1 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2 Chung-Ju Wu <jasonwucj@gmail.com>
4 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
5 * config/nds32/nds32-md-auxiliary.c
6 (symbolic_reference_mentioned_p): New.
7 (nds32_legitimize_ict_address): New.
8 (nds32_expand_ict_move): New.
9 (nds32_indirect_call_referenced_p): New.
10 (nds32_symbol_binds_local_p): Delete.
11 (nds32_long_call_p): Modify.
12 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
13 * config/nds32/nds32-protos.h
14 (symbolic_reference_mentioned_p): Declare.
15 (nds32_legitimize_ict_address): Declare.
16 (nds32_expand_ict_move): Declare.
17 (nds32_indirect_call_referenced_p): Declare.
18 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
19 (nds32_relax_group): Use nds32_ict_const_p as condition.
20 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
21 (nds32_asm_file_start): Output ict_model directive in asm code.
22 (nds32_legitimate_address_p): Consider indirect call.
23 (nds32_print_operand): Consider indirect call.
24 (nds32_print_operand_address): Consider indirect call.
25 (nds32_insert_attributes): Handle "indirect_call" attribute.
26 (TARGET_LEGITIMATE_ADDRESS_P): Define.
27 (TARGET_LEGITIMATE_CONSTANT_P): Define.
28 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
29 (TARGET_DELEGITIMIZE_ADDRESS): Define.
30 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
31 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
32 (TARGET_ICT_MODEL_SMALL): Define.
33 (TARGET_ICT_MODEL_LARGE): Define.
34 * config/nds32/nds32.md (movsi): Consider ict model.
35 (call, call_value): Consider ict model.
36 (sibcall, sibcall_value): Consider ict model.
37 * config/nds32/nds32.opt (mict-model): New option.
38 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
41 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
42 Monk Chiang <sh.chiang04@gmail.com>
43 Jim Wilson <jimw@sifive.com>
45 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
46 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
47 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
48 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
49 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
50 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
51 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
52 compute save_libcall_adjustment properly.
53 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
54 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
55 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
56 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
57 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
58 (ABI_SPEC): Handle mabi=ilp32e.
59 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
61 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
62 <-march>: Add rv32e as an example.
64 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
67 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
68 (intra_create_variable_infos): Handle C++ constructors.
70 2018-05-18 Martin Liska <mliska@suse.cz>
72 * passes.def: Remove a redundant pass.
74 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
77 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
79 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
81 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
83 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
85 (armv2, armv2a, armv3, armv3m): Delete architectures.
86 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
87 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
88 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
90 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
91 (*mulsidi3adddi): Likewise.
93 (*mulsidi3_nov6): Likewise.
94 (umulsidi3): Likewise.
95 (umulsidi3_nov6): Likewise.
96 (umaddsidi4): Likewise.
97 (*umulsidi3adddi): Likewise.
98 (smulsi3_highpart): Likewise.
99 (*smulsi3_highpart_nov6): Likewise.
100 (umulsi3_highpart): Likewise.
101 (*umulsi3_highpart_nov6): Likewise.
102 * config/arm/arm.h (arm_arch3m): Delete.
103 * config/arm/arm.c (arm_arch3m): Delete.
104 (arm_option_override_internal): Update armv3-related comment.
105 (arm_configure_build_target): Delete use of isa_bit_mode32.
106 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
107 (arm_rtx_costs_internal): Delete check of arm_arch3m.
108 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
111 * config/arm/arm-protos.h (arm_arch3m): Delete.
112 * config/arm/arm-tables.opt: Regenerate.
113 * config/arm/arm-tune.md: Likewise.
114 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
115 deleted architectures.
117 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
119 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
120 (armv5t, armv5te): New features.
121 (ARMv5, ARMv5e): Delete fgroups.
122 (ARMv5t, ARMv5te): Adjust for above changes.
124 (armv5, armv5e): Delete arches.
125 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
127 (*call_reg_arm): Likewise.
128 (*call_value_reg_armv5): Likewise.
129 (*call_value_reg_arm): Likewise.
130 (*call_symbol): Likewise.
131 (*call_value_symbol): Likewise.
132 (*sibcall_insn): Likewise.
133 (*sibcall_value_insn): Likewise.
135 (prefetch): Likewise.
136 (define_split and define_peephole2 dependent on arm_arch5):
138 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
140 (TARGET_ARM_QBIT): Likewise.
141 (TARGET_DSP_MULTIPLY): Likewise.
142 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
143 (arm_arch5, arm_arch5e): Delete.
144 (arm_arch5t, arm_arch5te): Declare.
145 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
146 (arm_arch5t): Declare.
147 (arm_option_reconfigure_globals): Update for the above.
148 (arm_options_perform_arch_sanity_checks): Update comment, replace
149 use of arm_arch5 with arm_arch5t.
150 (use_return_insn): Likewise.
151 (arm_emit_call_insn): Likewise.
152 (output_return_instruction): Likewise.
153 (arm_final_prescan_insn): Likewise.
154 (arm_coproc_builtin_available): Likewise.
155 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
156 arm_arch5e with arm_arch5t and arm_arch5te.
157 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
158 (arm_arch5t, arm_arch5te): Declare.
159 * config/arm/arm-tables.opt: Regenerate.
160 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
161 * config/arm/t-multilib: Likewise.
162 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
163 instead of arm_arch5.
164 (*call_reg_thumb1): Likewise.
165 (*call_value_reg_thumb1_v5): Likewise.
166 (*call_value_reg_thumb1): Likewise.
167 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
169 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
171 2018-05-18 Martin Liska <mliska@suse.cz>
173 PR gcov-profile/84846
174 * doc/gcov.texi: Document -t option of gcov tool.
176 2018-05-18 Martin Liska <mliska@suse.cz>
178 PR gcov-profile/84846
179 * gcov.c (print_usage): Add new -t option.
180 (process_args): Handle the option.
181 (generate_results): Use stdout as output when requested by
184 2018-05-18 Martin Liska <mliska@suse.cz>
186 PR gcov-profile/84846
187 * coverage.c (coverage_init): Write PWD to .gcno file.
188 * doc/gcov.texi: Document how working directory is printed.
189 * gcov-dump.c (dump_gcov_file): Print PWD.
190 * gcov.c (output_intermediate_file): Likewise.
191 (read_graph_file): Read PWD string.
192 (output_lines): Print PWD.
194 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
197 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
198 for retval and return false if all args to phi are zero.
200 2018-05-18 Richard Biener <rguenther@suse.de>
202 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
204 (evrp_dom_walker::before_dom_children): Call it.
206 2018-05-18 Richard Biener <rguenther@suse.de>
208 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
209 results when processing array refs with variable index.
211 2018-05-18 Toon Moene <toon@moene.org>
213 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
214 directly after that of -floop-interchange. Indicate that both
215 options are enabled by default when specifying -O3.
217 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
219 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
220 iterator. Delete separate integer-mode vec_set<mode> expander.
221 (aarch64_simd_vec_setv2di): Delete.
222 (vec_setv2di): Delete.
223 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
224 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
225 the "w, r" alternative.
227 2018-05-18 Martin Liska <mliska@suse.cz>
229 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
230 * tree-pass.h (make_pass_lower_switch_O0): New function.
231 * tree-switch-conversion.c (node_has_low_bound): Remove.
232 (node_has_high_bound): Likewise.
233 (node_is_bounded): Likewise.
234 (class pass_lower_switch): Make it a template type and create
236 (pass_lower_switch::execute): Add template argument.
237 (make_pass_lower_switch): New function.
238 (make_pass_lower_switch_O0): New function.
239 (do_jump_if_equal): Remove.
240 (emit_case_nodes): Simplify to just handle all 3 cases and leave
241 all the hard work to tree optimization passes.
243 2018-05-18 Martin Liska <mliska@suse.cz>
245 * dbgcnt.c (limit_low): Renamed from limit.
246 (limit_high): New variable.
247 (dbg_cnt_is_enabled): Check for upper limit.
248 (dbg_cnt): Adjust dumping.
249 (dbg_cnt_set_limit_by_index): Add new argument for high
251 (dbg_cnt_set_limit_by_name): Likewise.
252 (dbg_cnt_process_single_pair): Parse new format.
253 (dbg_cnt_process_opt): Use strtok.
254 (dbg_cnt_list_all_counters): Remove 'value' and add
256 * doc/invoke.texi: Document changes.
258 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
260 * doc/sourcebuild.texi (scalar_all_fma): Document.
261 * tree.def (FMA_EXPR): Delete.
262 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
263 * internal-fn.c (ternary_direct): New macro.
264 (expand_ternary_optab_fn): Likewise.
265 (direct_ternary_optab_supported_p): Likewise.
266 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
267 * builtins.c (fold_builtin_fma): Delete.
268 (fold_builtin_3): Don't call it.
269 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
270 * expr.c (expand_expr_real_2): Likewise.
271 * fold-const.c (operand_equal_p): Likewise.
272 (fold_ternary_loc): Likewise.
273 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
274 * gimple.c (DEFTREECODE): Likewise.
275 * gimplify.c (gimplify_expr): Likewise.
276 * optabs-tree.c (optab_for_tree_code): Likewise.
277 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
278 * tree-eh.c (operation_could_trap_p): Likewise.
279 (stmt_could_throw_1_p): Likewise.
280 * tree-inline.c (estimate_operator_cost): Likewise.
281 * tree-pretty-print.c (dump_generic_node): Likewise.
282 (op_code_prio): Likewise.
283 * tree-ssa-loop-im.c (stmt_cost): Likewise.
284 * tree-ssa-operands.c (get_expr_operands): Likewise.
285 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
286 * fold-const-call.h (fold_fma): Delete.
287 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
288 CFN_FNMA and CFN_FNMS.
290 * genmatch.c (combined_fn): New enum.
291 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
292 (commutative_op): New function.
293 (commutate): Use it. Handle more than 2 operands.
294 (dt_operand::gen_gimple_expr): Use commutative_op.
295 (parser::parse_expr): Allow :c to be used with non-binary
296 operators if the commutative operand is known.
297 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
298 CFN_FMS, CFN_FNMA and CFN_FNMS.
299 (backprop::process_assign_use): Remove FMA_EXPR handling.
300 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
301 (gen_hsa_fma): New function.
302 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
303 IFN_FNMA and IFN_FNMS.
304 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
305 * gimple-fold.h (follow_all_ssa_edges): Declare.
306 * gimple-fold.c (follow_all_ssa_edges): New function.
307 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
308 gimple_build interface and use follow_all_ssa_edges to fold the result.
309 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
310 instead of checking for optabs directly.
311 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
312 rather than FMA_EXPRs.
313 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
314 call to IFN_FMA instead of an FMA_EXPR.
316 2018-05-17 Jim Wilson <jimw@sifive.com>
318 * expr.c (do_tablejump): When converting index to Pmode, if we have a
319 sign extended promoted subreg, and the range does not have the sign bit
320 set, then do a sign extend.
322 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
323 test, check for sign extended subreg and/or constant operands, and
324 do a sign extend in that case.
326 2018-05-17 Steve Ellcey <sellcey@cavium.com>
328 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
329 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
331 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
332 Change logics_shift_reg to logics_shift_imm.
333 (thunderx2t99_fp_loadpair_basic): Delete.
334 (thunderx2t99_fp_storepair_basic): Delete.
335 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
336 (thunderx2t99_asimd_polynomial): Delete.
337 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
338 and neon_fp_mul_d_scalar_q.
339 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
340 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
341 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
342 (thunderx2t99_asimd_lut): Add missing tbl types.
343 (thunderx2t99_asimd_ext): Delete.
344 (thunderx2t99_asimd_load1_1_mult): Delete.
345 (thunderx2t99_asimd_load1_2_mult): Delete.
346 (thunderx2t99_asimd_load1_ldp): New.
347 (thunderx2t99_asimd_load1): New.
348 (thunderx2t99_asimd_load2): Add missing *load2* types.
349 (thunderx2t99_asimd_load3): New.
350 (thunderx2t99_asimd_load4): New.
351 (thunderx2t99_asimd_store1_1_mult): Delete.
352 (thunderx2t99_asimd_store1_2_mult): Delete.
353 (thunderx2t99_asimd_store2_mult): Delete.
354 (thunderx2t99_asimd_store2_onelane): Delete.
355 (thunderx2t99_asimd_store_stp): New.
356 (thunderx2t99_asimd_store1): New.
357 (thunderx2t99_asimd_store2): New.
358 (thunderx2t99_asimd_store3): New.
359 (thunderx2t99_asimd_store4): New.
361 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
363 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
364 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
366 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
367 Segher Boessenkool <segher@kernel.crashing.org>
370 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
373 2018-05-17 Richard Biener <rguenther@suse.de>
375 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
376 for pruning loop and prune defs feeding only already visited PHIs.
378 2018-05-17 Richard Biener <rguenther@suse.de>
380 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
382 2018-05-17 Bin Cheng <bin.cheng@arm.com>
383 Richard Biener <rguenther@suse.de>
385 PR tree-optimization/85793
386 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
387 for VMAT_ELEMENTWISE.
389 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
391 * internal-fn.h (lookup_internal_fn): Declare
392 * internal-fn.c (lookup_internal_fn): New function.
393 * gimple.c (gimple_build_call_from_tree): Handle calls to
395 * gimple-pretty-print.c (dump_gimple_call): Print "." before
396 internal function names.
397 * tree-pretty-print.c (dump_generic_node): Likewise.
398 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
400 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
402 * gimple-fold.h (gimple_build): Make the function forms take
403 combined_fn rather than built_in_function.
404 (gimple_simplify): Likewise.
405 * gimple-match-head.c (gimple_simplify): Likewise.
406 * gimple-fold.c (gimple_build): Likewise.
407 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
408 rather than gimple_build_call_internal.
409 (get_initial_defs_for_reduction): Likewise.
410 (vect_create_epilog_for_reduction): Likewise.
411 (vectorizable_live_operation): Likewise.
413 2018-05-17 Martin Liska <mliska@suse.cz>
415 * gimple-ssa-sprintf.c (format_directive): Do not use
416 space in between 'G_' and '('.
418 2018-05-17 Jakub Jelinek <jakub@redhat.com>
421 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
422 even if the mask is not all ones.
425 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
427 (ix86_gimple_fold_builtin): Likewise.
430 * config/i386/i386.c: Include tree-vector-builder.h.
431 (ix86_vector_shift_count): New function.
432 (ix86_fold_builtin): Fold shift builtins by scalar count.
433 (ix86_gimple_fold_builtin): Likewise.
435 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
436 _mm512_setzero): New intrinsics.
438 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
439 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
441 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
442 code generation for cases where splatting a value is not useful.
443 * simplify-rtx.c (simplify_ternary_operation): Simplify
444 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
445 mode to a vec_concat.
447 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
449 * config.gcc: Support "goldmont-plus".
450 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
452 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
453 PROCESSOR_GOLDMONT_PLUS.
454 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
455 (processor_target_table): Add "goldmont-plus".
456 (PTA_GOLDMONT_PLUS): Define.
457 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
458 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
459 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
460 (fold_builtin_cpu): Add "goldmont-plus".
461 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
462 (ix86_option_override_internal): Add "goldmont-plus".
463 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
464 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
465 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
466 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
468 2018-05-17 Richard Biener <rguenther@suse.de>
470 PR tree-optimization/85757
471 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
472 remove defs that only feed that PHI from further processing.
474 2018-05-16 Jim Wilson <jimw@sifive.com>
476 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
478 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
480 2018-05-16 Mark Wielaard <mark@klomp.org>
482 * dwarf2out.c (count_index_strings): New function.
483 (output_indirect_strings): Call count_index_strings and generate
484 header for dwarf_version >= 5.
486 2018-05-16 Mark Wielaard <mark@klomp.org>
488 * dwarf2out.c (dwarf_FORM): New function.
489 (set_indirect_string): Use dwarf_FORM.
490 (reset_indirect_string): Likewise.
491 (size_of_die): Likewise.
492 (value_format): Likewise.
493 (output_die): Likewise.
494 (add_skeleton_AT_string): Likewise.
495 (output_macinfo_op): Likewise.
496 (index_string): Likewise.
497 (output_index_string_offset): Likewise.
498 (output_index_string): Likewise.
499 (count_index_strings): Likewise.
501 2018-05-16 Carl Love <cel@us.ibm.com>
503 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
504 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
506 2018-05-16 Martin Jambor <mjambor@suse.cz>
508 * ipa-prop.c (ipa_free_all_edge_args): Remove.
509 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
511 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
513 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
514 (fnma<mode>4): Likewise.
515 (fms<mode>4): Likewise.
516 (fnms<mode>4): Likewise.
517 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
518 (aarch64_fnma<mode>4): Likewise.
519 (aarch64_fms<mode>4): Likewise.
520 (aarch64_fnms<mode>4): Likewise.
521 (aarch64_fnmadd<mode>4): Likewise.
523 2018-05-16 Jason Merrill <jason@redhat.com>
525 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
527 2018-05-16 Richard Biener <rguenther@suse.de>
529 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
530 (dump_stmt_cost): Declare.
531 (add_stmt_cost): Dump cost we add.
532 (add_stmt_costs): New function.
533 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
535 (vect_analyze_stmt): Adjust prototype.
536 (vectorizable_condition): Likewise.
537 (vectorizable_live_operation): Likewise.
538 (vectorizable_reduction): Likewise.
539 (vectorizable_induction): Likewise.
540 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
541 cost vector to pass to vectorizable_ and record afterwards.
542 (vect_model_reduction_cost): Take cost vector argument and adjust.
543 (vect_model_induction_cost): Likewise.
544 (vectorizable_reduction): Likewise.
545 (vectorizable_induction): Likewise.
546 (vectorizable_live_operation): Likewise.
547 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
548 SLP_TREE_NUMBER_OF_VEC_STMTS.
549 (vect_analyze_slp_cost_1): Remove.
550 (vect_analyze_slp_cost): Likewise.
551 (vect_slp_analyze_node_operations): Take visited args and
552 a target cost vector. Avoid processing already visited stmt sets.
553 (vect_slp_analyze_operations): Use a local cost vector to gather
554 costs and register those of non-discarded instances.
555 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
556 (vect_schedule_slp_instance): Remove copying of
557 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
559 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
560 adding cost. Record cost entry location.
561 (vect_prologue_cost_for_slp_op): Function to compute cost of
562 a constant or invariant generated for SLP vect in the prologue,
563 split out from vect_analyze_slp_cost_1.
