.
9528 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9529 * internal-fn.c (fold_left_direct): Define.
9530 (expand_fold_left_optab_fn): Likewise.
9531 (direct_fold_left_optab_supported_p): Likewise.
9532 * fold-const-call.c (fold_const_fold_left): New function.
9533 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9534 * tree-parloops.c (valid_reduction_p): New function.
9535 (gather_scalar_reductions): Use it.
9536 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9537 (vect_finish_replace_stmt): Declare.
9538 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9539 (needs_fold_left_reduction_p): New function, split out from...
9540 (vect_is_simple_reduction): ...here. Accept reductions that
9541 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9542 (vect_force_simple_reduction): Also store the reduction type in
9543 the assignment's STMT_VINFO_REDUC_TYPE.
9544 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9545 (merge_with_identity): New function.
9546 (vect_expand_fold_left): Likewise.
9547 (vectorize_fold_left_reduction): Likewise.
9548 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9549 scalar phi in place for it. Check for target support and reject
9550 cases that would reassociate the operation. Defer the transform
9551 phase to vectorize_fold_left_reduction.
9552 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9553 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9554 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9556 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9558 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9559 call to ifc_temp_var.
9561 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9562 Alan Hayward <alan.hayward@arm.com>
9563 David Sherwood <david.sherwood@arm.com>
9565 * target.def (legitimize_address_displacement): Take the original
9566 offset as a poly_int.
9567 * targhooks.h (default_legitimize_address_displacement): Update
9569 * targhooks.c (default_legitimize_address_displacement): Likewise.
9570 * doc/tm.texi: Regenerate.
9571 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9572 as an argument, moving assert of ad->disp == ad->disp_term to...
9573 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9574 Try calling targetm.legitimize_address_displacement before expanding
9575 the address rather than afterwards, and adjust for the new interface.
9576 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9577 Match the new hook interface. Handle SVE addresses.
9578 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9581 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9583 * Makefile.in (OBJS): Add early-remat.o.
9584 * target.def (select_early_remat_modes): New hook.
9585 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9586 * doc/tm.texi: Regenerate.
9587 * targhooks.h (default_select_early_remat_modes): Declare.
9588 * targhooks.c (default_select_early_remat_modes): New function.
9589 * timevar.def (TV_EARLY_REMAT): New timevar.
9590 * passes.def (pass_early_remat): New pass.
9591 * tree-pass.h (make_pass_early_remat): Declare.
9592 * early-remat.c: New file.
9593 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9595 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
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-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9602 vfm1 with a bound_epilog parameter.
9603 (vect_do_peeling): Update calls accordingly, and move the prologue
9604 call earlier in the function. Treat the base bound_epilog as 0 for
9605 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9606 this base when peeling for gaps.
9607 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9608 with fully-masked loops.
9609 (vect_estimate_min_profitable_iters): Handle the single peeled
9610 iteration in that case.
9612 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9613 Alan Hayward <alan.hayward@arm.com>
9614 David Sherwood <david.sherwood@arm.com>
9616 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9617 single-element interleaving even if the size is not a power of 2.
9618 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9619 accesses for single-element interleaving if the group size is
9622 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9623 Alan Hayward <alan.hayward@arm.com>
9624 David Sherwood <david.sherwood@arm.com>
9626 * doc/md.texi (fold_extract_last_@var{m}): Document.
9627 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9628 * optabs.def (fold_extract_last_optab): New optab.
9629 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9630 * internal-fn.c (fold_extract_direct): New macro.
9631 (expand_fold_extract_optab_fn): Likewise.
9632 (direct_fold_extract_optab_supported_p): Likewise.
9633 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9634 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9635 EXTRACT_LAST_REDUCTION.
9636 (get_initial_def_for_reduction): Do not create an initial vector
9637 for EXTRACT_LAST_REDUCTION reductions.
9638 (vectorizable_reduction): Leave the scalar phi in place for
9639 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9640 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9641 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9642 transform phase to vectorizable_condition.
9643 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9645 (vect_finish_stmt_generation): ...here.
9646 (vect_finish_replace_stmt): New function.
9647 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9648 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9650 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9652 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9653 Alan Hayward <alan.hayward@arm.com>
9654 David Sherwood <david.sherwood@arm.com>
9656 * doc/md.texi (extract_last_@var{m}): Document.
9657 * optabs.def (extract_last_optab): New optab.
9658 * internal-fn.def (EXTRACT_LAST): New internal function.
9659 * internal-fn.c (cond_unary_direct): New macro.
9660 (expand_cond_unary_optab_fn): Likewise.
9661 (direct_cond_unary_optab_supported_p): Likewise.
9662 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9663 loops using EXTRACT_LAST.
9664 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9665 (extract_last_<mode>): ...this optab.
9666 (vec_extract<mode><Vel>): Update accordingly.
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 * target.def (empty_mask_is_expensive): New hook.
9673 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9674 * doc/tm.texi: Regenerate.
9675 * targhooks.h (default_empty_mask_is_expensive): Declare.
9676 * targhooks.c (default_empty_mask_is_expensive): New function.
9677 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9678 if the target says that empty masks are expensive.
9679 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9681 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9683 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9684 Alan Hayward <alan.hayward@arm.com>
9685 David Sherwood <david.sherwood@arm.com>
9687 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9688 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9689 (vect_use_loop_mask_for_alignment_p): New function.
9690 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9691 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9692 niters_skip argument. Make sure that the first niters_skip elements
9693 of the first iteration are inactive.
9694 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9695 Update call to vect_set_loop_masks_directly.
9696 (get_misalign_in_elems): New function, split out from...
9697 (vect_gen_prolog_loop_niters): ...here.
9698 (vect_update_init_of_dr): Take a code argument that specifies whether
9699 the adjustment should be added or subtracted.
9700 (vect_update_init_of_drs): Likewise.
9701 (vect_prepare_for_masked_peels): New function.
9702 (vect_do_peeling): Skip prologue peeling if we're using a mask
9703 instead. Update call to vect_update_inits_of_drs.
9704 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9706 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9707 alignment. Do not include the number of peeled iterations in
9708 the minimum threshold in that case.
9709 (vectorizable_induction): Adjust the start value down by
9710 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9711 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9712 Take the number of skipped iterations into account when calculating
9714 * tree-vect-stmts.c (vect_gen_while_not): New function.
9716 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9717 Alan Hayward <alan.hayward@arm.com>
9718 David Sherwood <david.sherwood@arm.com>
9720 * doc/sourcebuild.texi (vect_fully_masked): Document.
9721 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9723 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9725 (vect_analyze_loop_2): ...here. Don't check the vectorization
9726 factor against the number of loop iterations if the loop is
9729 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9730 Alan Hayward <alan.hayward@arm.com>
9731 David Sherwood <david.sherwood@arm.com>
9733 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9734 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9735 (dump_groups): Update accordingly.
9736 (iv_use::mem_type): New member variable.
9737 (address_p): New function.
9738 (record_use): Add a mem_type argument and initialize the new
9740 (record_group_use): Add a mem_type argument. Use address_p.
9741 Remove obsolete null checks of base_object. Update call to record_use.
9742 (find_interesting_uses_op): Update call to record_group_use.
9743 (find_interesting_uses_cond): Likewise.
9744 (find_interesting_uses_address): Likewise.
9745 (get_mem_type_for_internal_fn): New function.
9746 (find_address_like_use): Likewise.
9747 (find_interesting_uses_stmt): Try find_address_like_use before
9748 calling find_interesting_uses_op.
9749 (addr_offset_valid_p): Use the iv mem_type field as the type
9750 of the addressed memory.
9751 (add_autoinc_candidates): Likewise.
9752 (get_address_cost): Likewise.
9753 (split_small_address_groups_p): Use address_p.
9754 (split_address_groups): Likewise.
9755 (add_iv_candidate_for_use): Likewise.
9756 (autoinc_possible_for_pair): Likewise.
9757 (rewrite_groups): Likewise.
9758 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9759 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9760 (get_alias_ptr_type_for_ptr_address): New function.
9761 (rewrite_use_address): Rewrite address uses in calls that were
9762 identified by find_address_like_use.
9764 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9765 Alan Hayward <alan.hayward@arm.com>
9766 David Sherwood <david.sherwood@arm.com>
9768 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9770 * gimple-expr.h (is_gimple_addressable: Likewise.
9771 * gimple-expr.c (is_gimple_address): Likewise.
9772 * internal-fn.c (expand_call_mem_ref): New function.
9773 (expand_mask_load_optab_fn): Use it.
9774 (expand_mask_store_optab_fn): Likewise.
9776 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9777 Alan Hayward <alan.hayward@arm.com>
9778 David Sherwood <david.sherwood@arm.com>
9780 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9781 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9782 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9783 (cond_umax@var{mode}): Document.
9784 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9785 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9786 (cond_umin_optab, cond_umax_optab): New optabs.
9787 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9788 (COND_IOR, COND_XOR): New internal functions.
9789 * internal-fn.h (get_conditional_internal_fn): Declare.
9790 * internal-fn.c (cond_binary_direct): New macro.
9791 (expand_cond_binary_optab_fn): Likewise.
9792 (direct_cond_binary_optab_supported_p): Likewise.
9793 (get_conditional_internal_fn): New function.
9794 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9795 Cope with reduction statements that are vectorized as calls rather
9797 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9798 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9799 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9800 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9801 (UNSPEC_COND_EOR): New unspecs.
9802 (optab): Add mappings for them.
9803 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9804 (sve_int_op, sve_fp_op): New int attributes.
9806 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9807 Alan Hayward <alan.hayward@arm.com>
9808 David Sherwood <david.sherwood@arm.com>
9810 * optabs.def (while_ult_optab): New optab.
9811 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9812 * internal-fn.def (WHILE_ULT): New internal function.
9813 * internal-fn.h (direct_internal_fn_supported_p): New override
9814 that takes two types as argument.
9815 * internal-fn.c (while_direct): New macro.
9816 (expand_while_optab_fn): New function.
9817 (convert_optab_supported_p): Likewise.
9818 (direct_while_optab_supported_p): New macro.
9819 * wide-int.h (wi::udiv_ceil): New function.
9820 * tree-vectorizer.h (rgroup_masks): New structure.
9821 (vec_loop_masks): New typedef.
9822 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9824 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9825 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9826 (vect_max_vf): New function.
9827 (slpeel_make_loop_iterate_ntimes): Delete.
9828 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9829 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9830 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9831 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9832 internal-fn.h, stor-layout.h and optabs-query.h.
9833 (vect_set_loop_mask): New function.
9834 (add_preheader_seq): Likewise.
9835 (add_header_seq): Likewise.
9836 (interleave_supported_p): Likewise.
9837 (vect_maybe_permute_loop_masks): Likewise.
9838 (vect_set_loop_masks_directly): Likewise.
9839 (vect_set_loop_condition_masked): Likewise.
9840 (vect_set_loop_condition_unmasked): New function, split out from
9841 slpeel_make_loop_iterate_ntimes.
9842 (slpeel_make_loop_iterate_ntimes): Rename to..
9843 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9844 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9845 (vect_do_peeling): Update call accordingly.
9846 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9848 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9849 mask_compare_type, can_fully_mask_p and fully_masked_p.
9850 (release_vec_loop_masks): New function.
9851 (_loop_vec_info): Use it to free the loop masks.
9852 (can_produce_all_loop_masks_p): New function.
9853 (vect_get_max_nscalars_per_iter): Likewise.
9854 (vect_verify_full_masking): Likewise.
9855 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9856 retries, and free the mask rgroups before retrying. Check loop-wide
9857 reasons for disallowing fully-masked loops. Make the final decision
9858 about whether use a fully-masked loop or not.
9859 (vect_estimate_min_profitable_iters): Do not assume that peeling
9860 for the number of iterations will be needed for fully-masked loops.
9861 (vectorizable_reduction): Disable fully-masked loops.
9862 (vectorizable_live_operation): Likewise.
9863 (vect_halve_mask_nunits): New function.
9864 (vect_double_mask_nunits): Likewise.
9865 (vect_record_loop_mask): Likewise.
9866 (vect_get_loop_mask): Likewise.
9867 (vect_transform_loop): Handle the case in which the final loop
9868 iteration might handle a partial vector. Call vect_set_loop_condition
9869 instead of slpeel_make_loop_iterate_ntimes.
9870 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9871 (check_load_store_masking): New function.
9872 (prepare_load_store_mask): Likewise.
9873 (vectorizable_store): Handle fully-masked loops.
9874 (vectorizable_load): Likewise.
9875 (supportable_widening_operation): Use vect_halve_mask_nunits for
9877 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9878 (vect_gen_while): New function.
9879 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9880 (aarch64_uqdec<mode>): New insn.
9882 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9883 Alan Hayward <alan.hayward@arm.com>
9884 David Sherwood <david.sherwood@arm.com>
9886 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9887 (reduc_xor_scal_optab): New optabs.
9888 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9889 (reduc_xor_scal_@var{m}): Document.
9890 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9891 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9893 * fold-const-call.c (fold_const_call): Handle them.
9894 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9895 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9896 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9897 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9898 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9899 (UNSPEC_XORV): New unspecs.
9900 (optab): Add entries for them.
9901 (BITWISEV): New int iterator.
9902 (bit_reduc_op): New int attributes.
9904 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9905 Alan Hayward <alan.hayward@arm.com>
9906 David Sherwood <david.sherwood@arm.com>
9908 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9909 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9910 * optabs.def (vec_shl_insert_optab): New optab.
9911 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9912 (duplicate_and_interleave): Likewise.
9913 * tree-vect-loop.c: Include internal-fn.h.
9914 (neutral_op_for_slp_reduction): New function, split out from
9915 get_initial_defs_for_reduction.
9916 (get_initial_def_for_reduction): Handle option 2 for variable-length
9917 vectors by loading the neutral value into a vector and then shifting
9918 the initial value into element 0.
9919 (get_initial_defs_for_reduction): Replace the code argument with
9920 the neutral value calculated by neutral_op_for_slp_reduction.
9921 Use gimple_build_vector for constant-length vectors.
9922 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9923 but the first group_size elements have a neutral value.
9924 Use duplicate_and_interleave otherwise.
9925 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9926 Update call to get_initial_defs_for_reduction. Handle SLP
9927 reductions for variable-length vectors by creating one vector
9928 result for each scalar result, with the elements associated
9929 with other scalar results stubbed out with the neutral value.
9930 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9931 Require IFN_VEC_SHL_INSERT for double reductions on
9932 variable-length vectors, or SLP reductions that have
9933 a neutral value. Require can_duplicate_and_interleave_p
9934 support for variable-length unchained SLP reductions if there
9935 is no neutral value, such as for MIN/MAX reductions. Also require
9936 the number of vector elements to be a multiple of the number of
9937 SLP statements when doing variable-length unchained SLP reductions.
9938 Update call to vect_create_epilog_for_reduction.
9939 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9940 and remove initial values.
9941 (duplicate_and_interleave): Make public.
9942 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9943 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9945 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9946 Alan Hayward <alan.hayward@arm.com>
9947 David Sherwood <david.sherwood@arm.com>
9949 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9950 (can_duplicate_and_interleave_p): New function.
9951 (vect_get_and_check_slp_defs): Take the vector of statements
9952 rather than just the current one. Remove excess parentheses.
9953 Restriction rejectinon of vect_constant_def and vect_external_def
9954 for variable-length vectors to boolean types, or types for which
9955 can_duplicate_and_interleave_p is false.
9956 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9957 (duplicate_and_interleave): New function.
9958 (vect_get_constant_vectors): Use gimple_build_vector for
9959 constant-length vectors and suitable variable-length constant
9960 vectors. Use duplicate_and_interleave for other variable-length
9961 vectors. Don't defer the update when inserting new statements.
9963 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9964 Alan Hayward <alan.hayward@arm.com>
9965 David Sherwood <david.sherwood@arm.com>
9967 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9968 min_profitable_iters doesn't go negative.
9970 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9971 Alan Hayward <alan.hayward@arm.com>
9972 David Sherwood <david.sherwood@arm.com>
9974 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9975 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9976 * optabs.def (vec_mask_load_lanes_optab): New optab.
9977 (vec_mask_store_lanes_optab): Likewise.
9978 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9979 (MASK_STORE_LANES): Likewise.
9980 * internal-fn.c (mask_load_lanes_direct): New macro.
9981 (mask_store_lanes_direct): Likewise.
9982 (expand_mask_load_optab_fn): Handle masked operations.
9983 (expand_mask_load_lanes_optab_fn): New macro.
9984 (expand_mask_store_optab_fn): Handle masked operations.
9985 (expand_mask_store_lanes_optab_fn): New macro.
9986 (direct_mask_load_lanes_optab_supported_p): Likewise.
9987 (direct_mask_store_lanes_optab_supported_p): Likewise.
9988 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9990 (vect_load_lanes_supported): Likewise.
9991 * tree-vect-data-refs.c (strip_conversion): New function.
9992 (can_group_stmts_p): Likewise.
9993 (vect_analyze_data_ref_accesses): Use it instead of checking
9994 for a pair of assignments.
9995 (vect_store_lanes_supported): Take a masked_p parameter.
9996 (vect_load_lanes_supported): Likewise.
9997 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9998 vect_store_lanes_supported and vect_load_lanes_supported.
9999 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10000 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10001 parameter. Don't allow gaps for masked accesses.
10002 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10003 and vect_load_lanes_supported.
10004 (get_load_store_type): Take a masked_p parameter and update
10005 call to get_group_load_store_type.
10006 (vectorizable_store): Update call to get_load_store_type.
10007 Handle IFN_MASK_STORE_LANES.
10008 (vectorizable_load): Update call to get_load_store_type.
10009 Handle IFN_MASK_LOAD_LANES.
10011 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10012 Alan Hayward <alan.hayward@arm.com>
10013 David Sherwood <david.sherwood@arm.com>
10015 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10017 * config/aarch64/aarch64-protos.h
10018 (aarch64_sve_struct_memory_operand_p): Declare.
10019 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10020 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10021 (VPRED, vpred): Handle SVE structure modes.
10022 * config/aarch64/constraints.md (Utx): New constraint.
10023 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10024 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10025 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10026 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10027 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10028 structure modes. Split into pieces after RA.
10029 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10030 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10032 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10033 SVE structure modes.
10034 (aarch64_classify_address): Likewise.
10035 (sizetochar): Move earlier in file.
10036 (aarch64_print_operand): Handle SVE register lists.
10037 (aarch64_array_mode): New function.
10038 (aarch64_sve_struct_memory_operand_p): Likewise.
10039 (TARGET_ARRAY_MODE): Redefine.
10041 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10042 Alan Hayward <alan.hayward@arm.com>
10043 David Sherwood <david.sherwood@arm.com>
10045 * target.def (array_mode): New target hook.
10046 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10047 * doc/tm.texi: Regenerate.
10048 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10049 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10050 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10051 targetm.array_mode.
10052 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10055 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10056 Alan Hayward <alan.hayward@arm.com>
10057 David Sherwood <david.sherwood@arm.com>
10059 * fold-const.c (fold_binary_loc): Check the argument types
10060 rather than the result type when testing for a vector operation.
10062 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10064 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10065 * doc/tm.texi: Regenerate.
10067 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10068 Alan Hayward <alan.hayward@arm.com>
10069 David Sherwood <david.sherwood@arm.com>
10071 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10072 (sve): Document new AArch64 extension.
10073 * doc/md.texi (w): Extend the description of the AArch64
10074 constraint to include SVE vectors.
10075 (Upl, Upa): Document new AArch64 predicate constraints.
10076 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10078 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10079 (msve-vector-bits=): New option.
10080 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10081 SVE when these are disabled.
10082 (sve): New extension.
10083 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10084 modes. Adjust their number of units based on aarch64_sve_vg.
10085 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10086 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10087 aarch64_addr_query_type.
10088 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10089 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10090 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10091 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10092 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10093 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10094 (aarch64_simd_imm_zero_p): Delete.
10095 (aarch64_check_zero_based_sve_index_immediate): Declare.
10096 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10097 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10098 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10099 (aarch64_sve_float_mul_immediate_p): Likewise.
10100 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10101 rather than an rtx.
10102 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10103 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10104 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10105 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10106 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10107 (aarch64_regmode_natural_size): Likewise.
10108 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10109 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10111 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10112 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10113 for VG and the SVE predicate registers.
10114 (V_ALIASES): Add a "z"-prefixed alias.
10115 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10116 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10117 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10118 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10119 (REG_CLASS_NAMES): Add entries for them.
10120 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10121 and the predicate registers.
10122 (aarch64_sve_vg): Declare.
10123 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10124 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10125 (REGMODE_NATURAL_SIZE): Define.
10126 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10128 * config/aarch64/aarch64.c: Include cfgrtl.h.
10129 (simd_immediate_info): Add a constructor for series vectors,
10130 and an associated step field.
10131 (aarch64_sve_vg): New variable.
10132 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10133 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10134 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10135 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10136 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10137 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10138 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10139 (aarch64_get_mask_mode): New functions.
10140 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10141 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10142 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10143 predicate modes and predicate registers. Explicitly restrict
10144 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10145 to store a vector mode if it is recognized by
10146 aarch64_classify_vector_mode.
10147 (aarch64_regmode_natural_size): New function.
10148 (aarch64_hard_regno_caller_save_mode): Return the original mode
10150 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10151 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10152 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10153 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10155 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10156 does not overlap dest if the function is frame-related. Handle
10158 (aarch64_split_add_offset): New function.
10159 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10160 them aarch64_add_offset.
10161 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10162 and update call to aarch64_sub_sp.
10163 (aarch64_add_cfa_expression): New function.
10164 (aarch64_expand_prologue): Pass extra temporary registers to the
10165 functions above. Handle the case in which we need to emit new
10166 DW_CFA_expressions for registers that were originally saved
10167 relative to the stack pointer, but now have to be expressed
10168 relative to the frame pointer.
10169 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10171 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10172 IP0 and IP1 values for SVE frames.
10173 (aarch64_expand_vec_series): New function.
10174 (aarch64_expand_sve_widened_duplicate): Likewise.
10175 (aarch64_expand_sve_const_vector): Likewise.
10176 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10177 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10178 into the register, rather than emitting a SET directly.
10179 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10180 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10181 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10182 (offset_9bit_signed_scaled_p): New functions.
10183 (aarch64_replicate_bitmask_imm): New function.
10184 (aarch64_bitmask_imm): Use it.
10185 (aarch64_cannot_force_const_mem): Reject expressions involving
10186 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10187 (aarch64_classify_index): Handle SVE indices, by requiring
10188 a plain register index with a scale that matches the element size.
10189 (aarch64_classify_address): Handle SVE addresses. Assert that
10190 the mode of the address is VOIDmode or an integer mode.
10191 Update call to aarch64_classify_symbol.
10192 (aarch64_classify_symbolic_expression): Update call to
10193 aarch64_classify_symbol.
10194 (aarch64_const_vec_all_in_range_p): New function.
10195 (aarch64_print_vector_float_operand): Likewise.
10196 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10197 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10198 and the FP immediates 1.0 and 0.5.
10199 (aarch64_print_address_internal): Handle SVE addresses.
10200 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10201 (aarch64_regno_regclass): Handle predicate registers.
10202 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10204 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10205 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10206 (aarch64_convert_sve_vector_bits): New function.
10207 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10208 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10209 rather than an rtx.
10210 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10211 Handle SVE vector and predicate modes. Accept VL-based constants
10212 that need only one temporary register, and VL offsets that require
10213 no temporary registers.
10214 (aarch64_conditional_register_usage): Mark the predicate registers
10215 as fixed if SVE isn't available.
10216 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10217 Return true for SVE vector and predicate modes.
10218 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10219 rather than an unsigned int. Handle SVE modes.
10220 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10222 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10224 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10225 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10226 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10227 (aarch64_sve_float_mul_immediate_p): New functions.
10228 (aarch64_sve_valid_immediate): New function.
10229 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10230 Explicitly reject structure modes. Check for INDEX constants.
10231 Handle PTRUE and PFALSE constants.
10232 (aarch64_check_zero_based_sve_index_immediate): New function.
10233 (aarch64_simd_imm_zero_p): Delete.
10234 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10235 vector modes. Accept constants in the range of CNT[BHWD].
10236 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10237 ask for an Advanced SIMD mode.
10238 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10239 (aarch64_simd_vector_alignment): Handle SVE predicates.
10240 (aarch64_vectorize_preferred_vector_alignment): New function.
10241 (aarch64_simd_vector_alignment_reachable): Use it instead of
10243 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10244 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10246 (MAX_VECT_LEN): Delete.
10247 (expand_vec_perm_d): Add a vec_flags field.
10248 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10249 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10250 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10252 (aarch64_evpc_rev): Rename to...
10253 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10254 (aarch64_evpc_rev_global): New function.
10255 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10256 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10258 (aarch64_evpc_sve_tbl): New function.
10259 (aarch64_expand_vec_perm_const_1): Update after rename of
10260 aarch64_evpc_rev. Handle SVE permutes too, trying
10261 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10262 than aarch64_evpc_tbl.
10263 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10264 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10265 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10266 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10267 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10268 (aarch64_expand_sve_vcond): New functions.
10269 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10270 of aarch64_vector_mode_p.
10271 (aarch64_dwarf_poly_indeterminate_value): New function.
10272 (aarch64_compute_pressure_classes): Likewise.
10273 (aarch64_can_change_mode_class): Likewise.
10274 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10275 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10276 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10277 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10278 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10279 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10280 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10281 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10283 (Dn, Dl, Dr): Accept const as well as const_vector.
10284 (Dz): Likewise. Compare against CONST0_RTX.
10285 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10286 of "vector" where appropriate.
10287 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10288 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10289 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10290 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10291 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10292 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10293 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10294 (v_int_equiv): Extend to SVE modes.
10295 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10297 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10298 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10299 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10300 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10301 (SVE_COND_FP_CMP): New int iterators.
10302 (perm_hilo): Handle the new unpack unspecs.
10303 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10305 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10306 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10307 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10308 (aarch64_equality_operator, aarch64_constant_vector_operand)
10309 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10310 (aarch64_sve_nonimmediate_operand): Likewise.
10311 (aarch64_sve_general_operand): Likewise.
10312 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10313 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10314 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10315 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10316 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10317 (aarch64_sve_float_arith_immediate): Likewise.
10318 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10319 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10320 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10321 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10322 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10323 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10324 (aarch64_sve_float_arith_operand): Likewise.
10325 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10326 (aarch64_sve_float_mul_operand): Likewise.
10327 (aarch64_sve_vec_perm_operand): Likewise.
10328 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10329 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10330 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10331 as well as const_vector.
10332 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10333 in file. Use CONST0_RTX and CONSTM1_RTX.
10334 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10335 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10336 Use aarch64_simd_imm_zero.
10337 * config/aarch64/aarch64-sve.md: New file.
10338 * config/aarch64/aarch64.md: Include it.
10339 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10340 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10341 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10342 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10343 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10344 (sve): New attribute.
10345 (enabled): Disable instructions with the sve attribute unless
10347 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10348 aarch64_expand_mov_immediate.
10349 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10350 CNT[BHSD] immediates.
10351 (movti): Split CONST_POLY_INT moves into two halves.
10352 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10353 Split additions that need a temporary here if the destination
10354 is the stack pointer.
10355 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10356 (*add<mode>3_poly_1): New instruction.
10357 (set_clobber_cc): New expander.
10359 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10361 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10362 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10363 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10364 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10365 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10366 Change innermode from fixed_mode_size to machine_mode.
10367 (simplify_subreg): Update call accordingly. Handle a constant-sized
10368 subreg of a variable-length CONST_VECTOR.
10370 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10371 Alan Hayward <alan.hayward@arm.com>
10372 David Sherwood <david.sherwood@arm.com>
10374 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10375 (add_offset_to_base): New function, split out from...
10376 (create_mem_ref): ...here. When handling a scale other than 1,
10377 check first whether the address is valid without the offset.
10378 Add it into the base if so, leaving the index and scale as-is.
10380 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10383 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10384 fold_for_warn before checking if arg2 is INTEGER_CST.
10386 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10388 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10389 (store_multiple_operation): Delete.
10390 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10391 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10392 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10393 guarded by TARGET_STRING.
10394 (rs6000_output_load_multiple): Delete.
10395 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10396 OPTION_MASK_STRING / TARGET_STRING handling.
10397 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10398 (const rs6000_opt_masks) <"string">: Change mask to 0.
10399 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10400 (MASK_STRING): Delete.
10401 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10403 (load_multiple): Delete.
10410 (store_multiple): Delete.
10417 (movmemsi_8reg): Delete.
10418 (corresponding unnamed define_insn): Delete.
10419 (movmemsi_6reg): Delete.
10420 (corresponding unnamed define_insn): Delete.
10421 (movmemsi_4reg): Delete.
10422 (corresponding unnamed define_insn): Delete.
10423 (movmemsi_2reg): Delete.
10424 (corresponding unnamed define_insn): Delete.
10425 (movmemsi_1reg): Delete.
10426 (corresponding unnamed define_insn): Delete.
10427 * config/rs6000/rs6000.opt (mno-string): New.
10428 (mstring): Replace by deprecation warning stub.