564 (vect_model_simple_cost): Make static. Adjust for SLP costing.
565 (vect_model_promotion_demotion_cost): Likewise.
566 (vect_model_store_cost): Likewise, make static.
567 (vect_model_load_cost): Likewise.
568 (vectorizable_bswap): Add cost vector arg and adjust.
569 (vectorizable_call): Likewise.
570 (vectorizable_simd_clone_call): Likewise.
571 (vectorizable_conversion): Likewise.
572 (vectorizable_assignment): Likewise.
573 (vectorizable_shift): Likewise.
574 (vectorizable_operation): Likewise.
575 (vectorizable_store): Likewise.
576 (vectorizable_load): Likewise.
577 (vectorizable_condition): Likewise.
578 (vectorizable_comparison): Likewise.
579 (can_vectorize_live_stmts): Likewise.
580 (vect_analyze_stmt): Likewise.
581 (vect_transform_stmt): Adjust calls to vectorizable_*.
582 * tree-vectorizer.c: Include gimple-pretty-print.h.
583 (dump_stmt_cost): New function.
585 2018-05-16 Richard Biener <rguenther@suse.de>
587 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
588 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
589 * tree-ssa-dse.c: Include tree-ssa-loop.h.
590 (check_name): New callback.
591 (dse_classify_store): Track cycles via a visited bitmap of PHI
592 defs and simplify handling of in-loop and across loop dead stores
593 and properly fail for loop-variant refs. Handle byte-tracking with
594 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
597 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
599 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
600 (vect_get_mask_type_for_stmt): Likewise.
601 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
603 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
604 to determine the statement's vector type and the vector type that
605 should be used for calculating nunits. Deal with cases in which
606 the type has to be deferred.
607 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
608 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
609 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
610 (vect_determine_vf_for_stmt): New functions, split out from...
611 (vect_determine_vectorization_factor): ...here.
612 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
613 (vect_get_mask_type_for_stmt): New functions, split out from
614 vect_determine_vectorization_factor.
616 2018-05-16 Richard Biener <rguenther@suse.de>
618 * tree-cfg.c (verify_gimple_assign_ternary): Properly
619 verify the [VEC_]COND_EXPR embedded comparison.
621 2018-05-15 Martin Sebor <msebor@redhat.com>
623 PR tree-optimization/85753
624 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
625 RECORD_TYPE in addition to ARRAY_TYPE.
627 2018-05-15 Martin Sebor <msebor@redhat.com>
630 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
632 2018-05-15 Richard Biener <rguenther@suse.de>
634 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
635 add by_clobber_p one. Change algorithm to collect all defs
636 representing uses we need to walk and try reducing them to
637 a single one before failing.
638 (dse_dom_walker::dse_optimize_stmt): Adjust.
640 2018-05-13 Mark Wielaard <mark@klomp.org>
642 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
643 (size_of_loc_descr): Likewise.
644 (output_loc_operands): Likewise.
645 (output_loc_operands_raw): Likewise.
646 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
647 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
648 (hash_loc_operands): Likewise.
649 (compare_loc_operands): Likewise.
651 2018-05-14 Mark Wielaard <mark@klomp.org>
653 * dwarf2out.c (count_index_addrs): New function.
654 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
656 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
658 PR tree-optimization/83648
659 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
660 return value as malloc candidate.
662 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
665 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
666 as true in call to suggest_attribute.
668 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
670 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
673 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
675 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
677 (aarch64_crypto_aesd_fused): Likewise.
679 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
681 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
682 (movsi_aarch64): Likewise.
683 (load_pairsi): Likewise.
684 (load_pairdi): Likewise.
685 (store_pairsi): Likewise.
686 (store_pairdi): Likewise.
687 (load_pairsf): Likewise.
688 (load_pairdf): Likewise.
689 (store_pairsf): Likewise.
690 (store_pairdf): Likewise.
691 (zero_extend): Likewise.
692 (trunc): Swap alternatives.
693 (fcvt_target): Add '?' to prefer w over r.
695 2018-05-14 Jakub Jelinek <jakub@redhat.com>
698 * config/i386/i386.md: Disallow non-commutative arithmetics in
699 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
700 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
701 in the peephole2 before it.
703 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
705 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
706 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
707 (ix86_handle_option): Handle -mcldemote.
708 * config.gcc: New header.
709 * config/i386/cldemoteintrin.h: New file.
710 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
711 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
713 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
714 OPTION_MASK_ISA_CLDEMOTE.
715 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
716 (ix86_valid_target_attribute_inner_p): Ditto.
717 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
718 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
719 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
720 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
721 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
723 * config/i386/i386.opt: Add -mcldemote.
724 * config/i386/x86intrin.h: New header.
725 * doc/invoke.texi: Add -mcldemote.
727 2018-05-14 Richard Biener <rguenther@suse.de>
729 * doc/match-and-simplify.texi: Adjust :s documentation.
731 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
733 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
734 intended memcpy size.
735 (REORDER_45): Likewise.
737 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
740 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
741 * vec.c (qsort_chk): Use gcc_qsort.
742 * Makefile.in (OBJS-libcommon): Add sort.o.
743 (build/sort.o): New target. Use it...
744 (BUILD_RTL): ... here, and...
745 (build/gencfn-macros): ... here, and...
746 (build/genmatch): ... here.
748 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
749 Chung-Ju Wu <jasonwucj@gmail.com>
751 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
752 * config/nds32/nds32-graywolf.md: New file.
753 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
754 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
756 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
757 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
758 * config/nds32/nds32.md (pipeline_model): Add graywolf.
759 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
760 * config/nds32/pipelines.md: Include n15 settings.
762 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
763 Chung-Ju Wu <jasonwucj@gmail.com>
765 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
766 * config/nds32/nds32-n13.md: New file.
767 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
768 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
770 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
771 * config/nds32/nds32.md (pipeline_model): Add n13.
772 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
773 * config/nds32/pipelines.md: Include n13 settings.
775 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
776 Chung-Ju Wu <jasonwucj@gmail.com>
778 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
779 * config/nds32/nds32-n10.md: New file.
780 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
781 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
783 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
784 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
785 * config/nds32/nds32.md (pipeline_model): Add n10.
786 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
787 * config/nds32/pipelines.md: Include n10 settings.
789 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
790 Kito Cheng <kito.cheng@gmail.com>
791 Chung-Ju Wu <jasonwucj@gmail.com>
793 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
794 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
795 Add enum values for DSP extension instructions.
796 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
798 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
799 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
801 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
802 * config/nds32/nds32-dspext.md: New file for DSP implementation.
803 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
804 * config/nds32/nds32-intrinsic.md: Likewise.
805 * config/nds32/nds32_intrinsic.h: Likewise.
806 * config/nds32/nds32-md-auxiliary.c: Likewise.
807 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
808 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
809 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
810 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
811 * config/nds32/nds32-protos.h: New declarations for DSP extension.
812 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
813 TYPE_DMAC in switch statement.
814 * config/nds32/nds32.c: New checking and implementation for DSP
815 extension instructions.
816 * config/nds32/nds32.h: Likewise.
817 * config/nds32/nds32.md: Likewise.
818 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
819 * config/nds32/predicates.md: Implement new predicates for DSP
822 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
824 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
825 Reformat alternatives and attributes so it is easier to identify
826 which constraints/attributes go with which instruction.
827 (mov<mode>_hardfloat32, FMOVE64): Likewise.
828 (mov<mode>_softfloat32, FMOVE64): Likewise.
829 (mov<mode>_hardfloat64, FMOVE64): Likewise.
830 (mov<mode>_softfloat64, FMOVE64): Likewise.
832 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
834 * doc/extend.texi (PowerPC Built-in Functions): Rename this
836 (Basic PowerPC Built-in Functions): The new name of the
837 subsection previously known as "PowerPC Built-in Functions".
838 (Basic PowerPC Built-in Functions Available on all Configurations):
840 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
841 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
842 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
843 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
845 2018-05-11 Martin Jambor <mjambor@suse.cz>
848 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
851 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
854 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
856 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
858 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
859 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
860 (ix86_handle_option): Handle -mwaitpkg.
861 * config.gcc: New header.
862 * config/i386/cpuid.h (bit_WAITPKG): New bit.
863 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
864 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
866 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
867 OPTION_MASK_ISA_WAITPKG.
868 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
869 (ix86_option_override_internal): Add PTA_WAITPKG.
870 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
871 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
873 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
874 __builtin_ia32_umwait and __builtin_ia32_tpause.
875 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
876 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
877 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
878 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
879 UNSPECV_TPAUSE): New.
880 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
881 * config/i386/i386.opt: Add -mwaitpkg.
882 * config/i386/waitpkgintrin.h: New file.
883 * config/i386/x86intrin.h: New header.
884 * doc/invoke.texi: Add -mwaitpkg.
886 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
889 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
891 (cortex-m0): Use armv6s-m isa.
892 (cortex-m0plus): Likewise.
893 (cortex-m1): Likewise.
894 (cortex-m0.small-multiply): Likewise.
895 (cortex-m0plus.small-multiply): Likewise.
896 (cortex-m1.small-multiply): Likewise.
898 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
899 Jakub Jelinek <jakub@redhat.com>
901 PR tree-optimization/85692
902 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
903 source permute as well.
905 2018-05-11 Martin Liska <mliska@suse.cz>
908 * doc/extend.texi: Document LLVM style format for no_sanitize
911 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
913 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
914 mode_supports_vsx_dform_quad to mode_supports_dq_form.
915 (mode_supports_vsx_dform_quad): Likewise.
916 (mode_supports_vmx_dform): Move these functions to be next to the
917 other mode_supports functions.
918 (mode_supports_dq_form): Likewise.
919 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
920 mode_supports_dq_form.
921 (reg_offset_addressing_ok_p): Likewise.
922 (offsettable_ok_by_alignment): Likewise.
923 (rs6000_legitimate_offset_address_p): Likewise.
924 (legitimate_lo_sum_address_p): Likewise.
925 (rs6000_legitimize_address): Likewise.
926 (rs6000_legitimize_reload_address): Likewise.
927 (rs6000_secondary_reload_inner): Likewise.
928 (rs6000_preferred_reload_class): Likewise.
929 (rs6000_output_move_128bit): Likewise.
931 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
933 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
934 Generate SImode target register for null target.
935 <case IX86_BUILTIN_XGETBV>: Ditto.
936 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
937 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
939 2018-05-10 Carl Love <cel@us.ibm.com>
941 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
942 dcbtt and dcbtstt if operands[2] is 0.
944 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
947 * config/i386/sse.md (usadv64qi): New expander.
949 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
951 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
952 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
953 -maltivec=be support.
954 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
955 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
956 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
957 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
958 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
959 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
960 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
961 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
962 altivec_vsumsws): Adjust.
963 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
964 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
965 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
967 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
968 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
969 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
970 (altivec_lve<VI_char>x): Delete expand.
971 (*altivec_lve<VI_char>x_internal): Rename to...
972 (altivec_lve<VI_char>x): ... this.
973 (altivec_lvxl_<mode>): Delete expand.
974 (*altivec_lvxl_<mode>_internal): Rename to ...
975 (altivec_lvxl_<mode>): ... this.
976 (altivec_stvxl_<mode>): Delete expand.
977 (*altivec_stvxl_<mode>_internal): Rename to ...
978 (altivec_stvxl_<mode>): ... this.
979 (altivec_stve<VI_char>x): Delete expand.
980 (*altivec_stve<VI_char>x_internal): Rename to ...
981 (altivec_stve<VI_char>x): ... this.
982 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
983 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
984 reduc_plus_scal_<mode>): Adjust.
985 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
987 (rs6000_cpu_cpp_builtins): Adjust.
988 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
989 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
990 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
991 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
992 -maltivec=be support.
993 (rs6000_split_vec_extract_var): Adjust.
994 (rs6000_split_v4si_init): Adjust.
995 (swap_selector_for_mode): Delete.
996 (altivec_expand_lvx_be, altivec_expand_stvx_be,
997 altivec_expand_stvex_be): Delete.
998 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
999 -maltivec=be support.
1000 (rs6000_gimple_fold_builtin): Ditto.
1001 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1003 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1004 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1005 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1006 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1007 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1008 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1009 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1010 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1011 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1012 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1013 anonymous split): Adjust.
1014 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1015 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1017 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1019 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1020 when --with-gxx-include-dir is also specified.
1021 * configure: Regenerate.
1023 2018-05-09 Jim Wilson <jimw@sifive.com>
1026 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1027 * config/riscv/t-withmultilib: New.
1028 * config/riscv/withmultilib.h: New.
1029 * doc/install.texi: Document RISC-V --with-multilib-list support.
1031 2018-05-09 Richard Biener <rguenther@suse.de>
1033 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1035 (vect_bb_vectorization_profitable_p): Adjust. Compute
1036 actual scalar cost using the cost vector and the add_stmt_cost
1039 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1041 PR rtl-optimization/85645
1042 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1043 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1045 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1047 PR rtl-optimization/85645
1048 * shrink-wrap.c (spread_components): Return a boolean saying if
1049 anything was changed.
1050 (try_shrink_wrapping_separate): Iterate spread_components until
1051 nothing changes anymore.
1053 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1055 PR rtl-optimization/85645
1056 * regrename.c (build_def_use): Also kill the chains that include the
1057 destination of a REG_CFA_REGISTER note.
1059 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1061 PR rtl-optimization/85645
1062 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1063 insn that has a REG_CFA_REGISTER note.
1065 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1067 * cfgexpand.c (expand_clobber): New function.
1068 (expand_gimple_stmt_1): Use it.
1069 * tree-vect-stmts.c (vect_clobber_variable): New function,
1071 (vectorizable_simd_clone_call): ...here.
1072 (vectorizable_store): Emit a clobber either side of an
1073 IFN_STORE_LANES sequence.
1074 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1076 2018-05-09 Tom de Vries <tom@codesourcery.com>
1079 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1080 (define_insn "trap_if_false"): Add exit after trap.
1082 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1084 PR rtl-optimization/85638
1085 * bb-reorder.c: Include common/common-target.h.
1086 (create_forwarder_block): New function extracted from...
1087 (fix_up_crossing_landing_pad): ...here. Rename into...
1088 (dw2_fix_up_crossing_landing_pad): ...this.
1089 (sjlj_fix_up_crossing_landing_pad): New function.
1090 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1091 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1092 from both partitions and exit the loop after one iteration.
1094 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1097 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1099 (Basic PowerPC Built-in Functions): The new name of the
1100 subsection previously known as "PowerPC Built-in Functions".
1101 (Basic PowerPC Built-in Functions Available on all Configurations):
1103 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1105 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1107 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1109 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1112 2018-05-08 Jim Wilson <jimw@sifive.com>
1114 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1115 (LD_EMUL_SUFFIX): New.
1116 (LINK_SPEC): Use it.
1118 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1120 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1122 (Basic PowerPC Built-in Functions): The new name of the
1123 subsection previously known as "PowerPC Built-in Functions".
1124 (Basic PowerPC Built-in Functions Available on all Configurations):
1126 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1128 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1130 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1132 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1135 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1138 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1139 after cmpelim optimization.
1141 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1143 * config.gcc: Support "goldmont".
1144 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1145 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1147 * config/i386/i386.c (m_GOLDMONT): Define.
1148 (processor_target_table): Add "goldmont".
1149 (PTA_GOLDMONT): Define.
1150 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1151 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1152 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1153 (fold_builtin_cpu): Add "goldmont".
1154 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1155 (ix86_option_override_internal): Add "goldmont".
1156 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1157 (processor_type): Add PROCESSOR_GOLDMONT.
1158 * config/i386/i386.md: Add CPU "glm".
1159 * config/i386/glm.md: New file.
1160 * config/i386/x86-tune.def: Add m_GOLDMONT.
1161 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1163 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1166 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1168 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1169 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1170 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1173 * config/i386/i386.c (ix86_fold_builtin): Handle
1174 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1177 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1178 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1180 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1183 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1184 (check_arch): Likewise.
1185 (check_fpu): Return the result rather than printing it.
1186 (end arch): Fix operator precedence.
1187 (end cpu): Likewise.
1188 (END): Print the result from check_fpu.
1190 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1191 Alan Hayward <alan.hayward@arm.com>
1192 David Sherwood <david.sherwood@arm.com>
1194 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1195 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1196 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1197 (*fcmuo<mode>_and): New patterns.
1199 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1201 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1202 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1203 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1204 (cmp_op, sve_imm_con): New code attributes.
1205 (SVE_COND_INT_CMP, imm_con): Delete.
1206 (cmp_op): Remove above unspecs from int attribute.
1207 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1209 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1210 comparison-specific unspecs.
1211 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1212 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1213 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1214 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1215 (*vec_fcm<cmp_op><mode>): Rename to...
1216 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1217 (*vec_fcmuo<mode>): Rename to...
1218 (*fcmuo<mode>): ...this and adjust likewise.
1219 (*pred_fcm<cmp_op><mode>): New pattern.
1220 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1221 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1223 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1225 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1226 (aarch64_emit_sve_predicated_cond): New function.
1227 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1228 (aarch64_emit_unspec_cond_or): Replace with...
1229 (aarch64_emit_sve_or_conds): ...this new function. Use
1230 aarch64_emit_sve_ptrue_op for the individual comparisons and
1231 aarch64_emit_binop to OR them together.
1232 (aarch64_emit_inverted_unspec_cond): Replace with...
1233 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1234 aarch64_emit_sve_ptrue_op for the comparison and
1235 aarch64_emit_unop to invert the result.
1236 (aarch64_expand_sve_vec_cmp_float): Update after the above
1237 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1239 2018-05-07 Nathan Sidwell <nathan@acm.org>
1241 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1242 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1243 (Backwards Compatibility): Likewise.
1245 2018-05-07 Luis Machado <luis.machado@linaro.org>
1249 2018-05-07 Luis Machado <luis.machado@linaro.org>
1251 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1252 <prefetch_dynamic_strides>: New const bool field.
1253 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1254 prefetch_dynamic_strides.
1255 (exynosm1_prefetch_tune): Likewise.
1256 (thunderxt88_prefetch_tune): Likewise.
1257 (thunderx_prefetch_tune): Likewise.
1258 (thunderx2t99_prefetch_tune): Likewise.
1259 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1260 (aarch64_override_options_internal): Update to set
1261 PARAM_PREFETCH_DYNAMIC_STRIDES.
1262 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1263 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1264 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1265 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1266 prefetch-dynamic-strides setting.