10429 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10431 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10433 * regrename.c (regrename_do_replace): If replacing the same
10434 reg multiple times, try to reuse last created gen_raw_REG.
10437 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10438 main to workaround a bug in GDB.
10440 2018-01-12 Tom de Vries <tom@codesourcery.com>
10443 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10445 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10447 PR rtl-optimization/80481
10448 * ira-color.c (get_cap_member): New function.
10449 (allocnos_conflict_by_live_ranges_p): Use it.
10450 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10451 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10453 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10456 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10457 (*saddl_se_1): Ditto.
10458 (*ssubsi_1): Ditto.
10459 (*ssubl_se_1): Ditto.
10461 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10463 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10464 rather than wi::to_widest for DR_INITs.
10465 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10466 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10467 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10469 (vect_analyze_group_access_1): Note that here.
10471 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10473 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10474 polynomial type sizes.
10476 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10478 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10479 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10480 (gimple_add_tmp_var): Likewise.
10482 2018-01-12 Martin Liska <mliska@suse.cz>
10484 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10485 (gimple_alloc_sizes): Likewise.
10486 (dump_gimple_statistics): Use PRIu64 in printf format.
10487 * gimple.h: Change uint64_t to int.
10489 2018-01-12 Martin Liska <mliska@suse.cz>
10491 * tree-core.h: Use uint64_t instead of int.
10492 * tree.c (tree_node_counts): Likewise.
10493 (tree_node_sizes): Likewise.
10494 (dump_tree_statistics): Use PRIu64 in printf format.
10496 2018-01-12 Martin Liska <mliska@suse.cz>
10498 * Makefile.in: As qsort_chk is implemented in vec.c, add
10499 vec.o to linkage of gencfn-macros.
10500 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10501 passing the info to record_node_allocation_statistics.
10502 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10504 * ggc-common.c (struct ggc_usage): Add operator== and use
10505 it in operator< and compare function.
10506 * mem-stats.h (struct mem_usage): Likewise.
10507 * vec.c (struct vec_usage): Remove operator< and compare
10508 function. Can be simply inherited.
10510 2018-01-12 Martin Jambor <mjambor@suse.cz>
10513 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10514 * tree-ssa-math-opts.c: Include domwalk.h.
10515 (convert_mult_to_fma_1): New function.
10516 (fma_transformation_info): New type.
10517 (fma_deferring_state): Likewise.
10518 (cancel_fma_deferring): New function.
10519 (result_of_phi): Likewise.
10520 (last_fma_candidate_feeds_initial_phi): Likewise.
10521 (convert_mult_to_fma): Added deferring logic, split actual
10522 transformation to convert_mult_to_fma_1.
10523 (math_opts_dom_walker): New type.
10524 (math_opts_dom_walker::after_dom_children): New method, body moved
10525 here from pass_optimize_widening_mul::execute, added deferring logic
10527 (pass_optimize_widening_mul::execute): Moved most of code to
10528 math_opts_dom_walker::after_dom_children.
10529 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10530 * config/i386/i386.c (ix86_option_override_internal): Added
10531 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10533 2018-01-12 Richard Biener <rguenther@suse.de>
10536 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10537 inline instance vars.
10539 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10542 * config/rx/rx.c (rx_is_restricted_memory_address):
10543 Handle SUBREG case.
10545 2018-01-12 Richard Biener <rguenther@suse.de>
10547 PR tree-optimization/80846
10548 * target.def (split_reduction): New target hook.
10549 * targhooks.c (default_split_reduction): New function.
10550 * targhooks.h (default_split_reduction): Declare.
10551 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10552 target requests first reduce vectors by combining low and high
10554 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10555 (get_vectype_for_scalar_type_and_size): Export.
10556 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10557 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10558 * doc/tm.texi: Regenerate.
10559 * config/i386/i386.c (ix86_split_reduction): Implement
10560 TARGET_VECTORIZE_SPLIT_REDUCTION.
10562 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10565 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10566 in PIC mode except for TARGET_VXWORKS_RTP.
10567 * config/sparc/sparc.c: Include cfgrtl.h.
10568 (TARGET_INIT_PIC_REG): Define.
10569 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10570 (sparc_pic_register_p): New predicate.
10571 (sparc_legitimate_address_p): Use it.
10572 (sparc_legitimize_pic_address): Likewise.
10573 (sparc_delegitimize_address): Likewise.
10574 (sparc_mode_dependent_address_p): Likewise.
10575 (gen_load_pcrel_sym): Remove 4th parameter.
10576 (load_got_register): Adjust call to above. Remove obsolete stuff.
10577 (sparc_expand_prologue): Do not call load_got_register here.
10578 (sparc_flat_expand_prologue): Likewise.
10579 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10580 (sparc_use_pseudo_pic_reg): New function.
10581 (sparc_init_pic_reg): Likewise.
10582 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10583 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10585 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10587 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10588 Add item for branch_cost.
10590 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10592 PR rtl-optimization/83565
10593 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10594 not extend the result to a larger mode for rotate operations.
10595 (num_sign_bit_copies1): Likewise.
10597 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10600 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10602 Use values-Xc.o for -pedantic.
10603 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10605 2018-01-12 Martin Liska <mliska@suse.cz>
10608 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10610 (possible_polymorphic_call_targets): Use it.
10611 (ipa_devirt): Likewise.
10613 2018-01-12 Martin Liska <mliska@suse.cz>
10615 * profile-count.h (enum profile_quality): Use 0 as invalid
10616 enum value of profile_quality.
10618 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10620 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10621 -mext-string options.
10623 2018-01-12 Richard Biener <rguenther@suse.de>
10625 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10626 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10627 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10629 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10631 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10633 * configure.ac (--with-long-double-format): Add support for the
10634 configuration option to change the default long double format on
10636 * config.gcc (powerpc*-linux*-*): Likewise.
10637 * configure: Regenerate.
10638 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10639 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10640 used without modification.
10642 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10644 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10645 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10646 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10647 MISC_BUILTIN_SPEC_BARRIER.
10648 (rs6000_init_builtins): Likewise.
10649 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10651 (speculation_barrier): New define_insn.
10652 * doc/extend.texi: Document __builtin_speculation_barrier.
10654 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10657 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10658 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10659 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10661 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10662 integral modes instead of "ss" and "sd".
10663 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10664 vectors with 32-bit and 64-bit elements.
10665 (vecdupssescalarmodesuffix): New mode attribute.
10666 (vec_dup<mode>): Use it.
10668 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10671 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10672 frame if argument is passed on stack.
10674 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10677 * ree.c (combine_reaching_defs): Optimize also
10678 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10679 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10681 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10683 PR middle-end/83189
10684 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10686 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10688 PR middle-end/83718
10689 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10690 after they are computed.
10692 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10694 PR tree-optimization/83695
10695 * gimple-loop-linterchange.cc
10696 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10697 reset cached scev information after interchange.
10698 (pass_linterchange::execute): Remove call to scev_reset_htab.
10700 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10702 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10703 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10704 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10705 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10706 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10707 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10708 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10709 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10710 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10711 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10712 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10713 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10714 (V_lane_reg): Likewise.
10715 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10717 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10718 (vfmal_lane_low<mode>_intrinsic,
10719 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10720 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10721 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10722 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10723 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10724 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10726 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10728 * config/arm/arm-cpus.in (fp16fml): New feature.
10729 (ALL_SIMD): Add fp16fml.
10730 (armv8.2-a): Add fp16fml as an option.
10731 (armv8.3-a): Likewise.
10732 (armv8.4-a): Add fp16fml as part of fp16.
10733 * config/arm/arm.h (TARGET_FP16FML): Define.
10734 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10736 * config/arm/arm-modes.def (V2HF): Define.
10737 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10738 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10739 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10740 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10741 vfmsl_low, vfmsl_high): New set of builtins.
10742 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10743 (vfml_op): New code attribute.
10744 (VFMLHALVES): New int iterator.
10745 (VFML, VFMLSEL): New mode attributes.
10746 (V_reg): Define mapping for V2HF.
10747 (V_hi, V_lo): New mode attributes.
10748 (VF_constraint): Likewise.
10749 (vfml_half, vfml_half_selector): New int attributes.
10750 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10752 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10753 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10755 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10756 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10757 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10758 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10760 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10761 Document new effective target and option set.
10763 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10765 * config/arm/arm-cpus.in (armv8_4): New feature.
10766 (ARMv8_4a): New fgroup.
10767 (armv8.4-a): New arch.
10768 * config/arm/arm-tables.opt: Regenerate.
10769 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10770 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10771 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10772 Add matching rules for -march=armv8.4-a and extensions.
10773 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10775 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10778 * config/rx/rx.md (BW): New mode attribute.
10779 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10781 2018-01-11 Richard Biener <rguenther@suse.de>
10783 PR tree-optimization/83435
10784 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10785 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10786 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10788 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10789 Alan Hayward <alan.hayward@arm.com>
10790 David Sherwood <david.sherwood@arm.com>
10792 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10794 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10795 (aarch64_print_address_internal): Use it to check for a zero offset.
10797 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10798 Alan Hayward <alan.hayward@arm.com>
10799 David Sherwood <david.sherwood@arm.com>
10801 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10802 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10803 Return a poly_int64 rather than a HOST_WIDE_INT.
10804 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10805 rather than a HOST_WIDE_INT.
10806 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10807 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10808 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10809 final_offset from HOST_WIDE_INT to poly_int64.
10810 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10811 to_constant when getting the number of units in an Advanced SIMD
10813 (aarch64_builtin_vectorized_function): Check for a constant number
10815 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10817 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10818 attribute instead of GET_MODE_NUNITS.
10819 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10820 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10821 GET_MODE_SIZE for fixed-size registers.
10822 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10823 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10824 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10825 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10826 (aarch64_print_operand, aarch64_print_address_internal)
10827 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10828 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10829 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10830 Handle polynomial GET_MODE_SIZE.
10831 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10832 wider than SImode without modification.
10833 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10834 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10835 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10836 passing and returning SVE modes.
10837 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10838 rather than GEN_INT.
10839 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10840 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10841 (aarch64_allocate_and_probe_stack_space): Likewise.
10842 (aarch64_layout_frame): Cope with polynomial offsets.
10843 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10844 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10845 polynomial offsets.
10846 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10847 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10848 poly_int64 rather than a HOST_WIDE_INT.
10849 (aarch64_get_separate_components, aarch64_process_components)
10850 (aarch64_expand_prologue, aarch64_expand_epilogue)
10851 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10852 (aarch64_anchor_offset): New function, split out from...
10853 (aarch64_legitimize_address): ...here.
10854 (aarch64_builtin_vectorization_cost): Handle polynomial
10855 TYPE_VECTOR_SUBPARTS.
10856 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10858 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10859 number of elements from the PARALLEL rather than the mode.
10860 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10861 rather than GET_MODE_BITSIZE.
10862 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10863 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10864 (aarch64_expand_vec_perm_const_1): Handle polynomial
10865 d->perm.length () and d->perm elements.
10866 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10867 Apply to_constant to d->perm elements.
10868 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10869 polynomial CONST_VECTOR_NUNITS.
10870 (aarch64_move_pointer): Take amount as a poly_int64 rather
10872 (aarch64_progress_pointer): Avoid temporary variable.
10873 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10874 the mode attribute instead of GET_MODE.
10876 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10877 Alan Hayward <alan.hayward@arm.com>
10878 David Sherwood <david.sherwood@arm.com>
10880 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10881 x exists before using it.
10882 (aarch64_add_constant_internal): Rename to...
10883 (aarch64_add_offset_1): ...this. Replace regnum with separate
10884 src and dest rtxes. Handle the case in which they're different,
10885 including when the offset is zero. Replace scratchreg with an rtx.
10886 Use 2 additions if there is no spare register into which we can
10887 move a 16-bit constant.
10888 (aarch64_add_constant): Delete.
10889 (aarch64_add_offset): Replace reg with separate src and dest
10890 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10891 Use aarch64_add_offset_1.
10892 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10893 an rtx rather than an int. Take the delta as a poly_int64
10894 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10895 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10896 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10897 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10898 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10899 and aarch64_add_sp.
10900 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10901 aarch64_add_constant.
10903 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10905 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10906 Use scalar_float_mode.
10908 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10910 * config/aarch64/aarch64-simd.md
10911 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10912 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10913 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10914 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10915 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10916 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10917 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10918 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10919 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10920 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10922 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10925 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10926 targ_options->x_arm_arch_string is non NULL.
10928 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10930 * config/aarch64/aarch64.h
10931 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10933 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10936 * expmed.c (emit_store_flag_force): Swap if const op0
10937 and change VOIDmode to mode of op0.
10939 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10941 PR rtl-optimization/83761
10942 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10943 than bytes to mode_for_size.
10945 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10947 PR middle-end/83189
10948 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10949 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10952 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10954 PR middle-end/83575
10955 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10956 when in layout mode.
10957 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10958 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10961 2018-01-10 Michael Collison <michael.collison@arm.com>
10963 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10964 * config/aarch64/aarch64-option-extension.def: Add
10965 AARCH64_OPT_EXTENSION of 'fp16fml'.
10966 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10967 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10968 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10969 * config/aarch64/constraints.md (Ui7): New constraint.
10970 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10971 (VFMLA_SEL_W): Ditto.
10974 (VFMLA16_LOW): New int iterator.
10975 (VFMLA16_HIGH): Ditto.
10976 (UNSPEC_FMLAL): New unspec.
10977 (UNSPEC_FMLSL): Ditto.
10978 (UNSPEC_FMLAL2): Ditto.
10979 (UNSPEC_FMLSL2): Ditto.
10980 (f16mac): New code attribute.
10981 * config/aarch64/aarch64-simd-builtins.def
10982 (aarch64_fmlal_lowv2sf): Ditto.
10983 (aarch64_fmlsl_lowv2sf): Ditto.
10984 (aarch64_fmlalq_lowv4sf): Ditto.
10985 (aarch64_fmlslq_lowv4sf): Ditto.
10986 (aarch64_fmlal_highv2sf): Ditto.
10987 (aarch64_fmlsl_highv2sf): Ditto.
10988 (aarch64_fmlalq_highv4sf): Ditto.
10989 (aarch64_fmlslq_highv4sf): Ditto.
10990 (aarch64_fmlal_lane_lowv2sf): Ditto.
10991 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10992 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10993 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10994 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10995 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10996 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10997 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10998 (aarch64_fmlal_lane_highv2sf): Ditto.
10999 (aarch64_fmlsl_lane_highv2sf): Ditto.
11000 (aarch64_fmlal_laneq_highv2sf): Ditto.
11001 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11002 (aarch64_fmlalq_lane_highv4sf): Ditto.
11003 (aarch64_fmlsl_lane_highv4sf): Ditto.
11004 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11005 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11006 * config/aarch64/aarch64-simd.md:
11007 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11008 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11009 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11010 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11011 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11012 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11013 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11014 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11015 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11016 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11017 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11018 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11019 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11020 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11021 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11022 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11023 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11024 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11025 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11026 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11027 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11028 (vfmlsl_low_u32): Ditto.
11029 (vfmlalq_low_u32): Ditto.
11030 (vfmlslq_low_u32): Ditto.
11031 (vfmlal_high_u32): Ditto.
11032 (vfmlsl_high_u32): Ditto.
11033 (vfmlalq_high_u32): Ditto.
11034 (vfmlslq_high_u32): Ditto.
11035 (vfmlal_lane_low_u32): Ditto.
11036 (vfmlsl_lane_low_u32): Ditto.
11037 (vfmlal_laneq_low_u32): Ditto.
11038 (vfmlsl_laneq_low_u32): Ditto.
11039 (vfmlalq_lane_low_u32): Ditto.
11040 (vfmlslq_lane_low_u32): Ditto.
11041 (vfmlalq_laneq_low_u32): Ditto.
11042 (vfmlslq_laneq_low_u32): Ditto.
11043 (vfmlal_lane_high_u32): Ditto.
11044 (vfmlsl_lane_high_u32): Ditto.
11045 (vfmlal_laneq_high_u32): Ditto.
11046 (vfmlsl_laneq_high_u32): Ditto.
11047 (vfmlalq_lane_high_u32): Ditto.
11048 (vfmlslq_lane_high_u32): Ditto.
11049 (vfmlalq_laneq_high_u32): Ditto.
11050 (vfmlslq_laneq_high_u32): Ditto.
11051 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11052 (AARCH64_FL_FOR_ARCH8_4): New.
11053 (AARCH64_ISA_F16FML): New ISA flag.
11054 (TARGET_F16FML): New feature flag for fp16fml.
11055 (doc/invoke.texi): Document new fp16fml option.
11057 2018-01-10 Michael Collison <michael.collison@arm.com>
11059 * config/aarch64/aarch64-builtins.c:
11060 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11061 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11062 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11063 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11064 (AARCH64_ISA_SHA3): New ISA flag.
11065 (TARGET_SHA3): New feature flag for sha3.
11066 * config/aarch64/iterators.md (sha512_op): New int attribute.
11067 (CRYPTO_SHA512): New int iterator.
11068 (UNSPEC_SHA512H): New unspec.
11069 (UNSPEC_SHA512H2): Ditto.
11070 (UNSPEC_SHA512SU0): Ditto.
11071 (UNSPEC_SHA512SU1): Ditto.
11072 * config/aarch64/aarch64-simd-builtins.def
11073 (aarch64_crypto_sha512hqv2di): New builtin.
11074 (aarch64_crypto_sha512h2qv2di): Ditto.
11075 (aarch64_crypto_sha512su0qv2di): Ditto.
11076 (aarch64_crypto_sha512su1qv2di): Ditto.
11077 (aarch64_eor3qv8hi): Ditto.
11078 (aarch64_rax1qv2di): Ditto.
11079 (aarch64_xarqv2di): Ditto.
11080 (aarch64_bcaxqv8hi): Ditto.
11081 * config/aarch64/aarch64-simd.md:
11082 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11083 (aarch64_crypto_sha512su0qv2di): Ditto.
11084 (aarch64_crypto_sha512su1qv2di): Ditto.
11085 (aarch64_eor3qv8hi): Ditto.
11086 (aarch64_rax1qv2di): Ditto.
11087 (aarch64_xarqv2di): Ditto.
11088 (aarch64_bcaxqv8hi): Ditto.
11089 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11090 (vsha512h2q_u64): Ditto.
11091 (vsha512su0q_u64): Ditto.
11092 (vsha512su1q_u64): Ditto.
11093 (veor3q_u16): Ditto.
11094 (vrax1q_u64): Ditto.
11095 (vxarq_u64): Ditto.
11096 (vbcaxq_u16): Ditto.
11097 * config/arm/types.md (crypto_sha512): New type attribute.
11098 (crypto_sha3): Ditto.
11099 (doc/invoke.texi): Document new sha3 option.
11101 2018-01-10 Michael Collison <michael.collison@arm.com>
11103 * config/aarch64/aarch64-builtins.c:
11104 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11105 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11106 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11107 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11108 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11109 (AARCH64_ISA_SM4): New ISA flag.
11110 (TARGET_SM4): New feature flag for sm4.
11111 * config/aarch64/aarch64-simd-builtins.def
11112 (aarch64_sm3ss1qv4si): Ditto.
11113 (aarch64_sm3tt1aq4si): Ditto.
11114 (aarch64_sm3tt1bq4si): Ditto.
11115 (aarch64_sm3tt2aq4si): Ditto.
11116 (aarch64_sm3tt2bq4si): Ditto.
11117 (aarch64_sm3partw1qv4si): Ditto.
11118 (aarch64_sm3partw2qv4si): Ditto.
11119 (aarch64_sm4eqv4si): Ditto.
11120 (aarch64_sm4ekeyqv4si): Ditto.
11121 * config/aarch64/aarch64-simd.md:
11122 (aarch64_sm3ss1qv4si): Ditto.
11123 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11124 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11125 (aarch64_sm4eqv4si): Ditto.
11126 (aarch64_sm4ekeyqv4si): Ditto.
11127 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11128 (sm3part_op): Ditto.
11129 (CRYPTO_SM3TT): Ditto.
11130 (CRYPTO_SM3PART): Ditto.
11131 (UNSPEC_SM3SS1): New unspec.
11132 (UNSPEC_SM3TT1A): Ditto.
11133 (UNSPEC_SM3TT1B): Ditto.
11134 (UNSPEC_SM3TT2A): Ditto.
11135 (UNSPEC_SM3TT2B): Ditto.
11136 (UNSPEC_SM3PARTW1): Ditto.
11137 (UNSPEC_SM3PARTW2): Ditto.
11138 (UNSPEC_SM4E): Ditto.
11139 (UNSPEC_SM4EKEY): Ditto.
11140 * config/aarch64/constraints.md (Ui2): New constraint.
11141 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11142 * config/arm/types.md (crypto_sm3): New type attribute.
11143 (crypto_sm4): Ditto.
11144 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11145 (vsm3tt1aq_u32): Ditto.
11146 (vsm3tt1bq_u32): Ditto.
11147 (vsm3tt2aq_u32): Ditto.
11148 (vsm3tt2bq_u32): Ditto.
11149 (vsm3partw1q_u32): Ditto.
11150 (vsm3partw2q_u32): Ditto.
11151 (vsm4eq_u32): Ditto.
11152 (vsm4ekeyq_u32): Ditto.
11153 (doc/invoke.texi): Document new sm4 option.
11155 2018-01-10 Michael Collison <michael.collison@arm.com>
11157 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11158 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11159 (AARCH64_FL_FOR_ARCH8_4): New.
11160 (AARCH64_FL_V8_4): New flag.
11161 (doc/invoke.texi): Document new armv8.4-a option.
11163 2018-01-10 Michael Collison <michael.collison@arm.com>
11165 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11166 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11167 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11168 * config/aarch64/aarch64-option-extension.def: Add
11169 AARCH64_OPT_EXTENSION of 'sha2'.
11170 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11171 (crypto): Disable sha2 and aes if crypto disabled.
11172 (crypto): Enable aes and sha2 if enabled.
11173 (simd): Disable sha2 and aes if simd disabled.
11174 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11176 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11177 (TARGET_SHA2): New feature flag for sha2.
11178 (TARGET_AES): New feature flag for aes.
11179 * config/aarch64/aarch64-simd.md:
11180 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11181 conditional on TARGET_AES.
11182 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11183 (aarch64_crypto_sha1hsi): Make pattern conditional
11185 (aarch64_crypto_sha1hv4si): Ditto.
11186 (aarch64_be_crypto_sha1hv4si): Ditto.
11187 (aarch64_crypto_sha1su1v4si): Ditto.
11188 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11189 (aarch64_crypto_sha1su0v4si): Ditto.
11190 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11191 (aarch64_crypto_sha256su0v4si): Ditto.
11192 (aarch64_crypto_sha256su1v4si): Ditto.
11193 (doc/invoke.texi): Document new aes and sha2 options.
11195 2018-01-10 Martin Sebor <msebor@redhat.com>
11197 PR tree-optimization/83781
11198 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11201 2018-01-11 Martin Sebor <msebor@gmail.com>
11202 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11204 PR tree-optimization/83501
11205 PR tree-optimization/81703
11207 * tree-ssa-strlen.c (get_string_cst): Rename...
11208 (get_string_len): ...to this. Handle global constants.
11209 (handle_char_store): Adjust.
11211 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11212 Jim Wilson <jimw@sifive.com>
11214 * config/riscv/riscv-protos.h (riscv_output_return): New.
11215 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11216 (riscv_attribute_table, riscv_output_return),
11217 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11218 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11219 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11221 (riscv_expand_prologue): Add early return for naked function.
11222 (riscv_expand_epilogue): Likewise.
11223 (riscv_function_ok_for_sibcall): Return false for naked function.
11224 (riscv_set_current_function): New.
11225 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11226 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11227 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11228 * doc/extend.texi (RISC-V Function Attributes): New.
11230 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11232 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11233 check for 128-bit long double before checking TCmode.
11234 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11235 128-bit long doubles before checking TFmode or TCmode.
11236 (FLOAT128_IBM_P): Likewise.
11238 2018-01-10 Martin Sebor <msebor@redhat.com>
11240 PR tree-optimization/83671
11241 * builtins.c (c_strlen): Unconditionally return zero for the empty
11243 Use -Warray-bounds for warnings.
11244 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11245 for non-constant array indices with COMPONENT_REF, arrays of
11246 arrays, and pointers to arrays.
11247 (gimple_fold_builtin_strlen): Determine and set length range for
11248 non-constant character arrays.
11250 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11252 PR middle-end/81897
11253 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11256 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11258 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11260 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11263 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11264 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11265 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11266 indexed_or_indirect_operand predicate.
11267 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11268 (*vsx_le_perm_load_v8hi): Likewise.
11269 (*vsx_le_perm_load_v16qi): Likewise.
11270 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11271 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11272 (*vsx_le_perm_store_v8hi): Likewise.
11273 (*vsx_le_perm_store_v16qi): Likewise.
11274 (eight unnamed splitters): Likewise.
11276 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11278 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11279 * config/rs6000/emmintrin.h: Likewise.
11280 * config/rs6000/mmintrin.h: Likewise.
11281 * config/rs6000/xmmintrin.h: Likewise.
11283 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11286 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11288 * tree.c (tree_nop_conversion): Return true for location wrapper
11290 (maybe_wrap_with_location): New function.
11291 (selftest::check_strip_nops): New function.
11292 (selftest::test_location_wrappers): New function.
11293 (selftest::tree_c_tests): Call it.
11294 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11295 (maybe_wrap_with_location): New decl.
11296 (EXPR_LOCATION_WRAPPER_P): New macro.
11297 (location_wrapper_p): New inline function.
11298 (tree_strip_any_location_wrapper): New inline function.
11300 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11303 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11304 stack_realign_offset for the largest alignment of stack slot
11306 (ix86_find_max_used_stack_alignment): New function.
11307 (ix86_finalize_stack_frame_flags): Use it. Set
11308 max_used_stack_alignment if we don't realign stack.
11309 * config/i386/i386.h (machine_function): Add
11310 max_used_stack_alignment.
11312 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11314 * config/arm/arm.opt (-mbranch-cost): New option.
11315 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11318 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11321 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11322 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11324 2018-01-10 Richard Biener <rguenther@suse.de>
11327 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11328 early out so it also covers the case where we have a non-NULL
11331 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11333 PR tree-optimization/83753
11334 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11335 for non-strided grouped accesses if the number of elements is 1.
11337 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11340 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11341 * i386.h (TARGET_USE_GATHER): Define.
11342 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11344 2018-01-10 Martin Liska <mliska@suse.cz>
11347 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11348 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11350 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11351 CLEANUP_NO_PARTITIONING is not set.
11353 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11355 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11356 for vectors, as a partial revert of r254296.
11357 * rtl.h (const_vec_p): Delete.
11358 (const_vec_duplicate_p): Don't test for vector CONSTs.
11359 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11360 * expmed.c (make_tree): Likewise.
11363 * common.md (E, F): Use CONSTANT_P instead of checking for
11365 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11366 checking for CONST_VECTOR.
11368 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11370 PR middle-end/83575
11371 * predict.c (force_edge_cold): Handle in more sane way edges
11372 with no prediction.
11374 2018-01-09 Carl Love <cel@us.ibm.com>
11376 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11378 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11379 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11380 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11381 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11382 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11383 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11384 * config/rs6000/rs6000-protos.h: Add extern defition for
11385 rs6000_generate_float2_double_code.
11386 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11388 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11389 (float2_v2df): Add define_expand.
11391 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11394 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11395 op_mode in the force_to_mode call.
11397 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11399 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11400 instead of checking each element individually.
11401 (aarch64_evpc_uzp): Likewise.
11402 (aarch64_evpc_zip): Likewise.
11403 (aarch64_evpc_ext): Likewise.
11404 (aarch64_evpc_rev): Likewise.
11405 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11406 instead of checking each element individually. Return true without
11408 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11409 whether all selected elements come from the same input, instead of
11410 checking each element individually. Remove calls to gen_rtx_REG,
11411 start_sequence and end_sequence and instead assert that no rtl is
11414 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11416 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11417 order of HIGH and CONST checks.
11419 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11421 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11422 if the destination isn't an SSA_NAME.
11424 2018-01-09 Richard Biener <rguenther@suse.de>
11426 PR tree-optimization/83668
11427 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11429 (canonicalize_loop_form): ... here, renamed from ...
11430 (canonicalize_loop_closed_ssa_form): ... this and amended to
11431 swap successor edges for loop exit blocks to make us use
11432 the RPO order we need for initial schedule generation.
11434 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11436 PR tree-optimization/64811
11437 * match.pd: When optimizing comparisons with Inf, avoid
11438 introducing or losing exceptions from comparisons with NaN.
11440 2018-01-09 Martin Liska <mliska@suse.cz>
11443 * asan.c (shadow_mem_size): Add gcc_assert.
11445 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11447 Don't save registers in main().
11450 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11451 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11452 * config/avr/avr.c (avr_set_current_function): Don't error if
11453 naked, OS_task or OS_main are specified at the same time.