1268 2018-05-07 Luis Machado <luis.machado@linaro.org>
1270 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1271 <minimum_stride>: New const int field.
1272 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1273 minimum_stride field.
1274 (exynosm1_prefetch_tune): Likewise.
1275 (thunderxt88_prefetch_tune): Likewise.
1276 (thunderx_prefetch_tune): Likewise.
1277 (thunderx2t99_prefetch_tune): Likewise.
1278 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1279 (aarch64_override_options_internal): Update to set
1280 PARAM_PREFETCH_MINIMUM_STRIDE.
1281 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1282 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1283 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1284 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1285 stride is constant and is below the minimum stride threshold.
1287 2018-05-07 Luis Machado <luis.machado@linaro.org>
1289 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1292 2018-05-07 Luis Machado <luis.machado@linaro.org>
1294 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1295 <prefetch_dynamic_strides>: New const bool field.
1296 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1297 prefetch_dynamic_strides.
1298 (exynosm1_prefetch_tune): Likewise.
1299 (thunderxt88_prefetch_tune): Likewise.
1300 (thunderx_prefetch_tune): Likewise.
1301 (thunderx2t99_prefetch_tune): Likewise.
1302 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1303 (aarch64_override_options_internal): Update to set
1304 PARAM_PREFETCH_DYNAMIC_STRIDES.
1305 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1306 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1307 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1308 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1309 prefetch-dynamic-strides setting.
1311 2018-05-07 Luis Machado <luis.machado@linaro.org>
1313 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1314 <minimum_stride>: New const int field.
1315 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1316 minimum_stride field.
1317 (exynosm1_prefetch_tune): Likewise.
1318 (thunderxt88_prefetch_tune): Likewise.
1319 (thunderx_prefetch_tune): Likewise.
1320 (thunderx2t99_prefetch_tune): Likewise.
1321 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1322 (aarch64_override_options_internal): Update to set
1323 PARAM_PREFETCH_MINIMUM_STRIDE.
1324 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1325 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1326 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1327 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1328 stride is constant and is below the minimum stride threshold.
1330 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1333 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1334 the type is addressable. Don't force op into register if it has
1337 2018-05-05 Roland McGrath <mcgrathr@google.com>
1340 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1341 any section for which we don't know a specific type it should have,
1342 regardless of name. Previously this was done only for the exact
1343 names ".init_array", ".fini_array", and ".preinit_array".
1344 (default_elf_asm_named_section): Add comment about
1345 relationship with default_section_type_flags and SECTION_NOTYPE.
1346 (get_section): Don't consider it a type conflict if one side has
1347 SECTION_NOTYPE and the other doesn't, as long as neither has the
1348 SECTION_BSS et al used in the default_section_type_flags logic.
1350 2018-05-05 Tom de Vries <tom@codesourcery.com>
1353 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1354 (workaround_barsyncs): New function.
1355 (nvptx_reorg): Use workaround_barsyncs.
1356 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1357 (define_expand "nvptx_membar_cta"): New define_expand.
1358 (define_insn "*nvptx_membar_cta"): New insn.
1360 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1362 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1363 To improve optimization opportunities.
1364 * builtin-types.def: The new needed builtin types for the above.
1366 2018-05-04 Richard Biener <rguenther@suse.de>
1368 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1369 * gimple-ssa-store-merging.c
1370 (imm_store_chain_info::output_merged_store): Remove redundant create,
1371 release split_store vector contents on failure.
1372 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1373 scalar stmt vector on cache hit.
1375 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1377 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1379 * config.gcc (powerpc-xilinx-eabi*): Remove.
1380 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1382 (fusion_addis_mem_combo_load): Ditto.
1383 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1385 (rs6000_cpu_cpp_builtins): Ditto.
1386 * config/rs6000/rs6000-linux.c
1387 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1388 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1389 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1391 (rs6000_setup_reg_addr_masks): Ditto.
1392 (rs6000_init_hard_regno_mode_ok): Ditto.
1393 (rs6000_option_override_internal): Ditto.
1394 (legitimate_lo_sum_address_p): Ditto.
1395 (rs6000_legitimize_address): Ditto.
1396 (rs6000_legitimize_reload_address): Ditto.
1397 (rs6000_legitimate_address_p): Ditto.
1398 (abi_v4_pass_in_fpr): Ditto.
1399 (setup_incoming_varargs): Ditto.
1400 (rs6000_gimplify_va_arg): Ditto.
1401 (rs6000_split_multireg_move): Ditto.
1402 (rs6000_savres_strategy): Ditto.
1403 (rs6000_emit_prologue_components): Ditto.
1404 (rs6000_emit_epilogue_components): Ditto.
1405 (rs6000_emit_prologue): Ditto.
1406 (rs6000_emit_epilogue): Ditto.
1407 (rs6000_elf_file_end): Ditto.
1408 (rs6000_function_value): Ditto.
1409 (rs6000_libcall_value): Ditto.
1410 * config/rs6000/rs6000.h: Ditto.
1411 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1412 (TARGET_MINMAX): ... this. New.
1413 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1414 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1415 (*movsi_internal1_single): Delete.
1416 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1417 mfpu=, mxilinx-fpu): Delete.
1418 * config/rs6000/singlefp.h: Delete.
1419 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1420 * config/rs6000/t-rs6000: Ditto.
1421 * config/rs6000/t-xilinx: Delete.
1422 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1423 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1424 (VStype_simple): Delete.
1425 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1426 * config/rs6000/xfpu.h: Delete.
1427 * config/rs6000/xfpu.md: Delete.
1428 * config/rs6000/xilinx.h: Delete.
1429 * config/rs6000/xilinx.opt: Delete.
1430 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1431 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1433 2018-05-04 Tom de Vries <tom@codesourcery.com>
1436 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1439 2018-05-04 Richard Biener <rguenther@suse.de>
1442 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1443 (expand_complex_div_wide): Likewise.
1444 (expand_complex_operations_1): Likewise.
1445 (expand_complex_libcall): Preserve EH info of the original stmt.
1446 (tree_lower_complex): Handle removed blocks.
1447 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1448 on complex multiplication and division libcall builtins.
1450 2018-05-04 Richard Biener <rguenther@suse.de>
1453 * fold-const.c (negate_expr_p): Restrict negation of operand
1454 zero of a division to when we know that can happen without
1456 (fold_negate_expr_1): Likewise.
1458 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1461 * real.h (real_nextafter): Declare.
1462 * real.c (real_nextafter): New function.
1463 * fold-const-call.c (fold_const_nextafter): New function.
1464 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1465 CASE_CFN_NEXTTOWARD.
1466 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1467 even when arg1_mode is different from arg0_mode.
1469 2018-05-03 Nathan Sidwell <nathan@acm.org>
1471 * doc/extend.texi (Deprecated Features): Remove
1473 (Backwards Compatibility): Likewise.
1474 * doc/invoke.texi (C++ Language Options): Likewise.
1475 (C++ Dialect Options): Likewise.
1477 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1480 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1481 _mm512_mask_mullox_epi64): New intrinsics.
1483 2018-05-03 Tom de Vries <tom@codesourcery.com>
1486 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1487 dump files): Add offload-tree.
1489 2018-05-03 Richard Biener <rguenther@suse.de>
1491 PR tree-optimization/85615
1492 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1493 to loops not nested in BBs loop father to avoid creating multi-entry
1496 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1498 PR tree-optimization/70291
1499 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1500 arguments. Change return type to tree. Emit libcall as a new
1501 statement rather than replacing existing one when inplace_p is true.
1502 (expand_complex_multiplication_components): New function.
1503 (expand_complex_multiplication): Expand floating-point complex
1504 multiplication using the above.
1505 (expand_complex_division): Rename inner_type parameter to type.
1506 Update expand_complex_libcall call-site.
1507 (expand_complex_operations_1): Update expand_complex_multiplication
1508 and expand_complex_division call-sites.
1510 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1513 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1514 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1515 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1516 the highest significant bit of the shift count mask is clear. In
1517 check whether and[sq]i3 is needed verify that all significant bits
1518 of the shift count other than the highest are set.
1520 2018-05-02 Tom de Vries <tom@codesourcery.com>
1523 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1524 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1525 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1526 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1527 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1528 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1529 __builtin_goacc_parlevel_size.
1531 2018-05-02 Richard Biener <rguenther@suse.de>
1533 PR tree-optimization/85597
1534 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1535 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1538 2018-05-02 Tom de Vries <tom@codesourcery.com>
1541 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1542 dump files): Add ltrans-tree.
1544 2018-05-02 Tom de Vries <tom@codesourcery.com>
1547 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1548 dump files): Add wpa-ipa.
1550 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1552 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1553 powerpc*-*-linux*paired* target.
1554 * config/rs6000/750cl.h: Delete.
1555 * config/rs6000/paired.h: Delete.
1556 * config/rs6000/paired.md: Delete.
1557 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1559 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1560 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1561 comment. Remove paired float support.
1562 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1563 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1565 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1566 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1568 * config/rs6000/rs6000.c: Remove paired float support.
1569 (paired_expand_vector_init, paired_expand_vector_move,
1570 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1571 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1572 paired_expand_builtin, paired_expand_predicate_builtin,
1573 paired_init_builtins): Delete.
1574 * config/rs6000/rs6000.h: Remove paired float support.
1575 * config/rs6000/rs6000.md: Remove paired float support.
1576 (move_from_CR_ov_bit): Delete.
1577 * config/rs6000/rs6000.opt (mpaired): Delete.
1578 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1579 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1581 2018-05-02 Richard Biener <rguenther@suse.de>
1584 * gimplify.c (gimplify_save_expr): When in SSA form allow
1585 SAVE_EXPRs to compute to SSA vars.
1587 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1590 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1591 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1592 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1593 clobber operands[2], instead use a new pseudo. Formatting fixes.
1595 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1597 PR tree-optimization/85586
1598 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1599 exit early for statements in the same group if the accesses are
1602 2018-05-02 Tom de Vries <tom@codesourcery.com>
1605 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1608 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1610 PR tree-optimization/85143
1611 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1613 2018-05-01 Tom de Vries <tom@codesourcery.com>
1616 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1617 not found" error message.
1619 2018-05-01 Tom de Vries <tom@codesourcery.com>
1622 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1623 * vec.c (test_ordered_remove_if): New function.
1624 (vec_c_tests): Call test_ordered_remove_if.
1625 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1626 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1627 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1628 VEC_ORDERED_REMOVE_IF.
1630 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1632 PR tree-optimization/82665
1633 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1634 pointer subtraction where arguments come from a memchr call.
1636 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1638 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1639 --push-state --as-needed and --pop-state instead of --as-needed and
1640 --no-as-needed if ld supports it.
1641 * configure: Regenerated.
1644 * doc/install.texi2html: Replace _002d with - and _002a with * in
1645 generated html files using sed.
1647 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1650 * gcc-rich-location.c (blank_line_before_p): New function.
1651 (use_new_line): New function.
1652 (gcc_rich_location::add_fixit_insert_formatted): New function.
1653 * gcc-rich-location.h
1654 (gcc_rich_location::add_fixit_insert_formatted): New function.
1656 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1658 * selftest.c (assert_streq): Rename "expected" and "actual" to
1659 "val1" and "val2". Extend NULL-handling to cover both inputs
1660 symmetrically, while still requiring both to be non-NULL for a pass.
1661 * selftest.h (assert_streq): Rename "expected" and "actual" to
1663 (ASSERT_EQ): Likewise.
1664 (ASSERT_EQ_AT): Likewise.
1665 (ASSERT_KNOWN_EQ): Likewise.
1666 (ASSERT_KNOWN_EQ_AT): Likewise.
1667 (ASSERT_NE): Likewise.
1668 (ASSERT_MAYBE_NE): Likewise.
1669 (ASSERT_MAYBE_NE_AT): Likewise.
1670 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1671 the assertion to pass.
1672 (ASSERT_STREQ_AT): Likewise.
1674 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1676 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1677 interaction with -pie.
1679 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1681 * selftest.h: Fix alphabetization of per-source-file selftest
1684 2018-04-30 Jason Merrill <jason@redhat.com>
1686 PR c++/61982 - dead stores to destroyed objects.
1687 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1690 2018-04-30 Jason Merrill <jason@redhat.com>
1692 * tree.c (build_clobber): New.
1693 * tree.h: Declare it.
1694 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1696 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1698 * diagnostic-show-locus.c (layout::layout): Update for
1699 location_get_source_line returning a char_span.
1700 (struct char_span): Move to input.h.
1701 (struct correction): Update for fields in char_span becoming
1703 (struct source_line): Update for location_get_source_line
1704 returning a char_span.
1705 (layout::print_line): Likewise.
1706 * edit-context.c (edited_file::print_content): Likewise.
1707 (edited_file::print_diff_hunk): Likewise.
1708 (edited_file::print_run_of_changed_lines): Likewise.
1709 (edited_file::get_num_lines): Likewise.
1710 (edited_line::edited_line): Likewise.
1711 * final.c (asm_show_source): Likewise.
1712 * input.c (location_get_source_line): Convert return type
1713 from const char * to char_span, losing the final "line_len"
1715 (dump_location_info): Update for the above.
1716 (get_substring_ranges_for_loc): Likewise. Use a char_span
1717 when handling the literal within the line.
1718 (test_reading_source_line): Update for location_get_source_line
1719 returning a char_span.
1720 * input.h (class char_span): Move here from
1721 diagnostic-show-locus.c, converting from a struct to a class.
1722 Make data members private.
1723 (char_span::operator bool): New.
1724 (char_span::length): New.
1725 (char_span::get_buffer): New.
1726 (char_span::operator[]): New.
1727 (char_span::subspan): Make const.
1728 (char_span::xstrdup): New.
1729 (location_get_source_line): Convert return type from const char *
1730 to char_span, losing the final "line_size" param.
1732 2018-04-30 Jan Hubicka <jh@suse.cz>
1734 * lto-wrapper.c (ltrans_priorities): New static var.
1735 (cmp_priority): New.
1736 (run_gcc): Read priorities and if doing parallel build order
1737 the Makefile by them.
1739 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1741 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1743 2018-04-30 Richard Biener <rguenther@suse.de>
1745 * tree-cfg.c (verify_address): Remove base argument, add
1746 flag whether to check TREE_ADDRESSABLE and do that.
1747 (verify_expr): Remove.
1748 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1749 (verify_gimple_assign_single): Likewise.
1750 (verify_gimple_switch): Likewise.
1751 (verify_expr_location_1): Dereference tp once. Add (disabled)
1752 piece from verify_expr.
1753 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1755 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1757 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1759 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1761 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1762 (small_data_pattern): Likewise.
1763 (arc_rewrite_small_data): Likewise.
1764 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1765 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1766 (get_symbol_alignment): New function.
1767 (legitimate_small_data_address_p): Likewise.
1768 (legitimate_scaled_address): Update, call
1769 legitimate_small_data_address_p.
1770 (output_sdata): New static variable.
1771 (arc_print_operand): Update how we handle small data operands.
1772 (arc_print_operand_address): Likewise.
1773 (arc_legitimate_address_p): Update, use
1774 legitimate_small_data_address_p.
1775 (arc_rewrite_small_data_p): Remove.
1776 (arc_rewrite_small_data_1): Likewise.
1777 (arc_rewrite_small_data): Likewise.
1778 (small_data_pattern): Likewise.
1779 (compact_sda_memory_operand): Update to use
1780 legitimate_small_data_address_p and get_symbol_alignment.
1781 (prepare_move_operands): Don't rewite sdata pattern.
1782 (prepare_extend_operands): Remove.
1783 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1785 (zero_extendqisi2): Likewise.
1786 (zero_extendhisi2): Likewise.
1787 (extendqihi2): Likewise.
1788 (extendqisi2): Likewise.
1789 (extendhisi2): Likewise.
1793 * config/arc/constraints.md (Usd): Change it to memory constraint.
1795 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1797 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1798 as source of std instructions.
1799 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1800 allow 6-bit constants as source for store instructions.
1801 (movdi_insn): Update instruction pattern to allow 6-bit constants
1802 as source for store instructions.
1804 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1806 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1808 2018-04-30 Nathan Sidwell <nathan@acm.org>
1809 Sandra Loosemore <sandra@codesourcery.com>
1811 * dumpfile.c (dump_open): Allow '-' for stdout.
1812 * doc/invoke.texi (Developer Options): Document dump filename
1813 determination early. Document stdin/stdout selection.
1815 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1817 Microblaze Target: PIC data text relative
1819 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1820 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1822 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1823 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1824 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1825 New addressing mode for data-text relative position indepenedent code.
1826 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1827 'ADDRESS_SYMBOLIC_TXT_REL'.
1828 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1829 (microblaze_legitimate_pic_operand): Exclude function calls from
1830 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1831 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1833 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1834 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1835 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1836 for 'address + offset'.
1837 (microblaze_expand_prologue): Add new function prologue call for
1839 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1840 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1841 table in case of TARGET_PIC_DATA_TEXT_REL.
1842 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1843 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1844 Add new macros 'UNSPEC_TEXT',
1845 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1846 + exclude function calls from 'UNSPEC_PLT' in case of data text
1848 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1849 new target hook for generating address diff vector tables in case of
1851 * doc/tm.texi : Regenerate.
1852 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1853 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1854 of addr diff vector generation.
1855 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1856 target hook definition.
1857 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1858 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1859 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1860 option for data text relative.
1862 2018-04-30 Richard Biener <rguenther@suse.de>
1864 * tree-chrec.h (evolution_function_is_constant_p): Remove
1866 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1868 2018-04-30 Richard Biener <rguenther@suse.de>
1871 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1873 2018-04-30 Richard Biener <rguenther@suse.de>
1875 PR tree-optimization/28364
1876 PR tree-optimization/85275
1877 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1878 copying first exit test.
1880 2018-04-28 Mark Wielaard <mark@klomp.org>
1882 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1884 (dwarf_AT): Handle DW_AT_addr_base.
1885 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1887 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1890 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1891 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1892 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1893 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1895 2018-04-28 Richard Biener <rguenther@suse.de>
1897 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1898 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1899 to reflect use. Only add interesting stmts.
1901 2018-04-27 Martin Jambor <mjambor@suse.cz>
1904 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1905 the jump function allows for passing through aggregate values.
1907 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1909 * input.h (in_system_header_at): Convert from macro to inline
1911 (from_macro_expansion_at): Likewise.
1912 (from_macro_definition_at): Likewise.
1914 2018-04-27 Jeff Law <law@redhat.com>
1916 * config.gcc: Mark tile* targets as deprecated/obsolete.