11454 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11456 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11458 * common/config/avr/avr-common.c (avr_option_optimization_table):
11459 Switch on -mmain-is-OS_task for optimizing compilations.
11461 2018-01-09 Richard Biener <rguenther@suse.de>
11463 PR tree-optimization/83572
11464 * graphite.c: Include cfganal.h.
11465 (graphite_transform_loops): Connect infinite loops to exit
11466 and remove fake edges at the end.
11468 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11470 * ipa-inline.c (edge_badness): Revert accidental checkin.
11472 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11475 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11476 symbols; not inline clones.
11478 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11481 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11482 hard registers. Formatting fixes.
11484 PR preprocessor/83722
11485 * gcc.c (try_generate_repro): Pass
11486 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11487 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11490 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11491 Kito Cheng <kito.cheng@gmail.com>
11493 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11494 (riscv_leaf_function_p): Delete.
11495 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11497 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11499 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11501 (do_ifelse): New function.
11502 (do_isel): New function.
11503 (do_sub3): New function.
11504 (do_add3): New function.
11505 (do_load_mask_compare): New function.
11506 (do_overlap_load_compare): New function.
11507 (expand_compare_loop): New function.
11508 (expand_block_compare): Call expand_compare_loop() when appropriate.
11509 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11510 option description.
11511 (-mblock-compare-inline-loop-limit): New option.
11513 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11516 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11517 Reverse order of second and third operands in first alternative.
11518 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11519 of first and second elements in UNSPEC_VPERMR vector.
11520 (altivec_expand_vec_perm_le): Likewise.
11522 2018-01-08 Jeff Law <law@redhat.com>
11524 PR rtl-optimizatin/81308
11525 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11526 (process_switch): If group_case_labels makes a change, then set
11528 (pass_convert_switch::execute): If a switch is converted, then
11529 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11531 PR rtl-optimization/81308
11532 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11535 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11537 PR target/83663 - Revert r255946
11538 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11539 generation for cases where splatting a value is not useful.
11540 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11541 across a vec_duplicate and a paradoxical subreg forming a vector
11542 mode to a vec_concat.
11544 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11546 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11547 -march=armv8.3-a variants.
11548 * config/arm/t-multilib: Likewise.
11549 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11551 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11553 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11555 (cceq_ior_compare_complement): Give it a name so I can use it, and
11556 change boolean_or_operator predicate to boolean_operator so it can
11557 be used to generate a crand.
11558 (eqne): New code iterator.
11559 (bd/bd_neg): New code_attrs.
11560 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11561 a single define_insn.
11562 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11563 decrement (bdnzt/bdnzf/bdzt/bdzf).
11564 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11565 with the new names of the branch decrement patterns, and added the
11566 names of the branch decrement conditional patterns.
11568 2018-01-08 Richard Biener <rguenther@suse.de>
11570 PR tree-optimization/83563
11571 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11574 2018-01-08 Richard Biener <rguenther@suse.de>
11576 PR middle-end/83713
11577 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11579 2018-01-08 Richard Biener <rguenther@suse.de>
11581 PR tree-optimization/83685
11582 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11583 references to abnormals.
11585 2018-01-08 Richard Biener <rguenther@suse.de>
11588 * dwarf2out.c (output_indirect_strings): Handle empty
11589 skeleton_debug_str_hash.
11590 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11592 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11594 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11595 (emit_store_direct): Likewise.
11596 (arc_trampoline_adjust_address): Likewise.
11597 (arc_asm_trampoline_template): New function.
11598 (arc_initialize_trampoline): Use asm_trampoline_template.
11599 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11600 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11601 * config/arc/arc.md (flush_icache): Delete pattern.
11603 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11605 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11606 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11609 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11612 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11613 by not USED_FOR_TARGET.
11614 (make_pass_resolve_sw_modes): Likewise.
11616 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11618 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11621 2018-01-08 Richard Biener <rguenther@suse.de>
11623 PR middle-end/83580
11624 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11626 2018-01-08 Richard Biener <rguenther@suse.de>
11628 PR middle-end/83517
11629 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11631 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11633 PR middle-end/81897
11634 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11635 basic blocks with a small number of successors.
11636 (convert_control_dep_chain_into_preds): Improve handling of
11638 (dump_predicates): Split apart into...
11639 (dump_pred_chain): ...here...
11640 (dump_pred_info): ...and here.
11641 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11642 (can_chain_union_be_invalidated_p): Improve check for invalidation
11644 (uninit_uses_cannot_happen): Avoid unnecessary if
11645 convert_control_dep_chain_into_preds yielded nothing.
11647 2018-01-06 Martin Sebor <msebor@redhat.com>
11649 PR tree-optimization/83640
11650 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11651 subtracting negative offset from size.
11652 (builtin_access::overlap): Adjust offset bounds of the access to fall
11653 within the size of the object if possible.
11655 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11657 PR rtl-optimization/83699
11658 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11659 extract_bit_field_as_subreg to cases in which the extracted
11660 value is also a vector.
11662 * lra-constraints.c (process_alt_operands): Test for the equivalence
11663 substitutions when detecting a possible reload cycle.
11665 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11668 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11669 by default if flag_selective_schedling{,2}. Formatting fixes.
11671 PR rtl-optimization/83682
11672 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11673 if it has non-VECTOR_MODE element mode.
11674 (vec_duplicate_p): Likewise.
11676 PR middle-end/83694
11677 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11678 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11680 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11683 * config/i386/i386-builtin.def
11684 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11685 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11686 Require also OPTION_MASK_ISA_AVX512F in addition to
11687 OPTION_MASK_ISA_GFNI.
11688 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11689 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11690 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11691 to OPTION_MASK_ISA_GFNI.
11692 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11693 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11694 OPTION_MASK_ISA_AVX512BW.
11695 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11696 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11697 addition to OPTION_MASK_ISA_GFNI.
11698 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11699 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11700 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11701 to OPTION_MASK_ISA_GFNI.
11702 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11703 a requirement for all ISAs rather than any of them with a few
11705 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11707 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11708 bitmasks to be enabled with 3 exceptions, instead of requiring any
11709 enabled ISA with lots of exceptions.
11710 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11711 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11712 Change avx512bw in isa attribute to avx512f.
11713 * config/i386/sgxintrin.h: Add license boilerplate.
11714 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11715 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11716 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11717 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11719 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11720 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11721 temporarily sse2 rather than sse if not enabled already.
11724 * config/i386/sse.md (VI248_VLBW): Rename to ...
11725 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11726 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11727 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11728 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11729 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11730 mode iterator instead of VI248_VLBW.
11732 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11734 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11735 (record_modified): Skip clobbers; add debug output.
11736 (param_change_prob): Use sreal frequencies.
11738 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11740 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11741 punt for user-aligned variables.
11743 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11745 * tree-chrec.c (chrec_contains_symbols): Return true for
11748 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11751 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11752 of (x|y) == x for BICS pattern.
11754 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11756 PR tree-optimization/83605
11757 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11758 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11761 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11763 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11764 * config/epiphany/rtems.h: New file.
11766 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11767 Uros Bizjak <ubizjak@gmail.com>
11770 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11771 QIreg_operand instead of register_operand predicate.
11772 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11773 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11774 comments instead of -fmitigate[-_]rop.
11776 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11779 * cgraphunit.c (symbol_table::compile): Switch to text_section
11780 before calling assembly_start debug hook.
11781 * run-rtl-passes.c (run_rtl_passes): Likewise.
11784 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11786 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11787 range_int_cst_p rather than !symbolic_range_p before calling
11788 extract_range_from_multiplicative_op_1.
11790 2018-01-04 Jeff Law <law@redhat.com>
11792 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11793 redundant test in assertion.
11795 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11797 * doc/rtl.texi: Document machine_mode wrapper classes.
11799 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11801 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11802 using tree_to_uhwi.
11804 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11806 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11807 the VEC_PERM_EXPR fold to fail.
11809 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11812 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11813 to switched_sections.
11815 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11818 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11819 test for d.testing.
11821 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11824 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11825 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11827 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11830 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11831 is BLKmode and bitpos not zero or mode change is needed.
11833 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11836 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11839 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11842 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11843 instead of MULT rtx. Update all corresponding splitters.
11844 (*saddl_se): Ditto.
11845 (*ssub<modesuffix>): Ditto.
11846 (*ssubl_se): Ditto.
11847 (*cmp_sadd_di): Update split patterns.
11848 (*cmp_sadd_si): Ditto.
11849 (*cmp_sadd_sidi): Ditto.
11850 (*cmp_ssub_di): Ditto.
11851 (*cmp_ssub_si): Ditto.
11852 (*cmp_ssub_sidi): Ditto.
11853 * config/alpha/predicates.md (const23_operand): New predicate.
11854 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11855 Look for ASHIFT, not MULT inner operand.
11856 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11858 2018-01-04 Martin Liska <mliska@suse.cz>
11860 PR gcov-profile/83669
11861 * gcov.c (output_intermediate_file): Add version to intermediate
11863 * doc/gcov.texi: Document new field 'version' in intermediate
11864 file format. Fix location of '-k' option of gcov command.
11866 2018-01-04 Martin Liska <mliska@suse.cz>
11869 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11871 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11873 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11875 2018-01-03 Martin Sebor <msebor@redhat.com>
11877 PR tree-optimization/83655
11878 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11879 checking calls with invalid arguments.
11881 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11883 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11884 (vectorizable_mask_load_store): Delete.
11885 (vectorizable_call): Return false for masked loads and stores.
11886 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11887 instead of gimple_assign_rhs1.
11888 (vectorizable_load): Handle IFN_MASK_LOAD.
11889 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11893 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11895 (vectorizable_mask_load_store): ...here.
11896 (vectorizable_load): ...and here.
11898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11900 * tree-vect-stmts.c (vect_build_all_ones_mask)
11901 (vect_build_zero_merge_argument): New functions, split out from...
11902 (vectorizable_load): ...here.
11904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11906 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11908 (vectorizable_mask_load_store): ...here.
11909 (vectorizable_store): ...and here.
11911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11913 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11915 (vectorizable_mask_load_store): ...here.
11917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11919 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11920 (vect_model_store_cost): Take a vec_load_store_type instead of a
11922 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11923 (vect_model_store_cost): Take a vec_load_store_type instead of a
11925 (vectorizable_mask_load_store): Update accordingly.
11926 (vectorizable_store): Likewise.
11927 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11931 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11932 IFN_MASK_LOAD calls here rather than...
11933 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11935 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11936 Alan Hayward <alan.hayward@arm.com>
11937 David Sherwood <david.sherwood@arm.com>
11939 * expmed.c (extract_bit_field_1): For vector extracts,
11940 fall back to extract_bit_field_as_subreg if vec_extract
11943 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11944 Alan Hayward <alan.hayward@arm.com>
11945 David Sherwood <david.sherwood@arm.com>
11947 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11948 they are variable or constant sized.
11949 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11950 slots for constant-sized data.
11952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11953 Alan Hayward <alan.hayward@arm.com>
11954 David Sherwood <david.sherwood@arm.com>
11956 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11957 handling COND_EXPRs with boolean comparisons, try to find a better
11958 basis for the mask type than the boolean itself.
11960 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11962 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11963 is calculated and how it can be overridden.
11964 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11965 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11967 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11971 Alan Hayward <alan.hayward@arm.com>
11972 David Sherwood <david.sherwood@arm.com>
11974 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11975 Remove the mode argument.
11976 (aarch64_simd_valid_immediate): Remove the mode and inverse
11978 * config/aarch64/iterators.md (bitsize): New iterator.
11979 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11980 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11981 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11982 aarch64_simd_valid_immediate.
11983 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11984 (aarch64_reg_or_bic_imm): Likewise.
11985 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11986 with an insn_type enum and msl with a modifier_type enum.
11987 Replace element_width with a scalar_mode. Change the shift
11988 to unsigned int. Add constructors for scalar_float_mode and
11989 scalar_int_mode elements.
11990 (aarch64_vect_float_const_representable_p): Delete.
11991 (aarch64_can_const_movi_rtx_p)
11992 (aarch64_simd_scalar_immediate_valid_for_move)
11993 (aarch64_simd_make_constant): Update call to
11994 aarch64_simd_valid_immediate.
11995 (aarch64_advsimd_valid_immediate_hs): New function.
11996 (aarch64_advsimd_valid_immediate): Likewise.
11997 (aarch64_simd_valid_immediate): Remove mode and inverse
11998 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11999 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12000 and aarch64_float_const_representable_p on the result.
12001 (aarch64_output_simd_mov_immediate): Remove mode argument.
12002 Update call to aarch64_simd_valid_immediate and use of
12003 simd_immediate_info.
12004 (aarch64_output_scalar_simd_mov_immediate): Update call
12007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12008 Alan Hayward <alan.hayward@arm.com>
12009 David Sherwood <david.sherwood@arm.com>
12011 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12012 (mode_nunits): Likewise CONST_MODE_NUNITS.
12013 * machmode.def (ADJUST_NUNITS): Document.
12014 * genmodes.c (mode_data::need_nunits_adj): New field.
12015 (blank_mode): Update accordingly.
12016 (adj_nunits): New variable.
12017 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12019 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12020 listed in adj_nunits.
12021 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12022 listed in adj_nunits. Don't emit case statements for such modes.
12023 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12024 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12025 nothing if adj_nunits is nonnull.
12026 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12027 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12028 (emit_mode_fbit): Update use of print_maybe_const_decl.
12029 (emit_move_size): Likewise. Treat the array as non-const
12031 (emit_mode_adjustments): Handle adj_nunits.
12033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12035 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12036 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12037 (VECTOR_MODES): Use it.
12038 (make_vector_modes): Take the prefix as an argument.
12040 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12041 Alan Hayward <alan.hayward@arm.com>
12042 David Sherwood <david.sherwood@arm.com>
12044 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12045 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12046 for MODE_VECTOR_BOOL.
12047 * machmode.def (VECTOR_BOOL_MODE): Document.
12048 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12049 (make_vector_bool_mode): New function.
12050 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12052 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12053 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12055 * stor-layout.c (int_mode_for_mode): Likewise.
12056 * tree.c (build_vector_type_for_mode): Likewise.
12057 * varasm.c (output_constant_pool_2): Likewise.
12058 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12059 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12060 for MODE_VECTOR_BOOL.
12061 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12062 of mode class checks.
12063 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12064 instead of a list of mode class checks.
12065 (expand_vector_scalar_condition): Likewise.
12066 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12068 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12069 Alan Hayward <alan.hayward@arm.com>
12070 David Sherwood <david.sherwood@arm.com>
12072 * machmode.h (mode_size): Change from unsigned short to
12074 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12075 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12076 or if measurement_type is not polynomial.
12077 (fixed_size_mode::includes_p): Check for constant-sized modes.
12078 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12079 return a poly_uint16 rather than an unsigned short.