1918 2018-04-27 Richard Biener <rguenther@suse.de>
1920 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1923 2018-04-27 Richard Biener <rguenther@suse.de>
1925 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1927 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1929 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1930 with Yd constraint. Set "preferred_for_speed" attribute from
1931 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1933 (*movdi_internal): Ditto.
1934 (movti_interunit splitters): Remove
1935 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1936 (movdi_interunit splitters): Ditto.
1937 * config/i386/constraints.md (Ye): Remove.
1938 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1940 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1943 * config/aarch64/constraints.md (Usg): Limit to 31.
1946 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1948 PR tree-optimization/85529
1949 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1950 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1951 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1952 zero extension or masking of the MSB bit.
1953 (optimize_range_tests): Add FIRST_BB argument, pass it through
1954 to optimize_range_tests_var_bound.
1955 (maybe_optimize_range_tests, reassociate_bb): Adjust
1956 optimize_range_tests callers.
1958 2018-04-26 Richard Biener <rguenther@suse.de>
1959 Jakub Jelinek <jakub@redhat.com>
1961 * cgraph.h (symbol_table): Just declare debug method here.
1962 * symtab.c (symbol_table::debug): Define.
1964 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1966 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1968 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1970 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1971 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1972 (*movdi_internal): Substitute Yi and Yj constraint with x
1973 and Ym and Yn constraint with y constraint. Update "isa"
1974 attribute and set "preferred_for_speed" attribute from
1975 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1976 (*movsi_internal): Ditto.
1977 (*movdf_internal): Ditto.
1978 (*movsf_internal): Ditto.
1979 (*zero_extendsidi2): Ditto.
1980 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1981 (sse2_loadld): Ditto.
1982 (*vec_extract<ssevecmodelower>_0): Ditto.
1983 (*vec_extractv4si_0_zext_sse4): Ditto.
1984 (vec_concatv2di): Ditto.
1985 (*vec_dup<mode>): Ditto.
1986 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1987 * config/i386/constraints.md (Yi): Remove.
1992 2018-04-26 Nathan Sidwell <nathan@acm.org>
1994 * dumpfile.c (dump_open): New.
1995 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1996 (dump_finish): Detect stdio/stderr by value not name.
1998 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2000 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2002 2018-04-26 Tom de Vries <tom@codesourcery.com>
2005 * config/nvptx/nvptx.c (verify_neutering_jumps)
2006 (verify_neutering_labels): New function
2007 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2009 2018-04-26 Tom de Vries <tom@codesourcery.com>
2012 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2013 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2016 2018-04-26 Richard Biener <rguenther@suse.de>
2017 Tom de Vries <tom@codesourcery.com>
2020 * lto-streamer-out.c (output_function): Fixup loops if required to match
2021 discovery done in the reader.
2023 2018-04-26 Richard Biener <rguenther@suse.de>
2025 PR tree-optimization/85116
2026 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2027 have a loop exit from the single latch predecessor. Remove
2028 case of header with just condition.
2029 (ch_base::copy_headers): Exclude infinite loops from any
2031 (pass_ch::execute): Record exits.
2033 2018-04-26 Richard Biener <rguenther@suse.de>
2035 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2036 prologue cost vector and pass it to vect_get_load_cost.
2037 (vect_get_peeling_costs_all_drs): Likewise.
2038 (vect_peeling_hash_get_lowest_cost): Likewise.
2039 (vect_enhance_data_refs_alignment): Likewise.
2041 2018-04-26 Richard Biener <rguenther@suse.de>
2044 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2045 checking of integer<->pointer conversions.
2046 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2047 sign-/zero-extending pointer types.
2048 (expand_omp_for_static_chunk): Likewise.
2050 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2051 Jean Lee <xiaoyur347@gmail.com>
2053 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2054 (TARGET_ASAN_SHADOW_OFFSET): Define.
2055 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2056 true for -fsanitize=address.
2058 2018-04-25 Mark Wielaard <mark@klomp.org>
2060 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2063 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2065 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2066 than "alu", remove explicit "memory" and "imm_disp" attributes.
2067 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2070 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2071 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2072 gen_lowpart_no_emit.
2074 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2077 * config/i386/i386.c (ix86_expand_builtin): Change memory
2078 operand to XI, extend p0 to Pmode.
2079 * config/i386/i386.md: Change unspec volatile and operand
2080 1 mode to XI, change operand 0 mode to P.
2082 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2084 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2085 GET_MODE_MASK before any checking.
2086 (nds32_can_use_bset_p): Likewise.
2087 (nds32_can_use_btgl_p): Likewise.
2089 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2091 * config/nds32/nds32-doubleword.md: New define_split pattern for
2092 illegal register number.
2094 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2096 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2098 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2100 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2102 2018-04-25 Richard Biener <rguenther@suse.de>
2104 * lto-streamer.h (LTO_major_version): Bump to 8.
2106 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2108 * BASE-VER: Set to 9.0.0.
2110 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2112 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2113 in __abskf2 and __powikf2.
2115 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2118 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2119 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2120 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2121 Use the above on operand 2. Reindent.
2122 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2124 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2127 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2129 * config/i386/i386.opt (mcet): Removed.
2130 * doc/install.texi: Remove -mcet documentation.
2131 * doc/invoke.texi: Likewise.
2133 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2136 * doc/install.texi: Remove -mcet from bootstrap-cet.
2138 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2141 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2142 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2146 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2147 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2148 containing a CONST_VECTOR.
2150 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2152 * doc/install.texi: Update newlib dependency for nvptx.
2154 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2157 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2158 instead of INTVAL when shifting x left.
2160 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2162 PR tree-optimization/85478
2163 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2164 vect_grouped_store_supported for single element vectors.
2166 2018-04-24 Richard Biener <rguenther@suse.de>
2169 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2170 load cost increase to the case of non-constant step.
2172 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2175 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2176 destination if any_malformed_asm.
2178 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2181 * expr.c (store_field): In the bitfield case, if the value comes from
2182 a function call and is returned in registers by means of a PARALLEL,
2183 do not change the mode of the temporary unless BLKmode and VOIDmode.
2185 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2187 PR rtl-optimization/85423
2188 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2189 dependencies to debug insns when the previous insn is non-debug.
2191 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2193 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2194 enums into a single definition.
2195 (fls): Fix predicates and printing.
2198 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2200 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2201 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2202 and short u6 immediate.
2203 (check_if_valid_sleep_operand): Remove.
2204 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2206 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2208 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2209 flag_always_save_lp condition.
2210 * config/nds32/nds32.opt (malways-save-lp): New option.
2212 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2214 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2215 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2216 * config/nds32/nds32.h
2217 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2218 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2220 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2222 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2223 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2225 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2226 Chung-Ju Wu <jasonwucj@gmail.com>
2228 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2230 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2231 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2233 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2235 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2237 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2239 * config/nds32/nds32-protos.h (nds32_data_alignment,
2240 nds32_local_alignment): Declare.
2241 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2242 nds32_local_alignment): New functions.
2243 (TARGET_CONSTANT_ALIGNMENT): Define.
2244 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2246 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2248 * config/nds32/nds32.c
2249 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2250 (TARGET_MODES_TIEABLE_P): Likewise.
2252 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2254 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2257 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2258 Chung-Ju Wu <jasonwucj@gmail.com>
2260 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2261 for unaligned access.
2262 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2264 * config/nds32/nds32-intrinsic.md: Likewise.
2265 * config/nds32/nds32_intrinsic.h: Likewise.
2266 * config/nds32/nds32.h (nds32_builtins): Likewise.
2267 * config/nds32/nds32.opt (munaligned-access): New option.
2268 * config/nds32/nds32.c (nds32_asm_file_start): Display
2269 flag_unaligned_access status.
2271 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2273 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2274 -mno-relax is present.
2275 * config/riscv/linux.h (LINK_SPEC): Ditto.
2277 2018-04-20 Martin Sebor <msebor@redhat.com>
2280 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2282 (gimple_fold_builtin_stxcpy_chk): Same.
2283 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2285 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2288 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2289 __powikf2 when long double is IEEE 128-bit.
2291 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2293 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2294 step to make sure stack always aligned.
2296 2018-04-20 Carl Love <cel@us.ibm.com>
2299 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2300 size check for arg0.
2302 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2303 Tom de Vries <tom@codesourcery.com>
2306 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2307 Emit insns for calls too.
2308 (nvptx_find_par): Always look for worker-level predecessor insn.
2309 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2311 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2312 (nvptx_process_pars): Propagate frames for calls.
2314 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2317 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2319 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2320 (ix86_handle_option): Don't handle OPT_mibt.
2321 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2323 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2325 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2326 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2327 (ix86_target_macros): Define __CET__ with flag_cf_protection
2328 for -fcf-protection.
2329 * config/i386/i386.c (isa2_opts): Remove -mibt.
2330 * config/i386/i386.h (TARGET_IBT): Removed.
2331 (TARGET_IBT_P): Likewise.
2332 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2333 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2334 * config/i386/i386.opt (mcet): Update help message.
2337 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2338 -mcet as an alias for -mshstk.
2340 2018-04-20 Richard Biener <rguenther@suse.de>
2343 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2344 complexity by forcing a single use of the multiply operand.
2346 2018-04-20 Martin Jambor <mjambor@suse.cz>
2349 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2350 recursion dependency to only apply to non-clones.
2352 2018-04-20 Martin Jambor <mjambor@suse.cz>
2355 * ipa-cp.c (create_specialized_node): Check that clones of
2356 self-recursive edges exist during IPA-CP.
2358 2018-04-19 Toon Moene <toon@moene.org>
2360 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2363 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2365 PR tree-optimization/85467
2366 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2367 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2368 VECTOR_CST element to type.
2370 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2373 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2374 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2375 (builtin_longjmp): Likewise.
2376 (save_stack_nonlocal): New pattern.
2377 (restore_stack_nonlocal): Likewise.
2379 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2382 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2383 Replace ASM_OUTPUT_LABEL with fprintf.
2385 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2388 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2389 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2390 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2391 define __IBT__ and __SHSTK__ for -fcf-protection.
2392 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2394 (ix86_trampoline_init): Likewise.
2395 (x86_output_mi_thunk): Likewise.
2396 (ix86_notrack_prefixed_insn_p): Likewise.
2397 (ix86_option_override_internal): Don't disallow -fcf-protection.
2398 * config/i386/i386.md (rdssp<mode>): Also enable for
2400 (incssp<mode>): Likewise.
2401 (nop_endbr): Likewise.
2402 * config/i386/i386.opt (mcet): Change help message to built-in
2406 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2407 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2408 enable CET built-in functions.
2410 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2412 * common/config/i386/i386-common.c
2413 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2414 OPTION_MASK_ISA_MOVDIRI_UNSET,
2415 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2416 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2417 * config.gcc (movdirintrin.h): New header.
2418 * config/i386/cpuid.h (bit_MOVDIRI,
2419 bit_MOVDIR64B): New bits.
2420 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2422 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2423 (VOID, PVOID, PCVOID)): New function types.
2424 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2425 __builtin_ia32_directstoreu_u64,
2426 __builtin_ia32_movdir64b): New builtins.
2427 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2428 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2430 (ix86_valid_target_attribute_inner_p): Ditto.
2431 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2432 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2433 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2434 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2435 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2436 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2437 (movdiri<mode>, movdir64b_<mode>): New.
2438 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2439 * config/i386/immintrin.h: Include movdirintrin.h.
2440 * config/i386/movdirintrin.h: New file.
2441 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2443 2018-04-19 Richard Biener <rguenther@suse.de>
2446 * cfg.c (clear_bb_flags): When loop state says we have
2447 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2449 2018-04-19 Richard Biener <rguenther@suse.de>
2451 PR tree-optimization/84737
2452 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2453 copying restrict info.
2454 (vect_setup_realignment): Use it.
2455 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2456 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2457 the first DR to all generated stores.
2458 (vectorizable_load): Likewise for loads.
2460 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2462 PR tree-optimization/85446
2463 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2464 the integral and pointer types to have the same precision.
2466 * doc/install.texi: Document --disable-cet being the default and
2469 2018-04-18 Martin Liska <mliska@suse.cz>
2471 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2474 2018-04-18 Martin Liska <mliska@suse.cz>
2477 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2480 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2481 arguments if they are comparable.
2483 2018-04-18 Martin Liska <mliska@suse.cz>
2486 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2489 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2492 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2495 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2496 ENDBR after calling __morestack.
2498 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2501 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2502 by using gcc_base_ver to generate a gcc_driver_version, and use
2503 it when generating GCC_DRIVER_NAME.
2504 * configure: Regenerate.
2506 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2509 * config.gcc: Obsolete powerpc*-*-*spe*.
2511 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2514 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2515 (stabstr_D): Change type of unum from unsigned int to
2516 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2519 2018-04-17 Jim Wilson <jimw@sifive.com>
2522 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2523 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2524 Set arg_pointer_offset after using pretend_args_size.
2526 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2528 PR rtl-optimization/85431
2529 * dse.c (record_store): Ignore zero width stores.
2532 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2533 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2534 __builtin_stack_restore rather than after it.
2535 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2536 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2537 argument instead of virtual_dynamic_stack_rtx.
2539 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2541 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2543 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2544 Add note to error message to explain internal mapping of overloaded
2545 built-in function name to non-overloaded built-in function name.
2546 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2549 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2552 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2553 where the inputs overlap with the output.
2555 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2557 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2558 (=v, v) alternative and explicit "memory" attribute.
2559 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2560 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2562 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2563 "sselog1" type instead of "sselog".
2564 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2565 "sselog". Remove explicit "memory" attribute.
2566 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2567 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2569 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2570 "sselog1" type instead of "sselog". Remove explicit "memory"
2572 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2573 use "sselog1" type instead of "sselog". Remove explicit "memory"
2575 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2576 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2578 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2579 "sselog1" type instead of "sselog". Remove explicit "memory"
2581 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2582 use "sselog1" type instead of "sselog". Remove explicit "memory"
2586 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2589 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2592 2018-04-17 Martin Jambor <mjambor@suse.cz>
2595 * ipa-cp.c (create_specialized_node): Call
2596 expand_all_artificial_thunks if necessary.
2598 2018-04-17 Martin Liska <mliska@suse.cz>
2601 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2602 in message, remote space in between '_G' and '('.
2604 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2607 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2608 avx512f_vmcmp<mode>3<round_saeonly_name>,
2609 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2610 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2611 avx512f_rndscale<mode><round_saeonly_name>,
2612 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2613 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2614 Use %<iptr>2 instead of %2 for -masm=intel.
2615 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2616 avx512f_vcvttss2usi<round_saeonly_name>,
2617 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2619 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2620 avx512f_vcvttsd2usi<round_saeonly_name>,
2621 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2622 Use %q1 instead of %1 for -masm=intel.
2623 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2624 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2625 of %3 for -masm=intel.
2626 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2628 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2630 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2632 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2633 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2635 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2637 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2638 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2639 %g1 and one with %0 and %1.
2640 (avx512er_vmrcp28<mode><round_saeonly_name>,
2641 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2643 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2644 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2645 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2646 of %0 and %{%4%} for -masm=intel.
2647 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2648 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2649 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2650 order of %0 and %{%5%}%{z%} for -masm=intel.
2652 2018-04-17 Jan Hubicka <jh@suse.cz>
2655 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2657 2018-04-17 Martin Liska <mliska@suse.cz>
2660 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2661 for target_clone error message. Make default implementation
2662 clone to be a local declaration.
2663 (separate_attrs): Add new argument and check for an empty
2665 (expand_target_clones): Handle it.
2666 (ipa_target_clone): Make redirection just for target_clones
2669 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2670 Tom de Vries <tom@codesourcery.com>
2673 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2674 tiled basic blocks without omp continue statements.
2676 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2679 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2680 vec_extract expression as having side effects to make sure it gets
2683 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2686 * config/i386/i386.c (get_builtin_code_for_version): Check
2689 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2692 * gcc/config.gcc: Support "skylake".
2693 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2695 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2696 (processor_target_table): Add "skylake".
2697 (ix86_option_override_internal): Add "skylake".
2698 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2699 PROCESSOR_CANNONLAKE.
2700 (get_builtin_code_for_version): Fix priority for
2701 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2702 PROCESSOR_SKYLAKE-AVX512.
2703 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2704 (processor_type): Add PROCESSOR_SKYLAKE.
2706 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2707 Jason Merrill <jason@redhat.com>
2710 * convert.c (convert_to_integer_1): Use direct recursion for
2711 enumeral types and types with a precision less than the number
2712 of bits in their mode.
2714 2018-04-16 Julia Koval <julia.koval@intel.com>
2717 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2718 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2720 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2723 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2724 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2725 and -mno-direct-move.
2727 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2730 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2731 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2732 Ensure negative shifts result in {0}.
2734 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2736 PR rtl-optimization/79916
2737 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2738 regs (if any) to define how to gnerate SD moves when LRA is in
2741 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2743 PR rtl-optimization/85393
2744 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2745 * except.c (expand_dw2_landing_pad_for_region): Make static.
2746 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2747 a label and unconditional jump to old_bb, rather than
2748 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2751 PR rtl-optimization/85376
2752 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2753 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2754 instead of a specific value.
2756 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2757 Bin Cheng <bin.cheng@arm.com>
2759 PR tree-optimization/82965
2760 PR tree-optimization/83991
2761 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2762 by_profile_only parameter.
2763 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2764 information if the loop was predicted to iterate too many times.
2765 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2767 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2770 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2773 2018-04-13 Martin Liska <mliska@suse.cz>
2774 Jakub Jelinek <jakub@redhat.com>
2777 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2778 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2779 * builtins.c (expand_builtin_memory_copy_args): Use
2780 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2781 handle dest_addr == pc_rtx.
2783 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2786 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2787 asked to not generate direct moves.
2788 (fix_trunc<mode>si2_stfiwx): Similar.
2789 (fix_trunc<mode>si2_internal): Similar.
2791 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2794 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2795 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2796 lookup if dest in some wider mode is known to be const0_rtx and
2797 if so, record permanent equivalence for it to be ZERO_EXTEND of
2798 the narrower mode destination.
2800 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2802 * lto-streamer-out.c (output_function): Revert 259346.
2803 * omp-expand.c (expand_oacc_for): Likewise.
2805 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2807 PR rtl-optimization/85354
2808 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2809 * sel-sched.c (sel_global_init): ... here.
2811 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2814 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2815 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2816 mode for PE-COFF targets.
2817 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2818 (i386_pe_asm_lto_end): Likewise.
2819 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2820 (TARGET_ASM_LTO_END): Likewise.
2821 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2822 (i386_pe_asm_lto_start): New function.
2823 (i386_pe_asm_lto_end): Likewise.
2825 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2826 Richard Biener <rguenther@suse.de>
2829 * lto-streamer-out.c (output_function): Fix CFG loop state before
2831 * omp-expand.c (expand_oacc_for): Handle calls to internal
2832 functions like regular functions.
2834 2018-04-12 Richard Biener <rguenther@suse.de>
2837 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2838 for the early LTO debug to properly generate references to it
2839 during DIE emission. Do not re-use that for the skeleton for
2841 (dwarf2out_early_finish): Likewise.
2843 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2846 * config/i386/sse.md
2847 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2848 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2849 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2850 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2851 and output is a reg, avoid creating invalid lowpart subreg, but
2852 instead split into a 512-bit move. Don't split if not AVX512VL,
2853 input is xmm16+ reg and output is a mem.
2854 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2855 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2856 xmm16+ reg and output is a mem.
2858 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2860 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2861 also for flag_dwarf2_cfi_asm.
2863 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2865 PR rtl-optimization/85342
2866 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2867 a bool scalar var inside of the loop instead. Don't try to update
2868 recog_data.operand after failed apply_change_group.
2870 2018-04-12 Tom de Vries <tom@codesourcery.com>
2873 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2874 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2875 array with flexible array member as array without given dimension.
2876 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2877 argument for undefined param to true.
2879 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2882 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2883 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2884 from PowerPC section.
2885 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2886 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2887 help text that is too long.
2888 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2889 help text that is too long.
2890 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2891 help text that is too long.
2893 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2895 * config/alpha/alpha.md (stack_probe_internal): Rename
2896 from "probe_stack". Update all callers.
2898 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2900 PR rtl-optimization/84566
2901 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2902 sched_macro_fuse_insns.
2904 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2907 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2908 (compute_block_dependences): ... from here.
2910 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2912 PR tree-optimization/85331
2913 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2914 from int to HOST_WIDE_INT.
2916 2018-04-11 Martin Jambor <mjambor@suse.cz>
2919 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2920 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2921 not the same as the source val.
2922 (cgraph_edge_brings_value_p): New parameter.
2923 (gather_edges_for_value): Pass destination value to
2924 cgraph_edge_brings_value_p.
2925 (perhaps_add_new_callers): Likewise.
2926 (get_info_about_necessary_edges): Likewise and exclude values brought
2927 only by self-recursive edges.
2928 (create_specialized_node): Redirect only clones of self-calling edges.
2929 (+self_recursive_pass_through_p): New function.
2930 (find_more_scalar_values_for_callers_subset): Use it.
2931 (find_aggregate_values_for_callers_subset): Likewise.
2932 (known_aggs_to_agg_replacement_list): Removed.
2933 (decide_whether_version_node): Re-calculate known constants for all
2934 remaining context clones.
2936 2018-04-11 Richard Biener <rguenther@suse.de>
2939 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2940 from early DWARF output.
2941 (dwarf2out_early_finish): Output line info unconditionally into
2942 early DWARF and add reference to it.
2944 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2947 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2948 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2949 other than V2DFmode using iptr mode attribute.
2950 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2952 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2954 PR rtl-optimization/84659
2955 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2957 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2960 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2962 (output_loc_list): Pass address of a dummy size variable even in the
2963 locview handling loop.
2964 (index_location_lists): Add comment on why skip_loc_list_entry can't
2967 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2970 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2973 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2976 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2977 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2978 and -mstring-compare-inline-limit.
2980 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2983 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2984 for stack clash protection in a register whenever we need it to be in
2987 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2989 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2990 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2992 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2995 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2997 (mlong-double-): Ditto.
2998 * config/rs6000/sysv4.opt (msdata=): Ditto.
2999 (mtls-size=): Ditto.
3001 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3003 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3004 erroneous entries for
3005 "vector int vec_ldl (int, long int *)", and
3006 "vector unsigned int vec_ldl (int, unsigned long int *)".
3007 Add comments and entries for
3008 "vector bool char vec_ldl (int, bool char *)",
3009 "vector bool short vec_ldl (int, bool short *)",
3010 "vector bool int vec_ldl (int, bool int *)",
3011 "vector bool long long vec_ldl (int, bool long long *)",
3012 "vector pixel vec_ldl (int, pixel *)",
3013 "vector long long vec_ldl (int, long long *)",
3014 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3015 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3016 type tree bool_long_long_type_node and correct definition of
3017 bool_V2DI_type_node to make reference to this new type tree.
3018 (rs6000_mangle_type): Replace erroneous reference to
3019 bool_long_type_node with bool_long_long_type_node.
3020 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3021 comments to emphasize sign distinctions for char and int types and
3022 replace RS6000_BTI_bool_long constant with
3023 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3024 use of RS6000_BTI_pixel.
3025 (bool_long_type_node): Remove this macro definition.
3026 (bool_long_long_type_node): New macro definition
3028 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3030 PR rtl-optimization/85300
3031 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3032 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3033 simplify_unary_operation fails.
3035 2018-04-10 Martin Liska <mliska@suse.cz>
3037 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3038 cgraph_edge and ipa_ref.
3040 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3044 * config/i386/sse.md
3045 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3046 computation of the VEC_MERGE selector from mask.
3047 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3048 Fix decoding of the VEC_MERGE selector into mask.
3050 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3052 PR tree-optimization/85286
3053 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3055 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3057 * final.c (final_1): Set insn_last_address as well as
3058 insn_current_address.
3060 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3063 * explow.c (emit_stack_probe): Call validize_mem on memory location
3064 before passing it to gen_probe_stack. Create address operand and
3065 legitimize it for the probe_stack_address case.
3067 2018-04-09 Jan Hubicka <jh@suse.cz>
3070 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3071 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3072 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3074 (free_lang_data): Rebuild type inheritance graph.
3076 2018-04-09 Martin Sebor <msebor@redhat.com>
3078 * invoke.texi (-finline-small-functions): Mention other optimization
3080 (-findirect-inlining, -fpartial-inlining): Same.
3081 (-finline-functions-called-once): Same.
3082 (-freorder-blocks-and-partition): Same.
3084 2018-04-09 Jan Hubicka <jh@suse.cz>
3087 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3088 jumps; choose last target that matches the criteria (i.e.
3089 no partition changes for non-crossing jumps).
3090 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3091 support for redirecting crossing jumps to non-crossing.
3093 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3095 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3096 also for naked functions.
3098 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3100 * config/arc/arc.md (add_shift): New pattern.
3101 (add_shift2): Likewise.
3102 (sub_shift): Likewise.
3103 (sub_shift_cmp0_noout): Likewise.
3104 (compare_si_ashiftsi): Likewise.
3105 (xbfu_cmp0_noout): New combine pattern.
3106 (xbfu_cmp0"): Likewise.
3107 (movsi_set_cc_insn): Place the predicable variant first.
3108 (commutative_binary_cmp0_noout): Remove clobber.
3109 (commutative_binary_cmp0): New pattern.
3110 (noncommutative_binary_cmp0): Likewise.
3111 (noncommutative_binary_cmp0_noout): Likewise.
3112 (noncommutative_binary_comparison_result_used): Removed.
3113 (rsub_cmp0): New pattern.
3114 (rsub_cmp0_noout): Likewise.
3115 (extzvsi): Changed, keep only meaningful variants.
3116 (SQH, SEZ): New iterators.
3117 (SQH_postfix): New mode attribute.
3118 (SEZ_prefix): New code attribute.
3119 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3120 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3121 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3123 (noncommutative_operator): Check the availability of barrel
3126 2018-04-09 Richard Biener <rguenther@suse.de>
3128 PR tree-optimization/85284
3129 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3130 Only use the niter constraining form of simple_iv when the exit
3133 2018-04-09 Tom de Vries <tom@codesourcery.com>
3136 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3137 (define_expand "*memory_barrier"): New define_expand.
3138 (define_insn "memory_barrier"): New insn.
3140 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3142 PR rtl-optimization/80463
3143 PR rtl-optimization/83972
3144 PR rtl-optimization/83480
3146 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3147 correct producer for the insn.
3148 (tidy_control_flow): Fixup seqnos in case of debug insns.
3150 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3152 PR rtl-optimization/83913
3154 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3155 different sched-times when merging exprs.
3157 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3159 PR rtl-optimization/83962
3161 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3162 tidy_fallthru_edge and tidy_control_flow.
3164 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3166 PR rtl-optimization/83530
3168 * sel-sched.c (force_next_insn): New global variable.
3169 (remove_insn_for_debug): When force_next_insn is true, also leave only
3170 next insn in the ready list.
3171 (sel_sched_region): When the region wasn't scheduled, make another pass
3172 over it with force_next_insn set to 1.
3174 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3176 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3178 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3179 for interrupt control.
3180 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3181 functions for interrupt control.
3182 * config/nds32/nds32-intrinsic.md: Likewise.
3183 * config/nds32/nds32_intrinsic.h: Likewise.
3184 * config/nds32/nds32.h (nds32_builtins): Likewise.
3186 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3188 * config/nds32/nds32.c (nds32_init_machine_status,
3189 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3190 strict_aligned_p field.
3191 (nds32_expand_to_rtl_hook): New function.
3192 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3193 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3195 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3196 Chung-Ju Wu <jasonwucj@gmail.com>
3198 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3199 * config/nds32/nds32-n7.md: New file.
3200 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3201 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3203 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3204 * config/nds32/nds32.md (pipeline_model): Add n7.
3205 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3206 * config/nds32/pipelines.md: Include n7 settings.
3208 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3209 Chung-Ju Wu <jasonwucj@gmail.com>
3211 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3212 * config/nds32/nds32-e8.md: New file.
3213 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3214 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3216 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3217 * config/nds32/nds32.md (pipeline_model): Add e8.
3218 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3219 * config/nds32/pipelines.md: Include e8 settings.
3221 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3222 Chung-Ju Wu <jasonwucj@gmail.com>
3224 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3225 * config/nds32/nds32-n8.md: New file.
3226 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3227 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3229 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3230 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3231 * config/nds32/nds32.md (pipeline_model): Add n8.
3232 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3233 * config/nds32/pipelines.md: Include n8 settings.
3235 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3236 Chung-Ju Wu <jasonwucj@gmail.com>
3238 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3239 * config/nds32/nds32-n9-2r1w.md: New file.
3240 * config/nds32/nds32-n9-3r2w.md: New file.
3241 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3242 nds32_register_ports): New or modify for cpu n9.
3243 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3245 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3246 * config/nds32/nds32-utils.c: New file.
3247 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3248 TARGET_MUL_SLOW): Define.
3249 * config/nds32/nds32.md (pipeline_model): New attribute.
3250 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3251 New options that support cpu n9.
3252 * config/nds32/pipelines.md: Include n9 settings.
3253 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3255 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3257 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3258 information if necessary.
3259 (output_cond_branch_compare_zero): Likewise.
3260 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3261 (nds32_target_alignment): Refine for alignment.
3262 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3263 (FUNCTION_BOUNDARY): Modify.
3264 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3266 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3268 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3270 * config/nds32/constants.md (unspec_volatile_element): Add values for
3271 TLB operation and data prefetch.
3272 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3273 functions for TLB operation and data prefetch.
3274 * config/nds32/nds32-intrinsic.md: Likewise.
3275 * config/nds32/nds32_intrinsic.h: Likewise.
3276 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3277 (nds32_print_operand): Likewise.
3278 * config/nds32/nds32.h (nds32_builtins): Likewise.
3280 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3281 Andrew Pinski <pinsika@gcc.gnu.org>
3284 * match.pd: Use constant_boolean_node of correct type instead of
3285 boolean_true_node or boolean_false_node for simplifying
3286 pointer comparisons to zero.
3288 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3290 PR tree-optimization/80021
3291 * tree.c (verify_type_variant): Make error call in verify_variant_match
3292 translatable and remove final full stop.
3294 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3296 * config/nds32/constants.md (unspec_volatile_element): Add
3297 UNSPEC_VOLATILE_EH_RETURN.
3298 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3299 nds32_output_stack_pop): Support dwarf exception handling process.
3300 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3301 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3302 exception handling process.
3303 (nds32_compute_stack_frame): Likewise.
3304 (nds32_return_addr_rtx): Likewise.
3305 (nds32_initial_elimination_offset): Likewise.
3306 (nds32_expand_prologue): Likewise.
3307 (nds32_expand_epilogue): Likewise.
3308 (nds32_dynamic_chain_address): New function.
3309 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3311 (DYNAMIC_CHAIN_ADDRESS): Define.
3312 (EH_RETURN_DATA_REGNO): Define.
3313 (EH_RETURN_STACKADJ_RTX): Define.
3314 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3315 patterns for dwarf exception handling.
3317 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3319 * config/nds32/nds32.h: Clean up obsolete macros.
3321 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3323 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3324 Add enum values for particular instructions.
3325 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3326 particular intrinsic functions.
3327 * config/nds32/nds32-intrinsic.md: Likewise.
3328 * config/nds32/nds32_intrinsic.h: Likewise.
3329 * config/nds32/nds32.h (nds32_builtins): Likewise.
3330 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3331 (btst, ave): New patterns for particular instructions.
3333 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3335 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3336 Add enum values for atomic load/store and memory sync.
3337 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3339 * config/nds32/nds32-intrinsic.md: Likewise.
3340 * config/nds32/nds32_intrinsic.h: Likewise.
3341 * config/nds32/nds32.h (nds32_builtins): Likewise.
3343 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3345 PR tree-optimization/85257
3346 * fold-const.c (native_encode_vector): If not all elts could fit
3347 and off is -1, return 0 rather than offset.
3348 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3349 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3350 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3351 adjust buffer in native_interpret_expr call.
3353 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3355 * config/nds32/constants.md (unspec_volatile_element): Add cache
3356 control enum values.
3357 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3358 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3359 * config/nds32/nds32.c (nds32_cctl_names): New.
3360 (nds32_print_operand): Handle cache control register names.
3361 * config/nds32/nds32.h (nds32_builtins): New enum values.
3362 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3364 * config/nds32/nds32.md (type): Add mmu.
3365 * config/nds32/pipelines.md (simple_insn): Add mmu.
3367 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3369 * config/nds32/nds32.md (type): Remove call.
3370 * config/nds32/pipelines.md (simple_insn): Likewise.
3372 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3374 * config/nds32/constants.md (unspec_volatile_element): Add
3375 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3376 UNSPEC_VOLATILE_FMFCFG.
3377 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3378 description for fmfcfg and fmfcsr.
3379 (bdesc_1arg): Add fmtcsr.
3380 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3381 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3382 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3383 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3384 unspec_fmfcfg): New patterns.
3385 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3386 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3387 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3388 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3389 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3390 __nds32__fmfcfg): Define.
3392 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3394 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3395 intrinsic register names.
3396 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3397 intrinsic register enum values and macros.
3399 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3401 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3402 for load/store addressing form.
3403 (nds32_print_operand_address): Likewise.
3405 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3408 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3409 based on LABEL_REF. Remove useless assertion.
3410 (pic_address_needs_scratch): Fix formatting.
3411 (sparc_legitimize_pic_address): Minor tweaks.
3412 (sparc_delegitimize_address): Adjust assertion accordingly.
3413 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3414 into symbolic_operand.
3415 (movsi_high_pic_label_ref): Likewise.
3416 (movsi_lo_sum_pic_label_ref): Likewise.
3417 (movdi_pic_label_ref): Likewise.
3418 (movdi_high_pic_label_ref): Likewise.
3419 (movdi_lo_sum_pic_label_ref): Likewise.
3421 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3423 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3424 custom LIB_SPEC setup.
3426 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3427 Kito Cheng <kito.cheng@gmail.com>
3429 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3430 * config/riscv/freebsd.h: New.
3432 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3434 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3435 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3438 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3439 Kito Cheng <kito.cheng@gmail.com>
3441 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3442 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3443 * config/nds32/nds32-protos.h (nds32_output_call,
3444 nds32_output_return): Declare.
3445 * config/nds32/nds32.md: Refine all the call and return patterns.
3447 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3450 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3451 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3453 PR rtl-optimization/84872
3454 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3455 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3458 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3460 * expr.c (copy_blkmode_to_reg): Revert 254862.
3461 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3463 2018-04-06 Richard Biener <rguenther@suse.de>
3466 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3467 after seeing a component reference with an adjacent field. Treat
3468 refs to arrays at struct end of external decls similar to
3469 refs to unconstrained commons.
3471 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3474 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3475 look through SAVE_EXPRs with non-side-effects argument. Adjust
3477 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3480 2018-04-06 Richard Biener <rguenther@suse.de>
3483 * alias.c (find_base_term): New wrapper around find_base_term
3484 unwinding CSELIB_VAL_PTR changes.
3485 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3488 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3490 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3492 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3493 constant definitions.
3494 ("nop"): lr 0,0 -> nopr r0
3495 ("nop_lr0", "nop_lr1"): New insn definitions.
3497 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3499 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3500 NDS32_V3PUSH_AVAILABLE_P macro.
3502 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3503 Chung-Ju Wu <jasonwucj@gmail.com>
3505 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3506 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3507 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3508 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3509 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3510 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3511 * config/nds32/constraints.md: New constraints and checking for hard
3512 float configuration.
3513 * config/nds32/iterators.md: New mode iterator and attribute for hard
3514 float configuration.
3515 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3517 * config/nds32/nds32-fpu.md: New file.
3518 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3519 deal with hard float code generation.
3520 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3522 (abi_type, float_reg_number): New enum type.
3523 * config/nds32/nds32-predicates.c: New predicates for hard float.
3524 * config/nds32/nds32-protos.h: Declare functions for hard float.
3525 * config/nds32/nds32.c: Implementation for hard float configuration.
3526 * config/nds32/nds32.h: Definitions for hard float configuration.
3527 * config/nds32/nds32.md: Include hard float machine description and
3528 modify patterns for hard float configuration.
3529 * config/nds32/nds32.opt: New options for hard float configuration.
3530 * config/nds32/predicates.md: New predicates for hard float
3533 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3535 * common/config/nds32/nds32-common.c
3536 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3538 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3541 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3542 CONSTRUCTOR_ELT (ctor, ...)->value.