12080 (emit_mode_size): Change the type of mode_size from unsigned short
12081 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12082 (emit_mode_adjustments): Cope with polynomial vector sizes.
12083 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12085 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12087 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12088 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12089 * caller-save.c (setup_save_areas): Likewise.
12090 (replace_reg_with_saved_mem): Likewise.
12091 * calls.c (emit_library_call_value_1): Likewise.
12092 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12093 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12094 (gen_lowpart_for_combine): Likewise.
12095 * convert.c (convert_to_integer_1): Likewise.
12096 * cse.c (equiv_constant, cse_insn): Likewise.
12097 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12098 (cselib_subst_to_values): Likewise.
12099 * dce.c (word_dce_process_block): Likewise.
12100 * df-problems.c (df_word_lr_mark_ref): Likewise.
12101 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12102 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12103 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12104 (rtl_for_decl_location): Likewise.
12105 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12106 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12107 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12108 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12109 (expand_expr_real_1): Likewise.
12110 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12111 (pad_below): Likewise.
12112 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12113 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12114 * ira.c (get_subreg_tracking_sizes): Likewise.
12115 * ira-build.c (ira_create_allocno_objects): Likewise.
12116 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12117 (ira_sort_regnos_for_alter_reg): Likewise.
12118 * ira-costs.c (record_operand_costs): Likewise.
12119 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12120 (resolve_simple_move): Likewise.
12121 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12122 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12123 (lra_constraints): Likewise.
12124 (CONST_POOL_OK_P): Reject variable-sized modes.
12125 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12126 (add_pseudo_to_slot, lra_spill): Likewise.
12127 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12128 * optabs-query.c (get_best_extraction_insn): Likewise.
12129 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12130 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12131 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12132 * recog.c (offsettable_address_addr_space_p): Likewise.
12133 * regcprop.c (maybe_mode_change): Likewise.
12134 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12135 * regrename.c (build_def_use): Likewise.
12136 * regstat.c (dump_reg_info): Likewise.
12137 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12138 (find_reloads, find_reloads_subreg_address): Likewise.
12139 * reload1.c (eliminate_regs_1): Likewise.
12140 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12141 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12142 (simplify_binary_operation_1, simplify_subreg): Likewise.
12143 * targhooks.c (default_function_arg_padding): Likewise.
12144 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12145 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12146 (verify_gimple_assign_ternary): Likewise.
12147 * tree-inline.c (estimate_move_cost): Likewise.
12148 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12149 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12150 (get_address_cost_ainc): Likewise.
12151 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12152 (vect_supportable_dr_alignment): Likewise.
12153 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12154 (vectorizable_reduction): Likewise.
12155 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12156 (vectorizable_operation, vectorizable_load): Likewise.
12157 * tree.c (build_same_sized_truth_vector_type): Likewise.
12158 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12159 * var-tracking.c (emit_note_insn_var_location): Likewise.
12160 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12161 (ADDR_VEC_ALIGN): Likewise.
12163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12164 Alan Hayward <alan.hayward@arm.com>
12165 David Sherwood <david.sherwood@arm.com>
12167 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12169 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12170 or if measurement_type is polynomial.
12171 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12172 * combine.c (make_extraction): Likewise.
12173 * dse.c (find_shift_sequence): Likewise.
12174 * dwarf2out.c (mem_loc_descriptor): Likewise.
12175 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12176 (extract_bit_field, extract_low_bits): Likewise.
12177 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12178 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12179 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12180 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12181 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12182 * reload.c (find_reloads): Likewise.
12183 * reload1.c (alter_reg): Likewise.
12184 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12185 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12186 * tree-if-conv.c (predicate_mem_writes): Likewise.
12187 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12188 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12189 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12190 * valtrack.c (dead_debug_insert_temp): Likewise.
12191 * varasm.c (mergeable_constant_section): Likewise.
12192 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12195 Alan Hayward <alan.hayward@arm.com>
12196 David Sherwood <david.sherwood@arm.com>
12198 * expr.c (expand_assignment): Cope with polynomial mode sizes
12199 when assigning to a CONCAT.
12201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12202 Alan Hayward <alan.hayward@arm.com>
12203 David Sherwood <david.sherwood@arm.com>
12205 * machmode.h (mode_precision): Change from unsigned short to
12207 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12209 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12210 or if measurement_type is not polynomial.
12211 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12212 in which the mode is already known to be a scalar_int_mode.
12213 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12214 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12216 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12217 for GET_MODE_PRECISION.
12218 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12219 for GET_MODE_PRECISION.
12220 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12222 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12223 (expand_field_assignment, make_extraction): Likewise.
12224 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12225 (get_last_value): Likewise.
12226 * convert.c (convert_to_integer_1): Likewise.
12227 * cse.c (cse_insn): Likewise.
12228 * expr.c (expand_expr_real_1): Likewise.
12229 * lra-constraints.c (simplify_operand_subreg): Likewise.
12230 * optabs-query.c (can_atomic_load_p): Likewise.
12231 * optabs.c (expand_atomic_load): Likewise.
12232 (expand_atomic_store): Likewise.
12233 * ree.c (combine_reaching_defs): Likewise.
12234 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12235 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12236 * tree.h (type_has_mode_precision_p): Likewise.
12237 * ubsan.c (instrument_si_overflow): Likewise.
12239 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12240 Alan Hayward <alan.hayward@arm.com>
12241 David Sherwood <david.sherwood@arm.com>
12243 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12244 polynomial numbers of units.
12245 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12246 (valid_vector_subparts_p): New function.
12247 (build_vector_type): Remove temporary shim and take the number
12248 of units as a poly_uint64 rather than an int.
12249 (build_opaque_vector_type): Take the number of units as a
12250 poly_uint64 rather than an int.
12251 * tree.c (build_vector_from_ctor): Handle polynomial
12252 TYPE_VECTOR_SUBPARTS.
12253 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12254 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12255 (build_vector_from_val): If the number of units is variable,
12256 use build_vec_duplicate_cst for constant operands and
12257 VEC_DUPLICATE_EXPR otherwise.
12258 (make_vector_type): Remove temporary is_constant ().
12259 (build_vector_type, build_opaque_vector_type): Take the number of
12260 units as a poly_uint64 rather than an int.
12261 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12263 * cfgexpand.c (expand_debug_expr): Likewise.
12264 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12265 (store_constructor, expand_expr_real_1): Likewise.
12266 (const_scalar_mask_from_tree): Likewise.
12267 * fold-const-call.c (fold_const_reduction): Likewise.
12268 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12269 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12270 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12271 (fold_relational_const): Likewise.
12272 (native_interpret_vector): Likewise. Change the size from an
12273 int to an unsigned int.
12274 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12275 TYPE_VECTOR_SUBPARTS.
12276 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12277 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12278 duplicating a non-constant operand into a variable-length vector.
12279 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12280 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12281 * ipa-icf.c (sem_variable::equals): Likewise.
12282 * match.pd: Likewise.
12283 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12284 * print-tree.c (print_node): Likewise.
12285 * stor-layout.c (layout_type): Likewise.
12286 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12287 * tree-cfg.c (verify_gimple_comparison): Likewise.
12288 (verify_gimple_assign_binary): Likewise.
12289 (verify_gimple_assign_ternary): Likewise.
12290 (verify_gimple_assign_single): Likewise.
12291 * tree-pretty-print.c (dump_generic_node): Likewise.
12292 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12293 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12294 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12295 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12296 (vect_shift_permute_load_chain): Likewise.
12297 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12298 (expand_vector_condition, optimize_vector_constructor): Likewise.
12299 (lower_vec_perm, get_compute_type): Likewise.
12300 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12301 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12302 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12303 (vect_recog_mask_conversion_pattern): Likewise.
12304 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12305 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12306 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12307 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12308 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12309 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12310 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12311 (supportable_widening_operation): Likewise.
12312 (supportable_narrowing_operation): Likewise.
12313 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12315 * varasm.c (output_constant): Likewise.
12317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12318 Alan Hayward <alan.hayward@arm.com>
12319 David Sherwood <david.sherwood@arm.com>
12321 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12322 so that both the length == 3 and length != 3 cases set up their
12323 own permute vectors. Add comments explaining why we know the
12324 number of elements is constant.
12325 (vect_permute_load_chain): Likewise.
12327 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12328 Alan Hayward <alan.hayward@arm.com>
12329 David Sherwood <david.sherwood@arm.com>
12331 * machmode.h (mode_nunits): Change from unsigned char to
12333 (ONLY_FIXED_SIZE_MODES): New macro.
12334 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12335 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12336 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12338 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12339 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12340 or if measurement_type is not polynomial.
12341 * genmodes.c (ZERO_COEFFS): New macro.
12342 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12344 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12345 Use ZERO_COEFFS when emitting initializers.
12346 * data-streamer.h (bp_pack_poly_value): New function.
12347 (bp_unpack_poly_value): Likewise.
12348 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12349 for GET_MODE_NUNITS.
12350 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12351 for GET_MODE_NUNITS.
12352 * tree.c (make_vector_type): Remove temporary shim and make
12353 the real function take the number of units as a poly_uint64
12354 rather than an int.
12355 (build_vector_type_for_mode): Handle polynomial nunits.
12356 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12357 * emit-rtl.c (const_vec_series_p_1): Likewise.
12358 (gen_rtx_CONST_VECTOR): Likewise.
12359 * fold-const.c (test_vec_duplicate_folding): Likewise.
12360 * genrecog.c (validate_pattern): Likewise.
12361 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12362 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12363 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12364 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12365 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12366 * rtlanal.c (subreg_get_info): Likewise.
12367 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12368 (vect_grouped_load_supported): Likewise.
12369 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12370 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12371 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12372 (simplify_const_unary_operation, simplify_binary_operation_1)
12373 (simplify_const_binary_operation, simplify_ternary_operation)
12374 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12375 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12376 instead of CONST_VECTOR_NUNITS.
12377 * varasm.c (output_constant_pool_2): Likewise.
12378 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12379 explicit-encoded elements in the XVEC for variable-length vectors.
12381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12383 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12386 Alan Hayward <alan.hayward@arm.com>
12387 David Sherwood <david.sherwood@arm.com>
12389 * coretypes.h (fixed_size_mode): Declare.
12390 (fixed_size_mode_pod): New typedef.
12391 * builtins.h (target_builtins::x_apply_args_mode)
12392 (target_builtins::x_apply_result_mode): Change type to
12393 fixed_size_mode_pod.
12394 * builtins.c (apply_args_size, apply_result_size, result_vector)
12395 (expand_builtin_apply_args_1, expand_builtin_apply)
12396 (expand_builtin_return): Update accordingly.
12398 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12400 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12401 * cselib.c (cselib_hash_rtx): Likewise.
12402 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12403 CONST_VECTOR encoding.
12405 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12406 Jeff Law <law@redhat.com>
12409 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12410 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12411 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12412 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12415 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12416 explicitly probe *sp in a noreturn function if there were any callee
12417 register saves or frame pointer is needed.
12419 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12422 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12423 BLKmode for ternary, binary or unary expressions.
12426 * var-tracking.c (delete_vta_debug_insn): New inline function.
12427 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12428 insns from get_insns () to NULL instead of each bb separately.
12429 Use delete_vta_debug_insn. No longer static.
12430 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12431 delete_vta_debug_insns callers.
12432 * rtl.h (delete_vta_debug_insns): Declare.
12433 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12434 instead of variable_tracking_main.
12436 2018-01-03 Martin Sebor <msebor@redhat.com>
12438 PR tree-optimization/83603
12439 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12440 arguments past the endof the argument list in functions declared
12441 without a prototype.
12442 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12443 Avoid checking when arguments are null.
12445 2018-01-03 Martin Sebor <msebor@redhat.com>
12448 * doc/extend.texi (attribute const): Fix a typo.
12449 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12450 issuing -Wsuggest-attribute for void functions.
12452 2018-01-03 Martin Sebor <msebor@redhat.com>
12454 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12455 offset_int::from instead of wide_int::to_shwi.
12456 (maybe_diag_overlap): Remove assertion.
12457 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12458 * gimple-ssa-sprintf.c (format_directive): Same.
12459 (parse_directive): Same.
12460 (sprintf_dom_walker::compute_format_length): Same.
12461 (try_substitute_return_value): Same.
12463 2018-01-03 Jeff Law <law@redhat.com>
12465 PR middle-end/83654
12466 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12467 non-constant residual for zero at runtime and avoid probing in
12468 that case. Reorganize code for trailing problem to mirror handling
12471 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12473 PR tree-optimization/83501
12474 * tree-ssa-strlen.c (get_string_cst): New.
12475 (handle_char_store): Call get_string_cst.
12477 2018-01-03 Martin Liska <mliska@suse.cz>
12479 PR tree-optimization/83593
12480 * tree-ssa-strlen.c: Include tree-cfg.h.
12481 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12482 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12483 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12485 (strlen_dom_walker::before_dom_children): Call
12486 gimple_purge_dead_eh_edges. Dump tranformation with details
12488 (strlen_dom_walker::before_dom_children): Update call by adding
12489 new argument cleanup_eh.
12490 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12492 2018-01-03 Martin Liska <mliska@suse.cz>
12495 * cif-code.def (VARIADIC_THUNK): New enum value.
12496 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12499 2018-01-03 Jan Beulich <jbeulich@suse.com>
12501 * sse.md (mov<mode>_internal): Tighten condition for when to use
12502 vmovdqu<ssescalarsize> for TI and OI modes.
12504 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12506 Update copyright years.
12508 2018-01-03 Martin Liska <mliska@suse.cz>
12511 * ipa-visibility.c (function_and_variable_visibility): Skip
12512 functions with noipa attribure.
12514 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12516 * gcc.c (process_command): Update copyright notice dates.
12517 * gcov-dump.c (print_version): Ditto.
12518 * gcov.c (print_version): Ditto.
12519 * gcov-tool.c (print_version): Ditto.
12520 * gengtype.c (create_file): Ditto.
12521 * doc/cpp.texi: Bump @copying's copyright year.
12522 * doc/cppinternals.texi: Ditto.
12523 * doc/gcc.texi: Ditto.
12524 * doc/gccint.texi: Ditto.
12525 * doc/gcov.texi: Ditto.
12526 * doc/install.texi: Ditto.
12527 * doc/invoke.texi: Ditto.
12529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12531 * vector-builder.h (vector_builder::m_full_nelts): Change from
12532 unsigned int to poly_uint64.
12533 (vector_builder::full_nelts): Update prototype accordingly.
12534 (vector_builder::new_vector): Likewise.
12535 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12536 (vector_builder::operator ==): Likewise.
12537 (vector_builder::finalize): Likewise.