3544 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3547 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3549 2018-04-05 Tom de Vries <tom@codesourcery.com>
3552 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3555 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3556 Kito Cheng <kito.cheng@gmail.com>
3558 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3559 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3560 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3561 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3563 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3564 Kito Cheng <kito.cheng@gmail.com>
3566 * config/nds32/constraints.md (Ufe): New memory constraint.
3567 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3568 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3569 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3571 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3572 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3574 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3576 * config/nds32/nds32.md: Use optimize_size in the condition for
3577 alu-shift instructions.
3579 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3581 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3583 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3585 * config/nds32/nds32.md (negsi2): Refine pattern.
3587 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3588 Chung-Ju Wu <jasonwucj@gmail.com>
3590 * config/nds32/iterators.md (shift_rotate): New code iterator.
3591 (shift): New code attribute.
3592 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3593 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3594 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3595 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3596 bit-wise operations.
3597 (andsi3, *andsi3): Ditto.
3598 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3599 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3600 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3601 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3602 nds32_ior_operand, nds32_xor_operand): New predicates.
3604 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3606 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3607 (addsi3, subsi3): ... this.
3609 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3611 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3613 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3615 * config/nds32/nds32.md: Adjust indention.
3617 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3619 * config/nds32/nds32.md (feature): New attribute.
3621 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3623 * config/nds32/nds32.md (subtype): New attribute.
3625 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3628 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3629 expansion to perform a bitwise AND of the argument followed by a
3630 boolean negation of the result.
3632 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3634 PR rtl-optimization/84878
3635 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3636 the basic block. Assert the use reference is not artificial and that
3637 it has an associated insn.
3639 2018-04-04 Michael Matz <matz@suse.de>
3641 * builtins.c (compute_objsize): Pass correct operand
3642 to array_at_struct_end_p.
3644 2018-04-04 Richard Biener <rguenther@suse.de>
3647 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3648 from contexts for DINFO_LEVEL_TERSE and below.
3650 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3652 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3653 resiter_operand condition.
3654 * config/nds32/nds32.md (*move<mode>): Ditto.
3656 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3657 Monk Chiang <sh.chiang04@gmail.com>
3659 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3661 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3663 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3665 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3666 Kito Cheng <kito.cheng@gmail.com>
3668 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3669 nds32_cond_code_str, output_cond_branch,
3670 output_cond_branch_compare_zero, nds32_expand_cbranch,
3671 nds32_expand_cstore, nds32_expand_movcc,
3672 nds32_output_cbranchsi4_equality_zero,
3673 nds32_output_cbranchsi4_equality_reg,
3674 nds32_output_cbranchsi4_equality_reg_or_const_int,
3675 nds32_output_cbranchsi4_greater_less_zero: New functions.
3676 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3677 nds32_expand_cstore, nds32_expand_movcc,
3678 nds32_output_cbranchsi4_equality_zero,
3679 nds32_output_cbranchsi4_equality_reg,
3680 nds32_output_cbranchsi4_equality_reg_or_const_int,
3681 nds32_output_cbranchsi4_greater_less_zero): Declare.
3682 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3683 nds32_rimm11s_operand): New predicates.
3684 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3685 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3688 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3690 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3691 * config/nds32/nds32.md: Ditto.
3692 * config/nds32/pipelines.md: Ditto.
3694 2018-04-04 Richard Biener <rguenther@suse.de>
3696 PR tree-optimization/85168
3697 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3698 propagating abnormals.
3700 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3702 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3704 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3705 Kito Cheng <kito.cheng@gmail.com>
3707 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3708 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3709 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3710 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3711 * config/nds32/nds32.md (sibcall_internal): New.
3712 (sibcall_register): Remove.
3713 (sibcall_immediate): Remove.
3714 (sibcall_value_internal): New.
3715 (sibcall_value_register): Remove.
3716 (sibcall_value_immediate): Remove.
3717 * config/nds32/predicates.md (nds32_general_register_operand): New.
3718 (nds32_call_address_operand): New.
3720 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3722 PR rtl-optimization/85167
3723 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3724 bb_defs if *split_p, instead preinitialize it to NULL.
3726 PR tree-optimization/85156
3727 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3728 evaluating the argument multiple times.
3730 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3732 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3734 (_mm_cvtpd_ps): Likewise.
3735 (_mm_cvttpd_epi32): Likewise.
3736 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3737 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3738 vector, pixel, and bool following altivec.h include.
3740 2018-04-03 Martin Sebor <msebor@redhat.com>
3742 * doc/extend.texi (Common Function Attributes): Clarify.
3743 (const attribute): Likewise.
3744 (pure attribute): Likewise.
3746 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3749 * config/i386/i386.c (ix86_expand_vector_set): Use
3750 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3752 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3754 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3755 instructions when changing rounding bits to preserve precision bits
3756 in the x87 control word.
3758 2018-04-03 Martin Liska <mliska@suse.cz>
3760 PR tree-optimization/82491
3761 * rtl.h (strip_offset_and_add): Replace += suboffset with
3762 poly_uint64 () + suboffset.
3764 2018-03-29 Martin Liska <mliska@suse.cz>
3765 Martin Jambor <mjambor@suse.cz>
3768 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3769 param_type is not an integral or pointer type.
3771 2018-04-03 Richard Biener <rguenther@suse.de>
3773 * sese.h (recompute_all_dominators): Remove.
3775 2018-04-02 Martin Sebor <msebor@redhat.com>
3777 * doc/invoke.texi (-Wrestrict): Fix typos.
3779 2018-04-02 Jim Wilson <jimw@sifive.com>
3781 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3782 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3783 (<optab>di3, <optab>si3_extend): Likewise.
3784 (<optab>si3_mask, <optab>si3_mask_1): New.
3785 (<optab>di3_mask, <optab>di3_mask_1): New.
3786 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3787 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3788 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3790 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3792 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3795 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3797 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3798 (nds32_canonicalize_comparison): New function.
3800 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3801 Kito Cheng <kito.cheng@gmail.com>
3802 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3804 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3805 * config/nds32/constants.md (unspec_volatile_element): Add
3806 UNSPEC_VOLATILE_RELAX_GROUP.
3807 * config/nds32/nds32-relax-opt.c: New file.
3808 * config/nds32/nds32-predicates.c
3809 (nds32_symbol_load_store_p): New function.
3810 * config/nds32/nds32-protos.h
3811 (nds32_symbol_load_store_p): Declare function.
3812 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3813 * config/nds32/nds32.c
3814 (nds32_register_pass): New function to register pass.
3815 (nds32_register_passes): New function to register passes.
3816 * config/nds32/nds32.md (relax_group): New pattern.
3817 * config/nds32/nds32.opt (mrelax-hint): New option.
3818 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3820 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3822 * config/nds32/t-nds32: Modify files dependency.
3824 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3826 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3827 (PROFILE_HOOK): Define its implementation.
3829 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3831 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3832 type and 32-bit size.
3834 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3837 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3838 (V_128_256): New mode iterator.
3839 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3840 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3841 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3843 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3844 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3846 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3849 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3850 NaN inputs correctly.
3852 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3855 * config/rs6000/vsx.md (??r): New mode attribute.
3856 (*vsx_mov<mode>_64bit): Use it.
3857 (*vsx_mov<mode>_32bit): Likewise.
3859 2018-03-30 Martin Sebor <msebor@redhat.com>
3861 PR tree-optimization/84818
3862 * builtins.c (check_access): Use warning_n.
3864 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3867 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3869 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3872 2018-03-30 Julia Koval <julia.koval@intel.com>
3875 * x86-tune.def (movx, partial_reg_dependency): Enable for
3878 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3881 * lra-constraints.c (process_alt_operands): Move setting
3882 this_alternative_matches below.
3884 2018-03-29 Martin Liska <mliska@suse.cz>
3887 * doc/invoke.texi: Document how LTO works with debug info.
3888 Describe auto-load support of binutils. Mention 'x86-64'
3889 as valid option value of -march option.
3891 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3893 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3896 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3897 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3898 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3901 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3904 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3905 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3906 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3907 for RS6000_BTM_POWERPC64.
3908 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3909 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3910 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3915 2018-03-28 Carl Love <cel@us.ibm.com>
3918 2017-09-27 Carl Love <cel@us.ibm.com>
3920 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3921 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3922 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3925 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3927 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3928 instead of __vector bool.
3929 (_mm_max_pu8): Likewise.
3930 (_mm_min_pi16): Likewise.
3932 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3935 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3936 (DIVWEUO): Likewise.
3938 (DIVDEUO): Likewise.
3939 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3940 DIVWEUO and DIVDEUO.
3941 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3942 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3943 (div_extend): Likewise.
3944 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3946 (__builtin_divweuo): Likewise.
3947 (__builtin_divdeo): Likewise.
3948 (__builtin_divdeuo): Likewise.
3950 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3953 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3954 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3956 PR tree-optimization/82004
3957 * gimple-match-head.c (optimize_pow_to_exp): New function.
3958 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3959 Don't fold to exp if optimize_pow_to_exp is false.
3961 2018-03-28 Martin Liska <mliska@suse.cz>
3964 * calls.c (initialize_argument_information): Fix trailing space.
3965 * common.opt: Fix typo and provide better explanation for
3966 -fsanitize-coverage option.
3967 * config/i386/i386.opt: Fix typo.
3969 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3970 Martin Liska <mliska@suse.cz>
3973 * gimplify.c (asan_poison_variable): Don't do the check for
3974 gimplify_omp_ctxp here.
3975 (gimplify_decl_expr): Do it here.
3976 (gimplify_target_expr): Likewise.
3978 2018-03-28 Martin Liska <mliska@suse.cz>
3981 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3982 chkp_type_bounds_count if MPX is not enabled.
3984 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3986 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3988 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3991 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3992 function to create the function decl for complex long double
3993 multiply and divide for -mabi=ieeelongdouble.
3994 (init_float128_ieee): Call it.
3996 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3999 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4000 -fcf-protection=branch -mibt.
4001 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4003 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4006 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4008 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4011 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4012 extern array declarations.
4014 2018-03-27 Richard Biener <rguenther@suse.de>
4017 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4018 explicit single_use checks.
4020 2018-03-27 Richard Biener <rguenther@suse.de>
4022 PR tree-optimization/85082
4023 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4026 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4028 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4029 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4030 Turn on fasynchronous-unwind-tables and funwind-tables.
4032 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4035 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4036 (*bmi_blsr_<mode>_ccz): Ditto.
4038 2018-03-26 Tom de Vries <tom@codesourcery.com>
4040 PR tree-optimization/85063
4041 * omp-general.c (offloading_function_p): New function. Factor out
4043 * omp-offload.c (pass_omp_target_link::gate): ... here.
4044 * omp-general.h (offloading_function_p): Declare.
4045 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4046 with attribute omp declare target for offloading functions.
4048 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4050 PR tree-optimization/84005
4051 * tree-data-ref.h (get_base_for_alignment): Declare.
4052 * tree-data-ref.c (get_base_for_alignment_1): New function.
4053 (get_base_for_alignment): Likewise.
4054 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4055 get_base_for_alignment to find a suitable base object, instead
4056 of always using drb->base_address.
4058 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4061 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4062 known size by default.
4064 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4067 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4068 and non BLKmode operands.
4070 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4073 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4074 Clean up attributes.
4076 2018-03-23 Richard Biener <rguenther@suse.de>
4079 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4080 we are going to emit early debug for LTO.
4082 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4085 * function.c (match_asm_constraints_1): Don't optimize if input
4086 doesn't satisfy general_operand predicate for output's mode.
4089 * alias.c (write_dependence_p): Don't require for x_canonicalized
4090 non-VOIDmode if x has VOIDmode.
4093 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4094 just don't try to optimize it rather than assert it never happens.
4096 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4098 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4099 macro expansions for definition of ST_INTERNAL_<mode> and
4100 LD_INTERNAL_<mode> builtins.
4101 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4103 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4105 (altivec_expand_st_builtin): Likewise.
4106 (altivec_expand_builtin): Remove calls to deleted functions.
4107 (rs6000_address_for_altivec): Delete this function.
4108 * config/rs6000/vector.md: Remove expands for
4109 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4111 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4114 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4115 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4116 re-computing once computed.
4117 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4118 (arm_init_machine_status): Initialize
4119 machine->static_chain_stack_bytes.
4121 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4124 * doc/extend.texi: Add four new prototypes for vec_ld.
4125 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4126 definitions for more logical presentation.
4127 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4128 entries for V1TI variants of __builtin_altivec_ld builtin.
4129 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4130 handling of V1TI variant of LVX icode pattern.
4131 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4132 (rs6000_gimple_fold_builtin): Likewise.
4133 (altivec_init_builtins): Add code to define
4134 __builtin_altivec_lvx_v1ti function.
4136 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4139 * function.c (match_asm_constraints_1): Don't do the optimization
4140 if input isn't a REG, SUBREG, MEM or constant.
4142 2018-03-22 Tom de Vries <tom@codesourcery.com>
4144 PR tree-optimization/84956
4145 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4146 bb_has_abnormal_pred.
4148 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4151 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4152 DECL_INITIAL (decl) to decl at the end.
4153 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4156 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4158 * doc/extend.texi (__builtin_tgmath): Document when complex
4159 integer types are treated as _Complex _Float64.
4161 2018-03-21 Tom de Vries <tom@codesourcery.com>
4163 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4165 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4167 PR tree-optimization/84960
4168 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4169 if it is ENTRY block, move them into single succ of ENTRY in that case.
4171 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4173 PR tree-optimization/84811
4174 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4175 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4176 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4177 (known_subrange_p): Update accordingly. Cast each value involved
4178 in the size comparison, rather than casting the result of the
4181 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4183 PR tree-optimization/84982
4184 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4185 by flipping the least significant bit rather than all bits from
4186 bitpos to bitpos + bitsize - 1.
4188 2018-03-21 Nathan Sidwell <nathan@acm.org>
4190 * doc/extend.texi (Deprecated Features): Remove mention of
4191 long-deleted deprecations.
4193 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4196 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4197 * configure: Regenerate.
4199 2018-03-21 Tom de Vries <tom@codesourcery.com>
4201 PR tree-optimization/83126
4202 * tree-parloops.c (num_phis): New function.
4203 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4205 2018-03-21 Nathan Sidwell <nathan@acm.org>
4207 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4208 mention anon-struct/union members and trailing attributes.
4210 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4212 PR tree-optimization/84969
4213 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4214 builtin memset partitions if they set different rhs values.
4216 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4218 PR rtl-optimization/84989
4219 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4220 VEC_DUPLICATE with scalar result mode.
4222 2018-03-21 Martin Liska <mliska@suse.cz>
4225 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4226 not intended return statement.
4228 2018-03-21 Martin Liska <mliska@suse.cz>
4231 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4232 (chkp_find_bound_slots_1): Limit number of iterations.
4234 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4237 * Minor grammar fixes for x86 options.
4239 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4242 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4243 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4245 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4248 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4249 (altivec_lvx_<mode>_1op): Likewise.
4250 (altivec_stvx_<mode>_2op): Likewise.
4251 (altivec_stvx_<mode>_1op): Likewise.
4252 (altivec_lvx_<VM2:mode>): New define_expand.
4253 (altivec_stvx_<VM2:mode>): Likewise.
4254 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4255 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4256 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4257 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4258 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4259 (rs6000_gen_lvx): Likewise.
4260 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4261 (altivec_expand_stv_builtin): Likewise.
4262 (altivec_expand_builtin): Likewise.
4263 * config/rs6000/vector.md: Likewise.
4265 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4268 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4271 2018-03-20 Richard Biener <rguenther@suse.de>
4274 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4275 sign-conversions as zero, fall back to standard scalar_stmt
4278 2018-03-20 Martin Liska <mliska@suse.cz>
4281 * predict.c (rebuild_frequencies): Handle case when we have
4282 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4284 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4287 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4288 flag_section_anchors.
4289 * varasm.c (use_blocks_for_decl_p): Remove hack for
4290 dw2_force_const_mem.
4293 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4295 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4296 be created, use lowpart_subreg of operands[0] rather than operands[0]
4298 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4299 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4300 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4301 and n constraint instead of aarch64_shift_imm_di and Usd.
4302 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4303 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4305 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4308 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4309 to favor GPR over NEON registers.
4310 (<shift>di3_neon): Likewise.
4312 2018-03-20 Tom de Vries <tom@codesourcery.com>
4315 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4316 (nvptx_process_pars): Emit bar.sync asap and alap.
4318 2018-03-20 Tom de Vries <tom@codesourcery.com>
4321 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4322 seen_label if seen_label is already set.
4324 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4327 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4328 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4329 Use 1U instead of 1. Formatting fixes.
4332 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4333 instead of TREE_TYPE (s1) for the return value.
4335 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4337 PR tree-optimization/84946
4338 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4339 bitsize + bitsize in poly_uint64 rather than poly_int64.
4342 * dwarf2asm.c: Include fold-const.c.
4343 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4344 of decl rather than decl itself.
4346 PR rtl-optimization/84643
4347 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4349 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4352 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4353 calling assemble_variable.
4355 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4358 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4359 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4361 2018-03-19 Jim Wilson <jimw@sifive.com>
4364 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4365 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4366 (riscv_first_stack_step): Likewise.
4367 (riscv_option_override): Use STACK_BOUNDARY instead of
4369 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4371 (BIGGEST_ALIGNMENT): Set to 128.
4372 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4373 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4376 2018-03-19 Richard Biener <rguenther@suse.de>
4378 PR tree-optimization/84933
4379 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4380 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4382 2018-03-19 Richard Biener <rguenther@suse.de>
4384 PR tree-optimization/84859
4385 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4386 (cond_if_else_store_replacement): Perform sinking operation on
4387 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4388 Generalize what a BB with a single eligible store is.
4390 2018-03-19 Richard Biener <rguenther@suse.de>
4392 PR tree-optimization/84929
4393 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4394 chrec_is_positive against non-chrec arg.
4396 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4399 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4401 2018-03-18 Martin Liska <mliska@suse.cz>
4403 PR rtl-optimization/84635
4404 * regrename.c (build_def_use): Use matches_mode only when
4407 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4409 PR tree-optimization/84913
4410 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4411 vectorize chains of COND_EXPRs.
4413 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4415 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4417 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4419 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4421 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4423 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4425 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4426 Kito Cheng <kito.cheng@gmail.com>
4428 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4429 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4430 (nds32_adjust_reg_alloc_order): New function.
4431 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4433 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4435 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4436 nds32_print_operand, nds32_print_operand_address): Use
4437 HOST_WIDE_INT_PRINT_DEC instead.
4439 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4441 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4443 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4446 * config/i386/i386.c (initial_ix86_tune_features,
4447 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4449 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4450 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4451 rather than 1u << ix86_tune. Formatting fix.