12538 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12539 Take the number of elements as a poly_uint64 rather than an
12541 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12542 from unsigned int to poly_uint64.
12543 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12544 (vec_perm_indices::new_vector): Likewise.
12545 (vec_perm_indices::length): Likewise.
12546 (vec_perm_indices::nelts_per_input): Likewise.
12547 (vec_perm_indices::input_nelts): Likewise.
12548 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12549 number of elements per input as a poly_uint64 rather than an
12550 unsigned int. Use the original encoding for variable-length
12551 vectors, rather than clamping each individual element.
12552 For the second and subsequent elements in each pattern,
12553 clamp the step and base before clamping their sum.
12554 (vec_perm_indices::series_p): Handle polynomial element counts.
12555 (vec_perm_indices::all_in_range_p): Likewise.
12556 (vec_perm_indices_to_tree): Likewise.
12557 (vec_perm_indices_to_rtx): Likewise.
12558 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12559 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12560 (tree_vector_builder::new_binary_operation): Handle polynomial
12561 element counts. Return false if we need to know the number
12562 of elements at compile time.
12563 * fold-const.c (fold_vec_perm): Punt if the number of elements
12564 isn't known at compile time.
12566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12568 * vec-perm-indices.h (vec_perm_builder): Change element type
12569 from HOST_WIDE_INT to poly_int64.
12570 (vec_perm_indices::element_type): Update accordingly.
12571 (vec_perm_indices::clamp): Handle polynomial element_types.
12572 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12573 (vec_perm_indices::all_in_range_p): Likewise.
12574 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12576 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12577 polynomial vec_perm_indices element types.
12578 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12579 * fold-const.c (fold_vec_perm): Likewise.
12580 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12581 * tree-vect-generic.c (lower_vec_perm): Likewise.
12582 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12583 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12584 element type to HOST_WIDE_INT.
12586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12587 Alan Hayward <alan.hayward@arm.com>
12588 David Sherwood <david.sherwood@arm.com>
12590 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12591 rather than an int. Use plus_constant.
12592 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12593 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12596 Alan Hayward <alan.hayward@arm.com>
12597 David Sherwood <david.sherwood@arm.com>
12599 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12600 a HOST_WIDE_INT to a poly_int64.
12602 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12603 Alan Hayward <alan.hayward@arm.com>
12604 David Sherwood <david.sherwood@arm.com>
12606 * calls.c (load_register_parameters): Cope with polynomial
12607 mode sizes. Require a constant size for BLKmode parameters
12608 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12609 forces a parameter to be padded at the lsb end in order to
12610 fill a complete number of words, require the parameter size
12611 to be ordered wrt UNITS_PER_WORD.
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 * reload1.c (spill_stack_slot_width): Change element type
12618 from unsigned int to poly_uint64_pod.
12619 (alter_reg): Treat mode sizes as polynomial.
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 * reload.c (complex_word_subreg_p): New function.
12626 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
12633 operands whose sizes aren't ordered.
12634 (match_reload): Refer to this check here.
12636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12637 Alan Hayward <alan.hayward@arm.com>
12638 David Sherwood <david.sherwood@arm.com>
12640 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12641 that the mode size is in the set {1, 2, 4, 8, 16}.
12643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12644 Alan Hayward <alan.hayward@arm.com>
12645 David Sherwood <david.sherwood@arm.com>
12647 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12648 Use plus_constant instead of gen_rtx_PLUS.
12650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12651 Alan Hayward <alan.hayward@arm.com>
12652 David Sherwood <david.sherwood@arm.com>
12654 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12655 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12656 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12657 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12658 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12659 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12660 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12661 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12662 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12663 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12665 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12666 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12667 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12668 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12669 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12670 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12671 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12672 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12673 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12674 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12676 * expr.c (emit_move_resolve_push): Treat the input and result
12677 of PUSH_ROUNDING as a poly_int64.
12678 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12679 (emit_push_insn): Likewise.
12680 * lra-eliminations.c (mark_not_eliminable): Likewise.
12681 * recog.c (push_operand): Likewise.
12682 * reload1.c (elimination_effects): Likewise.
12683 * rtlanal.c (nonzero_bits1): Likewise.
12684 * calls.c (store_one_arg): Likewise. Require the padding to be
12685 known at compile time.
12687 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12688 Alan Hayward <alan.hayward@arm.com>
12689 David Sherwood <david.sherwood@arm.com>
12691 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12692 Use plus_constant instead of gen_rtx_PLUS.
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 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12699 rather than an int.
12701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12702 Alan Hayward <alan.hayward@arm.com>
12703 David Sherwood <david.sherwood@arm.com>
12705 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12706 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12707 via stack temporaries. Treat the mode size as polynomial too.
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 * expr.c (expand_expr_real_2): When handling conversions involving
12714 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12715 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12716 as a poly_uint64 too.
12718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12719 Alan Hayward <alan.hayward@arm.com>
12720 David Sherwood <david.sherwood@arm.com>
12722 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12725 Alan Hayward <alan.hayward@arm.com>
12726 David Sherwood <david.sherwood@arm.com>
12728 * combine.c (can_change_dest_mode): Handle polynomial
12729 REGMODE_NATURAL_SIZE.
12730 * expmed.c (store_bit_field_1): Likewise.
12731 * expr.c (store_constructor): Likewise.
12732 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12733 and polynomial REGMODE_NATURAL_SIZE.
12734 (gen_lowpart_common): Likewise.
12735 * reginfo.c (record_subregs_of_mode): Likewise.
12736 * rtlanal.c (read_modify_subreg_p): Likewise.
12738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12739 Alan Hayward <alan.hayward@arm.com>
12740 David Sherwood <david.sherwood@arm.com>
12742 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12743 numbers of elements.
12745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12746 Alan Hayward <alan.hayward@arm.com>
12747 David Sherwood <david.sherwood@arm.com>
12749 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12756 in a POINTER_PLUS_EXPR.
12758 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12759 Alan Hayward <alan.hayward@arm.com>
12760 David Sherwood <david.sherwood@arm.com>
12762 * omp-simd-clone.c (simd_clone_subparts): New function.
12763 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12764 (ipa_simd_modify_function_body): Likewise.
12766 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12767 Alan Hayward <alan.hayward@arm.com>
12768 David Sherwood <david.sherwood@arm.com>
12770 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12771 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12772 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12773 (expand_vector_condition, vector_element): Likewise.
12774 (subparts_gt): New function.
12775 (get_compute_type): Use subparts_gt.
12776 (count_type_subparts): Delete.
12777 (expand_vector_operations_1): Use subparts_gt instead of
12778 count_type_subparts.
12780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12781 Alan Hayward <alan.hayward@arm.com>
12782 David Sherwood <david.sherwood@arm.com>
12784 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12785 (vect_compile_time_alias): ...this new function. Do the calculation
12786 on poly_ints rather than trees.
12787 (vect_prune_runtime_alias_test_list): Update call accordingly.
12789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12790 Alan Hayward <alan.hayward@arm.com>
12791 David Sherwood <david.sherwood@arm.com>
12793 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12795 (vect_schedule_slp_instance): Likewise.
12797 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12798 Alan Hayward <alan.hayward@arm.com>
12799 David Sherwood <david.sherwood@arm.com>
12801 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12802 constant and extern definitions for variable-length vectors.
12803 (vect_get_constant_vectors): Note that the number of units
12804 is known to be constant.
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_conversion): Treat the number
12811 of units as polynomial. Choose between WIDE and NARROW based
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 (simd_clone_subparts): New function.
12819 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12822 Alan Hayward <alan.hayward@arm.com>
12823 David Sherwood <david.sherwood@arm.com>
12825 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12826 vectors as polynomial. Use build_index_vector for
12827 IFN_GOMP_SIMD_LANE.
12829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12830 Alan Hayward <alan.hayward@arm.com>
12831 David Sherwood <david.sherwood@arm.com>
12833 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12834 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12835 for variable-length vectors.
12836 (vectorizable_mask_load_store): Treat the number of units as
12837 polynomial, asserting that it is constant if the condition has
12838 already been enforced.
12839 (vectorizable_store, vectorizable_load): Likewise.
12841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12842 Alan Hayward <alan.hayward@arm.com>
12843 David Sherwood <david.sherwood@arm.com>
12845 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12846 of units as polynomial. Punt if we can't tell at compile time
12847 which vector contains the final result.
12849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12850 Alan Hayward <alan.hayward@arm.com>
12851 David Sherwood <david.sherwood@arm.com>
12853 * tree-vect-loop.c (vectorizable_induction): Treat the number
12854 of units as polynomial. Punt on SLP inductions. Use an integer
12855 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12856 cast of such a series for variable-length floating-point
12859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12860 Alan Hayward <alan.hayward@arm.com>
12861 David Sherwood <david.sherwood@arm.com>
12863 * tree.h (build_index_vector): Declare.
12864 * tree.c (build_index_vector): New function.
12865 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12866 of units as polynomial, forcibly converting it to a constant if
12867 vectorizable_reduction has already enforced the condition.
12868 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12869 to create a {1,2,3,...} vector.
12870 (vectorizable_reduction): Treat the number of units as polynomial.
12871 Choose vectype_in based on the largest scalar element size rather
12872 than the smallest number of units. Enforce the restrictions
12875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12876 Alan Hayward <alan.hayward@arm.com>
12877 David Sherwood <david.sherwood@arm.com>
12879 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12880 number of units as polynomial.
12882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12883 Alan Hayward <alan.hayward@arm.com>
12884 David Sherwood <david.sherwood@arm.com>
12886 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12887 * target.def (autovectorize_vector_sizes): Return the vector sizes
12888 by pointer, using vector_sizes rather than a bitmask.
12889 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12890 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12891 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12893 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12894 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12895 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12896 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12897 * omp-general.c (omp_max_vf): Likewise.
12898 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12899 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12900 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12901 * tree-vect-slp.c (vect_slp_bb): Likewise.
12902 * doc/tm.texi: Regenerate.
12903 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12905 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12906 the vector size as a poly_uint64 rather than an unsigned int.
12907 (current_vector_size): Change from an unsigned int to a poly_uint64.
12908 (get_vectype_for_scalar_type): Update accordingly.
12909 * tree.h (build_truth_vector_type): Take the size and number of
12910 units as a poly_uint64 rather than an unsigned int.
12911 (build_vector_type): Add a temporary overload that takes
12912 the number of units as a poly_uint64 rather than an unsigned int.
12913 * tree.c (make_vector_type): Likewise.
12914 (build_truth_vector_type): Take the number of units as a poly_uint64
12915 rather than an unsigned int.
12917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12918 Alan Hayward <alan.hayward@arm.com>
12919 David Sherwood <david.sherwood@arm.com>
12921 * target.def (get_mask_mode): Take the number of units and length
12922 as poly_uint64s rather than unsigned ints.
12923 * targhooks.h (default_get_mask_mode): Update accordingly.
12924 * targhooks.c (default_get_mask_mode): Likewise.
12925 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12926 * doc/tm.texi: Regenerate.
12928 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12929 Alan Hayward <alan.hayward@arm.com>
12930 David Sherwood <david.sherwood@arm.com>
12932 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12933 * omp-general.c (omp_max_vf): Likewise.
12934 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12935 (expand_omp_simd): Handle polynomial safelen.
12936 * omp-low.c (omplow_simd_context): Add a default constructor.
12937 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12938 (lower_rec_simd_input_clauses): Update accordingly.
12939 (lower_rec_input_clauses): Likewise.
12941 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12942 Alan Hayward <alan.hayward@arm.com>
12943 David Sherwood <david.sherwood@arm.com>
12945 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12946 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12947 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12948 (vect_analyze_slp_cost): Likewise.
12949 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12950 (vect_model_load_cost): Likewise.
12952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12953 Alan Hayward <alan.hayward@arm.com>
12954 David Sherwood <david.sherwood@arm.com>
12956 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12957 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12958 from an unsigned int * to a poly_uint64_pod *.
12959 (calculate_unrolling_factor): New function.
12960 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12963 Alan Hayward <alan.hayward@arm.com>
12964 David Sherwood <david.sherwood@arm.com>
12966 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12967 from an unsigned int to a poly_uint64.
12968 (_loop_vec_info::slp_unrolling_factor): Likewise.
12969 (_loop_vec_info::vectorization_factor): Change from an int
12971 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12972 (vect_get_num_vectors): New function.
12973 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12974 (vect_get_num_copies): Use vect_get_num_vectors.
12975 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12976 to an unsigned int *.
12977 (vect_analyze_data_refs): Change min_vf from an int * to a
12979 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12980 than an unsigned HOST_WIDE_INT.
12981 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12982 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12983 to an unsigned int *.
12984 (vect_analyze_data_ref_dependences): Likewise.
12985 (vect_compute_data_ref_alignment): Handle polynomial vf.
12986 (vect_enhance_data_refs_alignment): Likewise.
12987 (vect_prune_runtime_alias_test_list): Likewise.
12988 (vect_shift_permute_load_chain): Likewise.
12989 (vect_supportable_dr_alignment): Likewise.
12990 (dependence_distance_ge_vf): Take the vectorization factor as a
12991 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12992 (vect_analyze_data_refs): Change min_vf from an int * to a
12994 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12995 vfm1 as a poly_uint64 rather than an int. Make the same change
12996 for the returned bound_scalar.
12997 (vect_gen_vector_loop_niters): Handle polynomial vf.
12998 (vect_do_peeling): Likewise. Update call to
12999 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13000 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13002 * tree-vect-loop.c (vect_determine_vectorization_factor)
13003 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13004 (vect_get_known_peeling_cost): Likewise.
13005 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13006 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13007 (vect_transform_loop): Likewise. Use the lowest possible VF when
13008 updating the upper bounds of the loop.
13009 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13010 rather than an int.
13011 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13012 polynomial unroll factors.
13013 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13014 (vect_make_slp_decision): Likewise.
13015 (vect_supported_load_permutation_p): Likewise, and polynomial
13017 (vect_analyze_slp_cost): Handle polynomial vf.
13018 (vect_slp_analyze_node_operations): Likewise.
13019 (vect_slp_analyze_bb_1): Likewise.
13020 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13021 than an unsigned HOST_WIDE_INT.
13022 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13023 (vectorizable_load): Handle polynomial vf.
13024 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13026 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13029 Alan Hayward <alan.hayward@arm.com>
13030 David Sherwood <david.sherwood@arm.com>
13032 * match.pd: Handle bit operations involving three constants
13033 and try to fold one pair.
13035 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13037 * tree-vect-loop-manip.c: Include gimple-fold.h.
13038 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13039 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13040 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13041 Add a path that uses a step of VF instead of 1, but disable it
13043 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13044 and niters_no_overflow parameters. Update calls to
13045 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13046 Create a new SSA name if the latter choses to use a ste other
13047 than zero, and return it via niters_vector_mult_vf_var.