4452 (ix86_option_override_internal): Change ix86_arch_mask from
4453 unsigned int to unsigned HOST_WIDE_INT, initialize to
4454 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4455 (ix86_function_specific_restore): Likewise.
4457 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4460 * postreload.c (reload_combine_recognize_pattern): Perform
4461 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4462 truncate_int_for_mode the result for the destination's mode.
4465 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4466 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4468 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4471 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4474 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4475 regno_allocno_class_array and sorted_pseudos.
4476 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4477 insns where regno is used.
4479 2018-03-16 Martin Liska <mliska@suse.cz>
4482 * multiple_target.c (create_dispatcher_calls): Redirect
4483 reference in the symbol table.
4485 2018-03-16 Martin Liska <mliska@suse.cz>
4488 * multiple_target.c (create_dispatcher_calls): Redirect also
4491 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4495 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4496 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4497 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4499 2018-03-16 Julia Koval <julia.koval@intel.com>
4501 * doc/invoke.texi (Skylake Server): Add CLWB.
4502 Cannonlake): Remove CLWB.
4504 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4506 PR tree-optimization/84841
4507 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4509 (FLOAT_ONE_CONST_TYPE): Define.
4510 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4511 (sort_by_operand_rank): Put entries with higher constant_type last
4512 rather than first to match comments.
4514 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4516 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4519 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4522 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4523 check and use address of target always.
4525 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4528 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4529 (indirect_thunk_bnd_needed): Likewise.
4530 (indirect_thunks_used): Likewise.
4531 (indirect_thunks_bnd_used): Likewise.
4532 (indirect_return_needed): New.
4533 (indirect_return_bnd_needed): Likewise.
4534 (output_indirect_thunk_function): Add a bool argument for
4536 (output_indirect_thunk_function): Don't generate alias for
4537 function return thunk.
4538 (ix86_code_end): Call output_indirect_thunk_function to generate
4539 function return thunks.
4540 (ix86_output_function_return): Set indirect_return_bnd_needed
4541 and indirect_return_needed instead of indirect_thunk_bnd_needed
4542 and indirect_thunk_needed.
4544 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4546 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4547 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4548 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4550 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4551 Paul Hua <paul.hua.gm@gmail.com>
4554 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4556 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4558 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4559 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4562 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4565 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4566 instead of GET_MODE_SIZE when comparing Units.
4568 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4571 * varasm.c (hash_section): Return an unchangeble hash value
4572 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4573 Return !aarch64_can_use_per_function_literal_pools_p ().
4575 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4578 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4579 rather than address of cmode as last argument to prepare_cmp_insn.
4581 2018-03-15 Julia Koval <julia.koval@intel.com>
4583 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4584 F_AVX512VNNI, F_AVX512BITALG): New.
4586 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4589 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4590 insn for floating-point loads and stores.
4592 2018-03-14 Carl Love <cel@us.ibm.com>
4594 * config/rs6000/rs6000-c.c: Add macro definitions for
4595 ALTIVEC_BUILTIN_VEC_PERMXOR.
4596 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4597 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4598 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4599 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4601 * config/doc/extend.texi: Add prototypes for vec_permxor.
4603 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4606 * diagnostic-show-locus.c (class layout_point): Convert m_line
4607 from int to linenum_type.
4608 (line_span::comparator): Use linenum "compare" function when
4609 comparing line numbers.
4610 (test_line_span): New function.
4611 (layout_range::contains_point): Convert param "row" from int to
4613 (layout_range::intersects_line_p): Likewise.
4614 (layout::will_show_line_p): Likewise.
4615 (layout::print_source_line): Likewise.
4616 (layout::should_print_annotation_line_p): Likewise.
4617 (layout::print_annotation_line): Likewise.
4618 (layout::print_leading_fixits): Likewise.
4619 (layout::annotation_line_showed_range_p): Likewise.
4620 (struct line_corrections): Likewise for field m_row.
4621 (line_corrections::line_corrections): Likewise for param "row".
4622 (layout::print_trailing_fixits): Likewise.
4623 (layout::get_state_at_point): Likewise.
4624 (layout::get_x_bound_for_row): Likewise.
4625 (layout::print_line): Likewise.
4626 (diagnostic_show_locus): Likewise for locals "last_line" and
4628 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4629 * input.c (selftest::test_linenum_comparisons): New function.
4630 (selftest::input_c_tests): Call it.
4631 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4632 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4633 * selftest.h (ASSERT_GT): New macro.
4634 (ASSERT_GT_AT): New macro.
4635 (ASSERT_LT): New macro.
4636 (ASSERT_LT_AT): New macro.
4638 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4640 PR rtl-optimization/84780
4641 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4643 2018-03-14 Martin Liska <mliska@suse.cz>
4645 * tree.c (record_node_allocation_statistics): Use
4646 get_stats_node_kind.
4647 (get_stats_node_kind): New function extracted from
4648 record_node_allocation_statistics.
4649 (free_node): Use get_stats_node_kind.
4651 2018-03-14 Richard Biener <rguenther@suse.de>
4653 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4654 that the value-set of ANTIC_IN doesn't grow.
4657 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4659 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4660 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4662 2018-03-14 Julia Koval <julia.koval@intel.com>
4664 * config.gcc (icelake-client, icelake-server): New.
4666 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4667 (initial_ix86_arch_features): Ditto.
4668 (PTA_SKYLAKE): Add SGX.
4669 (PTA_ICELAKE): Remove.
4670 (PTA_ICELAKE_CLIENT): New.
4671 (PTA_ICELAKE_SERVER): New.
4672 (ix86_option_override_internal): Split up icelake on icelake client and
4674 (get_builtin_code_for_version): Ditto.
4675 (fold_builtin_cpu): Ditto.
4676 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4677 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4678 * config/i386/i386.h (processor_type): Ditto.
4679 * doc/invoke.texi: Ditto.
4681 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4684 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4685 INTEGER_CST offset, add it together with bitpos / 8 and
4686 sign extend based on POINTER_SIZE.
4690 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4693 * config/i386/constraints.md (Yc): New register constraint.
4694 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4695 Use Yc constraint for alternative 2 of operand 0. Remove
4696 preferred_for_speed attribute.
4698 2018-03-14 Richard Biener <rguenther@suse.de>
4700 PR tree-optimization/84830
4701 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4702 with the old one to avoid oscillations.
4704 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4707 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4709 (assign_by_spills): Return a flag of reload assignment failure.
4710 Do not process the reload assignment failures. Do not spill other
4711 reload pseudos if they has the same reg class. Update n if
4713 (lra_assign): Add a return arg. Set up from the result of
4714 assign_by_spills call.
4715 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4716 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4717 usage_insns if it is not NULL.
4718 (spill_hard_reg_in_range): New function.
4719 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4720 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4721 function prototypes.
4722 (lra_assign): Change prototype.
4723 * lra.c (lra): Add code to deal with fails by splitting hard reg
4726 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4728 * config/riscv/riscv.opt (mrelax): New option.
4729 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4730 "norelax" when riscv_mrelax is disabled.
4731 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4733 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4736 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4737 reassociation for int modes.
4739 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4741 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4742 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4744 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4745 * config/aarch64/aarch64-sve.md
4746 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4747 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4748 (*extend<mode><Vwide>2): Rename to...
4749 (aarch64_sve_extend<mode><Vwide>2): ...this.
4750 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4751 renaming the old pattern to...
4752 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4754 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4755 define_expand, renaming the old pattern to...
4756 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4757 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4758 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4759 account when deciding which SVE instruction the optab should use.
4760 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4762 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4764 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4765 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4766 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4767 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4768 (tlsdesc_small_<mode>): Turn a define_expand and use
4769 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4770 (tlsdesc_small_advsimd_<mode>): ...this.
4771 (tlsdesc_small_sve_<mode>): New pattern.
4773 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4775 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4776 (UNSPEC_UMUL_HIGHPART): New constants.
4777 (MUL_HIGHPART): New int iteraor.
4778 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4779 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4781 (*<su>mul<mode>3_highpart): New define_insn.
4783 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4786 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4789 2018-03-13 Martin Liska <mliska@suse.cz>
4792 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4794 (sem_item_optimizer::~sem_item_optimizer): Release it.
4795 (sem_item_optimizer::merge_classes): Register variable aliases.
4796 (sem_item_optimizer::fixup_pt_set): New function.
4797 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4798 * ipa-icf.h: Declare new variables and functions.
4800 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4803 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4804 integer_pow2p@2 and test integer_pow2p in condition.
4805 (A < 0 ? C : 0): Similarly for @1.
4808 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4809 characters starting at p contain '\0' character, don't look beyond
4813 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4814 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4817 * reg-stack.c (change_stack): Change update_end var from int to
4818 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4819 also call set_block_for_insn on the newly added insns and rescan.
4822 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4823 on the last operand.
4826 * tree.c (stabilize_reference_1): Return save_expr (e) for
4827 STATEMENT_LIST even if it doesn't have side-effects.
4829 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4831 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4833 2018-03-12 Renlin Li <renlin.li@arm.com>
4835 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4836 aarch64_output_scalar_simd_mov_immediate.
4838 2018-03-12 Martin Sebor <msebor@redhat.com>
4840 PR tree-optimization/83456
4841 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4842 for perfectly overlapping calls to memcpy.
4843 (gimple_fold_builtin_memory_chk): Same.
4844 (gimple_fold_builtin_strcpy): Handle no-warning.
4845 (gimple_fold_builtin_stxcpy_chk): Same.
4846 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4848 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4850 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4851 parameter. Use it for SFmode.
4852 (rs6000_function_arg_advance_1): Adjust.
4853 (rs6000_function_arg): Adjust.
4854 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4856 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4858 PR rtl-optimization/84169
4859 PR rtl-optimization/84780
4860 * combine.c (can_combine_p): Check for a 2-insn combination whether
4861 the destination register is used between the two insns, too.
4863 2018-03-12 Richard Biener <rguenther@suse.de>
4865 PR tree-optimization/84803
4866 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4867 for refs DR analysis didn't process.
4869 2018-03-12 Richard Biener <rguenther@suse.de>
4871 PR tree-optimization/84777
4872 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4873 force-vectorize loops ignore whether we are optimizing for size.
4875 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4877 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4878 (TARGET_MD_ASM_ADJUST): Define.
4880 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4881 Kito Cheng <kito.cheng@gmail.com>
4882 Chung-Ju Wu <jasonwucj@gmail.com>
4884 * config/nds32/nds32.c (nds32_compute_stack_frame,
4885 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4886 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4887 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4888 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4889 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4890 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4891 * config/nds32/nds32.md (prologue, epilogue): Use macro
4892 NDS32_V3PUSH_AVAILABLE_P to do checking.
4894 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4897 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4898 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4899 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4900 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4901 addition of most attributes on !orig_type_die or the attribute not
4902 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4904 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4905 Chung-Ju Wu <jasonwucj@gmail.com>
4907 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4909 * config/nds32/nds32.opt (mvh): New option.
4911 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4912 Chung-Ju Wu <jasonwucj@gmail.com>
4914 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4916 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4917 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4920 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4921 Chung-Ju Wu <jasonwucj@gmail.com>
4923 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4925 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4926 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4928 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4929 Kito Cheng <kito.cheng@gmail.com>
4930 Chung-Ju Wu <jasonwucj@gmail.com>
4932 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4933 UNSPEC_FFMISM and UNSPEC_FLMISM.
4934 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4935 for ffb, ffmism and flmism.
4936 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4937 (unspec_ffmism): Ditto.
4938 (unspec_flmism): Ditto.
4939 (nds32_expand_builtin_impl): Check if string extension is available.
4940 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4941 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4943 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4946 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4949 * lra-assigns.c (assign_by_spills): Return a flag of reload
4950 assignment failure. Do not process the reload assignment
4951 failures. Do not spill other reload pseudos if they has the same
4953 (lra_assign): Add a return arg. Set up from the result of
4954 assign_by_spills call.
4955 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4956 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4957 usage_insns if it is not NULL.
4958 (spill_hard_reg_in_range): New function.
4959 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4960 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4961 function prototypes.
4962 (lra_assign): Change prototype.
4963 * lra.c (lra): Add code to deal with fails by splitting hard reg
4966 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4969 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4971 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4974 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4975 (dw_val_node): Add val_symbolic_view.
4976 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4977 (symview_upper_bound): New.
4978 (new_line_info_table): Initialize symviews_since_reset.
4979 (dwarf2out_source_line): Count symviews_since_reset and set
4980 symview_upper_bound.
4981 (dw_val_equal_p): Handle symview.
4982 (add_AT_symview): New.
4983 (print_dw_val): Handle symview.
4984 (attr_checksum, attr_checksum_ordered): Likewise.
4985 (same_dw_val_p, size_of_die): Likewise.
4986 (value_format, output_die): Likewise.
4987 (add_high_low_attributes): Use add_AT_symview for entry_view.
4988 (dwarf2out_finish): Reset symview_upper_bound, clear
4991 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4994 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4995 Add strict argument and use it.
4996 (rs6000_split_multireg_move): Update for new strict argument.
4997 (mem_operand_gpr): Disallow all non-offsettable addresses.
4998 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5000 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5003 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5004 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5005 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5008 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5009 decl, use remap_type if we want to use the type.
5011 2018-03-09 Martin Sebor <msebor@redhat.com>
5013 PR tree-optimization/84526
5014 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5016 (builtin_access::generic_overlap): Be prepared to handle non-array
5019 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5021 PR rtl-optimization/84682
5022 * lra-constraints.c (process_address_1): Check is_address flag
5023 for address constraints.
5024 (process_alt_operands): Likewise.
5025 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5026 preprocess_constraints.
5027 * recog.h (preprocess_constraints): Add oploc parameter.
5029 * recog.c (preprocess_constraints): Test address_operand for
5030 CT_ADDRESS constraints.
5032 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5035 * lra-assigns.c (assign_by_spills): Return a flag of reload
5036 assignment failure. Do not process the reload assignment
5037 failures. Do not spill other reload pseudos if they has the same
5039 (lra_assign): Add a return arg. Set up from the result of
5040 assign_by_spills call.
5041 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5042 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5043 usage_insns if it is not NULL.
5044 (spill_hard_reg_in_range): New function.
5045 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5046 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5047 function prototypes.
5048 (lra_assign): Change prototype.
5049 * lra.c (lra): Add code to deal with fails by splitting hard reg
5052 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5055 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5056 Accept complain bool parameter. Only emit errors if it is true.
5057 (arm_parse_cpu_option_name): Likewise.
5058 (arm_target_thumb_only): Adjust callers of the above.
5059 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5060 prototype to take a default true bool parameter.
5061 (arm_parse_arch_option_name): Likewise.
5063 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5064 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5068 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5069 * configure: Regenerate.
5070 * configure.ac ("linker --version-script option"): New.
5071 ("linker soname option"): New.
5073 2018-03-09 Richard Biener <rguenther@suse.de>
5075 PR tree-optimization/84775
5076 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5077 immediate uses of predicate stmts and mark them modified.
5080 PR tree-optimization/84178
5081 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5083 (version_loop_for_if_conversion): Delay update_ssa call.
5084 (tree_if_conversion): Delay update_ssa until after predicate
5087 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5090 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5091 when the function accesses prior frames.
5093 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5096 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5097 gen_llsym, otherwise call maybe_gen_llsym.
5100 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5101 has ',' character inside of it.
5103 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5106 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5107 as clobbering CC_REGNUM.
5109 2018-03-08 Richard Biener <rguenther@suse.de>
5112 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5113 (follow_copies_to_constant): Do not follow SSA names registered
5116 2018-03-08 Richard Biener <rguenther@suse.de>
5118 PR tree-optimization/84178
5119 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5121 (version_loop_for_if_conversion): Delay update_ssa call.
5122 (tree_if_conversion): Delay update_ssa until after predicate
5125 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5127 PR tree-optimization/84178
5128 * tree-if-conv.c (release_bb_predicate): Remove the
5129 the assertion that the stmts have NULL use_ops.
5130 Discard the statements, asserting that they haven't
5131 yet been added to a BB.
5133 2018-03-08 Richard Biener <rguenther@suse.de>
5135 PR tree-optimization/84746
5136 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5137 (phi_translate): Pass in destination ANTIC_OUT set.
5138 (phi_translate_1): Likewise. For a simplified result lookup
5139 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5140 (phi_translate_set): Adjust.
5141 (do_pre_regular_insertion): Likewise.
5142 (do_pre_partial_partial_insertion): Likewise.
5144 2018-03-08 Martin Liska <mliska@suse.cz>
5146 PR gcov-profile/84735
5147 * doc/gcov.texi: Document usage of profile files.
5148 * gcov-io.h: Document changes in the format.
5150 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5154 * dwarf2out.c (struct dw_line_info_table): Update comments for
5156 (FORCE_RESET_NEXT_VIEW): New.
5157 (FORCE_RESETTING_VIEW_P): New.
5158 (RESETTING_VIEW_P): Check for -1 too.
5159 (ZERO_VIEW_P): Likewise.
5160 (new_line_info_table): Force-reset next view.
5161 (dwarf2out_begin_function): Likewise.
5162 (dwarf2out_source_line): Simplify zero_view_p initialization.
5163 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5164 view directly. Omit view when omitting .loc at line 0.
5166 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5168 PR tree-optimization/84740
5169 * tree-switch-conversion.c (process_switch): Call build_constructors
5170 only if info.phi_count is non-zero.
5172 PR tree-optimization/84739
5173 * tree-tailcall.c (find_tail_calls): Check call arguments against
5174 DECL_ARGUMENTS (current_function_decl) rather than
5175 DECL_ARGUMENTS (func) when checking for tail recursion.
5177 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5179 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5180 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5181 Volker Reichelt's entry and add entries for people that perform
5182 GCC fuzzy testing and report numerous bugs.
5184 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5187 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5188 readonly data in sdata, if that is disabled.
5189 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5190 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5191 -mreadonly-in-sdata option.
5193 2018-03-07 Martin Sebor <msebor@redhat.com>
5195 PR tree-optimization/84468
5196 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5197 basic block when looking for nul assignment.
5199 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5202 * except.h (output_function_exception_table): Adjust prototype.
5203 * except.c (output_function_exception_table): Remove FNNAME parameter
5204 and add SECTION parameter. Ouput one part of the table at a time.
5205 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5206 the first part of the exception table and emit unwind directives.