13048 * tree-vect-loop.c (vect_transform_loop): Update calls to
13049 vect_do_peeling, vect_gen_vector_loop_niters and
13050 slpeel_make_loop_iterate_ntimes.
13051 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13052 (vect_gen_vector_loop_niters): Update declarations after above changes.
13054 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13056 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13057 128-bit round to integer instructions.
13058 (ceil<mode>2): Likewise.
13059 (btrunc<mode>2): Likewise.
13060 (round<mode>2): Likewise.
13062 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13064 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13065 unaligned VSX load/store on P8/P9.
13066 (expand_block_clear): Allow the use of unaligned VSX
13067 load/store on P8/P9.
13069 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13071 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13073 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13074 swap associated with both a load and a store.
13076 2018-01-02 Andrew Waterman <andrew@sifive.com>
13078 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13079 * config/riscv/riscv.md (clear_cache): Use it.
13081 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13083 * web.c: Remove out-of-date comment.
13085 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13087 * expr.c (fixup_args_size_notes): Check that any existing
13088 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13089 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13090 (emit_single_push_insn): ...here.
13092 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13094 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13095 (const_vector_encoded_nelts): New function.
13096 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13097 (const_vector_int_elt, const_vector_elt): Declare.
13098 * emit-rtl.c (const_vector_int_elt_1): New function.
13099 (const_vector_elt): Likewise.
13100 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13101 of CONST_VECTOR_ELT.
13103 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13105 * expr.c: Include rtx-vector-builder.h.
13106 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13107 directly on the tree encoding.
13108 (const_vector_from_tree): Likewise.
13109 * optabs.c: Include rtx-vector-builder.h.
13110 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13111 sequence of "u" values.
13112 * vec-perm-indices.c: Include rtx-vector-builder.h.
13113 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13114 directly on the vec_perm_indices encoding.
13116 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13118 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13119 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13120 * rtx-vector-builder.h: New file.
13121 * rtx-vector-builder.c: Likewise.
13122 * rtl.h (rtx_def::u2): Add a const_vector field.
13123 (CONST_VECTOR_NPATTERNS): New macro.
13124 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13125 (CONST_VECTOR_DUPLICATE_P): Likewise.
13126 (CONST_VECTOR_STEPPED_P): Likewise.
13127 (CONST_VECTOR_ENCODED_ELT): Likewise.
13128 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13129 (unwrap_const_vec_duplicate): Likewise.
13130 (const_vec_series_p): Check for a non-duplicated vector encoding.
13131 Say that the function only returns true for integer vectors.
13132 * emit-rtl.c: Include rtx-vector-builder.h.
13133 (gen_const_vec_duplicate_1): Delete.
13134 (gen_const_vector): Call gen_const_vec_duplicate instead of
13135 gen_const_vec_duplicate_1.
13136 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13137 (gen_const_vec_duplicate): Use rtx_vector_builder.
13138 (gen_const_vec_series): Likewise.
13139 (gen_rtx_CONST_VECTOR): Likewise.
13140 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13141 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13142 Build a new vector rather than modifying a CONST_VECTOR in-place.
13143 (handle_special_swappables): Update call accordingly.
13144 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13145 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13146 Build a new vector rather than modifying a CONST_VECTOR in-place.
13147 (handle_special_swappables): Update call accordingly.
13149 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13151 * simplify-rtx.c (simplify_const_binary_operation): Use
13152 CONST_VECTOR_ELT instead of XVECEXP.
13154 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13156 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13157 the selector elements to be different from the data elements
13158 if the selector is a VECTOR_CST.
13159 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13160 ssizetype for the selector.
13162 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13164 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13165 before testing each element individually.
13166 * tree-vect-generic.c (lower_vec_perm): Likewise.
13168 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13170 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13171 * selftest-run-tests.c (selftest::run_tests): Call it.
13172 * vector-builder.h (vector_builder::operator ==): New function.
13173 (vector_builder::operator !=): Likewise.
13174 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13175 (vec_perm_indices::all_from_input_p): New function.
13176 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13177 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13178 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13179 instead of reading the VECTOR_CST directly. Detect whether both
13180 vector inputs are the same before constructing the vec_perm_indices,
13181 and update the number of inputs argument accordingly. Use the
13182 utility functions added above. Only construct sel2 if we need to.
13184 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13186 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13187 the broadcast of the low byte.
13188 (expand_mult_highpart): Use an explicit encoding for the permutes.
13189 * optabs-query.c (can_mult_highpart_p): Likewise.
13190 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13191 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13192 (vectorizable_bswap): Likewise.
13193 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13194 explicit encoding for the power-of-2 permutes.
13195 (vect_permute_store_chain): Likewise.
13196 (vect_grouped_load_supported): Likewise.
13197 (vect_permute_load_chain): Likewise.
13199 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13201 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13202 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13203 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13204 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13205 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13206 (vect_gen_perm_mask_any): Likewise.
13208 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13210 * int-vector-builder.h: New file.
13211 * vec-perm-indices.h: Include int-vector-builder.h.
13212 (vec_perm_indices): Redefine as an int_vector_builder.
13213 (auto_vec_perm_indices): Delete.
13214 (vec_perm_builder): Redefine as a stand-alone class.
13215 (vec_perm_indices::vec_perm_indices): New function.
13216 (vec_perm_indices::clamp): Likewise.
13217 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13218 (vec_perm_indices::new_vector): New function.
13219 (vec_perm_indices::new_expanded_vector): Update for new
13220 vec_perm_indices class.
13221 (vec_perm_indices::rotate_inputs): New function.
13222 (vec_perm_indices::all_in_range_p): Operate directly on the
13223 encoded form, without computing elided elements.
13224 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13225 encoding. Update for new vec_perm_indices class.
13226 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13227 the given vec_perm_builder.
13228 (expand_vec_perm_var): Update vec_perm_builder constructor.
13229 (expand_mult_highpart): Use vec_perm_builder instead of
13230 auto_vec_perm_indices.
13231 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13232 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13233 or double series encoding as appropriate.
13234 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13235 vec_perm_indices instead of auto_vec_perm_indices.
13236 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13237 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13238 (vect_permute_store_chain): Likewise.
13239 (vect_grouped_load_supported): Likewise.
13240 (vect_permute_load_chain): Likewise.
13241 (vect_shift_permute_load_chain): Likewise.
13242 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13243 (vect_transform_slp_perm_load): Likewise.
13244 (vect_schedule_slp_instance): Likewise.
13245 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13246 (vectorizable_mask_load_store): Likewise.
13247 (vectorizable_bswap): Likewise.
13248 (vectorizable_store): Likewise.
13249 (vectorizable_load): Likewise.
13250 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13251 vec_perm_indices instead of auto_vec_perm_indices. Use
13252 tree_to_vec_perm_builder to read the vector from a tree.
13253 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13254 vec_perm_builder instead of a vec_perm_indices.
13255 (have_whole_vector_shift): Use vec_perm_builder and
13256 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13257 truncation to calc_vec_perm_mask_for_shift.
13258 (vect_create_epilog_for_reduction): Likewise.
13259 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13260 from auto_vec_perm_indices to vec_perm_indices.
13261 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13262 instead of changing individual elements.
13263 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13264 the vector in d.perm.
13265 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13266 from auto_vec_perm_indices to vec_perm_indices.
13267 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13268 instead of changing individual elements.
13269 (arm_vectorize_vec_perm_const): Use new_vector to install
13270 the vector in d.perm.
13271 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13272 Update vec_perm_builder constructor.
13273 (rs6000_expand_interleave): Likewise.
13274 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13275 (rs6000_expand_interleave): Likewise.
13277 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13279 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13280 to qimode could truncate the indices.
13281 * optabs.c (expand_vec_perm_var): Likewise.
13283 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13285 * Makefile.in (OBJS): Add vec-perm-indices.o.
13286 * vec-perm-indices.h: New file.
13287 * vec-perm-indices.c: Likewise.
13288 * target.h (vec_perm_indices): Replace with a forward class
13290 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13291 * optabs.h: Include vec-perm-indices.h.
13292 (expand_vec_perm): Delete.
13293 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13294 (expand_vec_perm_const): Declare.
13295 * target.def (vec_perm_const_ok): Replace with...
13296 (vec_perm_const): ...this new hook.
13297 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13298 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13299 * doc/tm.texi: Regenerate.
13300 * optabs.def (vec_perm_const): Delete.
13301 * doc/md.texi (vec_perm_const): Likewise.
13302 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13303 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13304 expand_vec_perm for constant permutation vectors. Assert that
13305 the mode of variable permutation vectors is the integer equivalent
13306 of the mode that is being permuted.
13307 * optabs-query.h (selector_fits_mode_p): Declare.
13308 * optabs-query.c: Include vec-perm-indices.h.
13309 (selector_fits_mode_p): New function.
13310 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13311 is defined, instead of checking whether the vec_perm_const_optab
13312 exists. Use targetm.vectorize.vec_perm_const instead of
13313 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13314 fit in the vector mode before using a variable permute.
13315 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13316 vec_perm_indices instead of an rtx.
13317 (expand_vec_perm): Replace with...
13318 (expand_vec_perm_const): ...this new function. Take the selector
13319 as a vec_perm_indices rather than an rtx. Also take the mode of
13320 the selector. Update call to shift_amt_for_vec_perm_mask.
13321 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13322 Use vec_perm_indices::new_expanded_vector to expand the original
13323 selector into bytes. Check whether the indices fit in the vector
13324 mode before using a variable permute.
13325 (expand_vec_perm_var): Make global.
13326 (expand_mult_highpart): Use expand_vec_perm_const.
13327 * fold-const.c: Includes vec-perm-indices.h.
13328 * tree-ssa-forwprop.c: Likewise.
13329 * tree-vect-data-refs.c: Likewise.
13330 * tree-vect-generic.c: Likewise.
13331 * tree-vect-loop.c: Likewise.
13332 * tree-vect-slp.c: Likewise.
13333 * tree-vect-stmts.c: Likewise.
13334 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13336 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13337 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13338 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13339 (aarch64_vectorize_vec_perm_const): ...this new function.
13340 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13341 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13342 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13343 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13344 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13345 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13346 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13348 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13349 check for NEON modes.
13350 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13351 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13352 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13353 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13355 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13356 the old VEC_PERM_CONST conditions.
13357 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13358 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13359 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13360 (ia64_vectorize_vec_perm_const_ok): Merge into...
13361 (ia64_vectorize_vec_perm_const): ...this new function.
13362 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13363 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13364 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13365 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13366 * config/mips/mips.c (mips_expand_vec_perm_const)
13367 (mips_vectorize_vec_perm_const_ok): Merge into...
13368 (mips_vectorize_vec_perm_const): ...this new function.
13369 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13370 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13371 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13372 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13373 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13374 (rs6000_expand_vec_perm_const): Delete.
13375 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13377 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13378 (altivec_expand_vec_perm_const_le): Take each operand individually.
13379 Operate on constant selectors rather than rtxes.
13380 (altivec_expand_vec_perm_const): Likewise. Update call to
13381 altivec_expand_vec_perm_const_le.
13382 (rs6000_expand_vec_perm_const): Delete.
13383 (rs6000_vectorize_vec_perm_const_ok): Delete.
13384 (rs6000_vectorize_vec_perm_const): New function.
13385 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13386 an element count and rtx array.
13387 (rs6000_expand_extract_even): Update call accordingly.
13388 (rs6000_expand_interleave): Likewise.
13389 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13390 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13391 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13392 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13393 (rs6000_expand_vec_perm_const): Delete.
13394 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13395 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13396 (altivec_expand_vec_perm_const_le): Take each operand individually.
13397 Operate on constant selectors rather than rtxes.
13398 (altivec_expand_vec_perm_const): Likewise. Update call to
13399 altivec_expand_vec_perm_const_le.
13400 (rs6000_expand_vec_perm_const): Delete.
13401 (rs6000_vectorize_vec_perm_const_ok): Delete.
13402 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13403 reference to the SPE evmerge intructions.
13404 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13405 an element count and rtx array.
13406 (rs6000_expand_extract_even): Update call accordingly.
13407 (rs6000_expand_interleave): Likewise.
13408 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13409 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13411 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13413 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13415 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13416 vector mode and that that mode matches the mode of the data
13418 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13419 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13420 directly using expand_vec_perm_1 when forcing selectors into
13422 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13424 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13426 * optabs-query.h (can_vec_perm_p): Delete.
13427 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13428 * optabs-query.c (can_vec_perm_p): Split into...
13429 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13430 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13431 particular selector is valid.
13432 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13433 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13434 (vect_grouped_load_supported): Likewise.
13435 (vect_shift_permute_load_chain): Likewise.
13436 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13437 (vect_transform_slp_perm_load): Likewise.
13438 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13439 (vectorizable_bswap): Likewise.
13440 (vect_gen_perm_mask_checked): Likewise.
13441 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13442 implementations of variable permutation vectors into account
13443 when deciding which selector to use.
13444 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13445 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13446 with a false third argument.
13447 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13448 to test whether the constant selector is valid and can_vec_perm_var_p
13449 to test whether a variable selector is valid.
13451 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13453 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13454 * optabs-query.c (can_vec_perm_p): Likewise.
13455 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13456 instead of vec_perm_indices.
13457 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13458 (vect_gen_perm_mask_checked): Likewise,
13459 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13460 (vect_gen_perm_mask_checked): Likewise,
13462 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13464 * optabs-query.h (qimode_for_vec_perm): Declare.
13465 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13466 (qimode_for_vec_perm): ...this new function.
13467 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13469 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13471 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13472 does not have a conditional at the top.
13474 2018-01-02 Richard Biener <rguenther@suse.de>
13476 * ipa-inline.c (big_speedup_p): Fix expression.
13478 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13481 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13484 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13488 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13489 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13490 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13491 cond_taken_branch_cost 3->4.
13493 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13495 PR tree-optimization/83581
13496 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13497 TODO_cleanup_cfg if any changes have been made.
13499 PR middle-end/83608
13500 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13501 convert_modes if target mode has the right side, but different mode
13504 PR middle-end/83609
13505 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13506 last argument when extracting from CONCAT. If either from_real or
13507 from_imag is NULL, use expansion through memory. If result is not
13508 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13509 the parts directly to inner mode, if even that fails, use expansion
13512 PR middle-end/83623
13513 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13514 check for bswap in mode rather than HImode and use that in expand_unop
13517 Copyright (C) 2018 Free Software Foundation, Inc.
13519 Copying and distribution of this file, with or without modification,
13520 are permitted in any medium without royalty provided the copyright
13521 notice and this notice are preserved.
projects / gcc.git / blob