5207 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5208 (i386_pe_seh_cold_init): Likewise.
5209 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5210 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5211 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5212 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5213 * config/i386/winnt.c: Include except.h.
5214 (struct seh_frame_state): Add reg_offset, after_prologue and
5215 in_cold_section fields.
5216 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5217 (i386_pe_seh_cold_init): New function.
5218 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5219 to seh->in_cold_section.
5220 (seh_emit_push): Record the offset of the push.
5221 (seh_emit_save): Record the offet of the save.
5222 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5223 Test seh->after_prologue to disregard the epilogue.
5224 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5225 (i386_pe_end_cold_function): New function.
5227 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5230 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5231 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5234 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5236 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5237 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5240 * multiple_target.c: Include tree-inline.h and intl.h.
5241 (expand_target_clones): Diagnose and fail if node->definition and
5242 !tree_versionable_function_p (node->decl).
5244 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5246 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5248 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5249 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5250 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5252 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5255 * combine.c (try_combine): Use reg_or_subregno instead of handling
5256 just paradoxical SUBREGs and REGs.
5258 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5260 * config/arc/arc.c (arc_finalize_pic): Remove function.
5261 (arc_must_save_register): We use single base PIC register, remove
5262 checks to save/restore the PIC register.
5263 (arc_expand_prologue): Likewise.
5264 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5266 (arc_verify_short): Likewise.
5267 (arc_attr_type): Likewise.
5268 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5269 (walk_stores): Likewise.
5270 (arc_address_cost): Make it static.
5271 (arc_verify_short): Likewise.
5272 (branch_dest): Likewise.
5273 (arc_attr_type): Likewise.
5274 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5275 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5276 (arc_final_prescan_insn): Remove inserting the nops due to
5277 hardware hazards. It is done in reorg step.
5278 (insn_length_variant_t): Remove.
5279 (insn_length_parameters_t): Likewise.
5280 (arc_insn_length_parameters): Likewise.
5281 (arc_get_insn_variants): Likewise.
5282 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5284 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5287 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5290 PR tree-optimization/84687
5291 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5293 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5295 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5297 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5298 Rename to ppc_speculation_barrier.
5299 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5300 __builtin_ppc_speculation_barrier.
5302 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5305 * combine.c (combine_simplify_rtx): Don't try to simplify if
5306 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5309 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5311 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5312 to 32 bytes when compiling for POWER9.
5314 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5317 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5318 regparm >= 3 with no arg reg available also for calls with
5319 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5322 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5324 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5326 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5329 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5331 2018-03-05 Richard Biener <rguenther@suse.de>
5333 PR tree-optimization/84486
5334 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5335 When inserting a __builtin_assume_aligned call set the LHS
5336 SSA name alignment info accordingly.
5338 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5340 PR tree-optimization/84114
5341 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5342 Avoid reassociation of FLOAT_MODE addition.
5344 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5346 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5347 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5348 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5349 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5350 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5351 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5352 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5354 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5355 __builtin_ia32_wbinvd): New builtins.
5356 (SPECIAL_ARGS2): New.
5357 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5358 (SPECIAL_ARGS2): New.
5359 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5360 (ix86_valid_target_attribute_inner_p): Ditto.
5361 (ix86_init_mmx_sse_builtins): Add special_args2.
5362 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5363 TARGET_WBNOINVD_P): New.
5364 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5365 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5366 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5367 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5368 * config/i386/pconfigintrin.h: New file.
5369 * config/i386/wbnoinvdintrin.h: Ditto.
5370 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5371 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5373 2018-03-05 Richard Biener <rguenther@suse.de>
5375 PR tree-optimization/84670
5376 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5378 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5379 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5380 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5381 doesn't grow if all successors (recursively) were visited at least
5384 2018-03-05 Richard Biener <rguenther@suse.de>
5386 PR tree-optimization/84650
5387 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5388 if executed in the loop pipeline.
5390 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5392 * doc/configfiles.texi (Configuration Files): Move info about
5393 conditionalizing $target-protos.h to...
5394 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5395 differs from $target-protos.h.
5397 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5398 Chung-Ju Wu <jasonwucj@gmail.com>
5400 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5401 * config/nds32/nds32-multiple.md (setmemsi): Define.
5402 * config/nds32/nds32-memory-manipulation.c
5403 (nds32_gen_dup_4_byte_to_word_value): New.
5404 (emit_setmem_word_loop): New.
5405 (emit_setmem_byte_loop): New.
5406 (nds32_expand_setmem_loop): New.
5407 (nds32_expand_setmem_loop_v3m): New.
5408 (nds32_expand_setmem_unroll): New.
5409 (nds32_expand_setmem): New.
5411 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5412 Chung-Ju Wu <jasonwucj@gmail.com>
5414 * config/nds32/nds32-memory-manipulation.c
5415 (nds32_emit_load_store): New.
5416 (nds32_emit_post_inc_load_store): New.
5417 (nds32_emit_mem_move): New.
5418 (nds32_emit_mem_move_block): New.
5419 (nds32_expand_movmemsi_loop_unknown_size): New.
5420 (nds32_expand_movmemsi_loop_known_size): New.
5421 (nds32_expand_movmemsi_loop): New.
5422 (nds32_expand_movmemsi_unroll): New.
5423 (nds32_expand_movmemqi): Rename ...
5424 (nds32_expand_movmemsi): ... to this.
5425 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5426 (movmemsi): ... to this.
5427 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5428 (nds32_expand_movmemsi): ... to this.
5430 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5431 Monk Chiang <sh.chiang04@gmail.com>
5432 Chung-Ju Wu <jasonwucj@gmail.com>
5434 * config/nds32/nds32-protos.h
5435 (nds32_expand_load_multiple): New arguments.
5436 (nds32_expand_store_multiple): Ditto.
5437 (nds32_valid_multiple_load_store): Rename ...
5438 (nds32_valid_multiple_load_store_p): ... to this.
5439 * config/nds32/nds32-memory-manipulation.c
5440 (nds32_expand_load_multiple): Refine implementation.
5441 (nds32_expand_store_multiple): Ditto.
5442 * config/nds32/nds32-multiple.md
5443 (load_multiple): Update nds32_expand_load_multiple interface.
5444 (store_multiple): Update nds32_expand_store_multiple interface.
5445 * config/nds32/nds32-predicates.c
5446 (nds32_valid_multiple_load_store): Rename ...
5447 (nds32_valid_multiple_load_store_p): ... to this and refine
5449 * config/nds32/predicates.md
5450 (nds32_load_multiple_and_update_address_operation): New predicate.
5451 (nds32_store_multiple_and_update_address_operation): New predicate.
5453 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5454 Chung-Ju Wu <jasonwucj@gmail.com>
5456 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5457 (combo): New attribute.
5458 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5460 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5462 * config/nds32/nds32.opt: Change -mcmodel= default value.
5464 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5465 Monk Chiang <sh.chiang04@gmail.com>
5466 Chung-Ju Wu <jasonwucj@gmail.com>
5468 * config/nds32/constants.md (unspec_element): New enum.
5469 * config/nds32/constraints.md (Umw): New constraint.
5470 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5471 * config/nds32/nds32-intrinsic.md: Likewise.
5472 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5473 (nds32_valid_smw_lwm_base_p): New.
5474 (nds32_output_smw_single_word): New.
5475 (nds32_output_lmw_single_word): New.
5476 (nds32_expand_unaligned_load): New.
5477 (nds32_expand_unaligned_store): New.
5478 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5479 (nds32_output_smw_single_word): Declare.
5480 (nds32_output_lmw_single_word): Declare.
5481 (nds32_expand_unaligned_load): Declare.
5482 (nds32_expand_unaligned_store): Declare.
5483 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5484 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5485 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5486 NDS32_BUILTIN_UASTORE_DW.
5487 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5490 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5491 Kito Cheng <kito.cheng@gmail.com>
5492 Chung-Ju Wu <jasonwucj@gmail.com>
5494 * config/nds32/nds32-intrinsic.c
5495 (nds32_expand_builtin_null_ftype_reg): Delete.
5496 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5497 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5498 (nds32_read_argument): New.
5499 (nds32_legitimize_target): Ditto.
5500 (nds32_legitimize_argument): Ditto.
5501 (nds32_check_constant_argument): Ditto.
5502 (nds32_expand_unop_builtin): Ditto.
5503 (nds32_expand_unopimm_builtin): Ditto.
5504 (nds32_expand_binop_builtin): Ditto.
5505 (nds32_builtin_decl_impl): Ditto.
5506 (builtin_description): Ditto.
5507 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5508 (nds32_init_builtins_impl): Ditto.
5509 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5510 (nds32_builtin_decl): New.
5511 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5512 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5514 2018-03-02 Jeff Law <law@redhat.com>
5516 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5517 (redundant_insn, fill_simple_delay_slots): Likewise.
5518 (fill_slots_from_thread): Likewise.
5519 * resource.c (mark_referenced_resources): Likewise.
5520 (mark_set_resources, find_dead_or_set_registers): Likewise.
5522 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5524 * substring-locations.h (format_warning_va): Formatting fix for
5526 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5528 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5529 * substring-locations.c: Include intl.h.
5530 (format_warning_va): Turned into small wrapper around
5531 format_warning_n_va, renamed to ...
5532 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5533 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5535 (format_warning_at_substring_n): New function.
5536 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5537 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5538 format_warning_at_substring with just a shorter name instead of
5539 const function pointer.
5540 (fmtwarn_n): New function.
5541 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5542 appropriate, get rid of all the fmtstr temporaries, move conditionals
5543 with G_() wrapped string literals directly into fmtwarn arguments,
5544 cast dir.len to (int), formatting fixes.
5546 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5548 * doc/invoke.texi: Remove "Cilk Plus" references.
5550 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5551 Richard Biener <rguenther@suse.de>
5554 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5555 for error or warning attributes if CALL_FROM_THUNK_P is set.
5558 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5561 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5562 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5565 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5566 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5568 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5569 instead of -1U in last predictors element's probability member.
5571 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5574 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5575 arguments if they are comparable.
5577 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5579 PR tree-optimization/84634
5580 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5581 masks and masked_loop_p with a single loop_masks, making sure it's
5582 null for bb vectorization.
5584 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5586 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5587 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5588 loop->force_vectorize to check whether there is no alias.
5590 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5593 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5595 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5597 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5598 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5599 instead of a loop around prev_real_insn.
5600 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5604 * config/i386/i386.c (ix86_print_operand): Use conditional
5605 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5608 2018-03-02 Richard Biener <rguenther@suse.de>
5610 PR tree-optimization/84427
5611 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5612 (bitmap_set_subtract_values): Rewrite to handle multiple
5615 (prune_clobbered_mems): Likewise.
5616 (phi_translate): Take edge instead of pred/phiblock.
5617 (phi_translate_1): Likewise.
5618 (phi_translate_set): Likewise. Insert all translated
5619 exprs for a value into the set, keeping possibly multiple
5620 expressions per value.
5621 (compute_antic_aux): Adjust for phi_translate changes.
5622 When intersecting union the expressions and prune those
5623 not in the final value set, keeping possibly multiple
5624 expressions per value. Do not use value-insertion
5625 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5626 all expressions. Add verification that the value-sets
5627 only shrink during iteration.
5628 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5629 (do_pre_regular_insertion): Likewise.
5630 (do_pre_partial_partial_insertion): Likewise.
5632 2018-03-02 Richard Biener <rguenther@suse.de>
5635 * config/darwin.c (saved_debug_info_level): New static global.
5636 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5637 (darwin_asm_lto_end): Restore debug info generation settings.
5639 2018-03-01 Martin Liska <mliska@suse.cz>
5642 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5645 2018-03-01 Richard Biener <rguenther@suse.de>
5648 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5649 type annotation with LTO when debug was disabled at compile-time.
5651 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5653 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5655 (mips_final_postscan_insn): Likewise.
5657 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5659 PR rtl-optimization/84528
5660 * alias.c (init_alias_target): Add commentary.
5661 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5662 a unique base value if the frame pointer is not eliminated
5663 to the stack pointer.
5665 2018-03-01 Tom de Vries <tom@codesourcery.com>
5667 PR rtl-optimization/83327
5668 * lra-int.h (hard_regs_spilled_into): Declare.
5669 * lra.c (hard_regs_spilled_into): Define.
5670 (init_reg_info): Init hard_regs_spilled_into.
5671 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5672 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5673 (process_bb_lives): Handle hard_regs_spilled_into.
5674 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5675 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5677 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5679 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5680 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5681 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5682 * config/rs6000/aix72.h: New file.
5684 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5686 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5687 instead of warning_at with conditional singular and plural messages
5691 * stor-layout.c (update_alignment_for_field): For
5692 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5693 && !DECL_PACKED (field), do the alignment update, just use
5694 only desired_align instead of MAX (type_align, desired_align)
5696 (place_field): Don't do known_align < desired_align handling
5697 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5698 is non-NULL, instead do it after rli->prev_field handling and
5699 only if not within a bitfield word. For DECL_PACKED (field)
5700 use type_align of BITS_PER_UNIT.
5702 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5704 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5705 superfluous parentheses and trailing spaces.
5707 2018-02-28 Richard Biener <rguenther@suse.de>
5709 PR tree-optimization/84584
5710 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5711 SCoPs with fake exit edge.
5713 2018-02-28 Martin Liska <mliska@suse.cz>
5716 * timevar.c (timer::print): Fix format to properly print 100%
5719 2018-02-28 Richard Biener <rguenther@suse.de>
5722 * genmatch.c (capture_info::walk_match): Do not mark
5723 captured expressions without operands as expr_p given
5724 they act more like predicates and should be subject to
5725 "lost tail" side-effect preserving.
5727 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5729 PR rtl-optimization/81611
5730 * auto-inc-dec.c (attempt_change): Move dead note from
5731 mem_insn if it's the next use of regno
5732 (find_address): Take address use of reg holding
5733 non-incremented value. Add parm to limit search to the named
5735 (merge_in_block): Attempt to use a mem insn that is the next
5736 use of the original regno.
5738 2018-02-27 Martin Sebor <msebor@redhat.com>
5741 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5742 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5744 2018-02-27 Martin Sebor <msebor@redhat.com>
5746 PR translation/84207
5747 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5748 n argument to unsigned HOST_WIDE_INT.
5749 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5750 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5751 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5752 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5754 2018-02-27 Richard Biener <rguenther@suse.de>
5756 PR tree-optimization/84512
5757 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5758 Do not use the estimate returned from record_stmt_cost for
5759 the scalar iteration cost but sum properly using add_stmt_cost.
5761 2018-02-27 Richard Biener <rguenther@suse.de>
5763 PR tree-optimization/84466
5764 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5765 Adjust last change to less strictly validate use operands.
5767 2018-02-27 Martin Liska <mliska@suse.cz>
5769 PR gcov-profile/84548
5770 * gcov.c (process_file): Allow partial overlap and consider it
5771 also as group functions.
5772 (output_lines): Properly calculate range of lines for a group.
5774 2018-02-27 Martin Liska <mliska@suse.cz>
5776 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5777 'ggc' suffixes. Change first column width.
5778 (timer::print): Fix formatting of the column.
5780 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5782 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5783 preserve inline entry blocks for the sake of debug inline
5784 entry point markers alone.
5785 (remove_unused_locals): Suggest in comments a better place to
5786 force the preservation of inline entry blocks that are
5787 otherwise unused, but do not preserve them.
5789 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5791 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5793 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5796 * config/i386/constraints.md (Bs): Replace
5797 ix86_indirect_branch_register with
5798 TARGET_INDIRECT_BRANCH_REGISTER.
5800 * config/i386/i386.md (indirect_jump): Likewise.
5801 (tablejump): Likewise.
5802 (*sibcall_memory): Likewise.
5803 (*sibcall_value_memory): Likewise.
5804 Peepholes of indirect call and jump via memory: Likewise.
5805 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5806 (*sibcall_value_GOT_32): Likewise.
5807 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5808 (GOT_memory_operand): Likewise.
5809 (call_insn_operand): Likewise.
5810 (sibcall_insn_operand): Likewise.
5811 (GOT32_symbol_operand): Likewise.
5812 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5814 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5816 PR rtl-optimization/83496
5817 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5818 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5819 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5820 redundant insn, if any.
5821 (relax_delay_slots): Likewise.
5822 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5824 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5826 PR tree-optimization/83965
5827 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5828 that grouped statements are part of a reduction chain. Return
5829 true if the statement is not marked as a reduction itself but
5831 (vect_recog_dot_prod_pattern): Don't check whether the statement
5832 is part of a group here.
5833 (vect_recog_sad_pattern): Likewise.
5834 (vect_recog_widen_sum_pattern): Likewise.
5836 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5839 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5841 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5844 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5846 (ix86_output_indirect_function_return): New prototype.
5847 (ix86_split_simple_return_pop_internal): Likewise.
5848 * config/i386/i386.c (indirect_return_via_cx): New.
5849 (indirect_return_via_cx_bnd): Likewise.
5850 (indirect_thunk_name): Handle return va CX_REG.
5851 (output_indirect_thunk_function): Create alias for
5852 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5853 (ix86_output_indirect_jmp): Remove the bool argument.
5854 (ix86_output_indirect_function_return): New function.
5855 (ix86_split_simple_return_pop_internal): Likewise.
5856 * config/i386/i386.md (*indirect_jump): Don't pass false
5857 to ix86_output_indirect_jmp.
5858 (*tablejump_1): Likewise.
5859 (simple_return_pop_internal): Change it to define_insn_and_split.
5860 Call ix86_split_simple_return_pop_internal to split it for
5862 (simple_return_indirect_internal): Call
5863 ix86_output_indirect_function_return instead of
5864 ix86_output_indirect_jmp.
5866 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5869 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5870 memset and value initialization afterwards.
5872 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5874 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5876 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5879 * common/config/aarch64/aarch64-common.c
5880 (aarch_option_optimization_table[]): Switch
5881 off fomit-frame-pointer
5883 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5884 Chung-Ju Wu <jasonwucj@gmail.com>
5886 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5888 (store_multiple): Ditto.
5890 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5892 * config.gcc: Add --with-cpu support for nds32 target.
5893 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5894 * config/nds32/nds32.opt: Add -mcpu= option.
5896 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5898 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5899 isel=yes): Warn for these deprecated options.
5901 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5903 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5904 ISA_2_5_MASKS_EMBEDDED.
5906 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5908 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5909 p->max as pointers rather than using iterative_hash_expr.
5911 2018-02-23 Carl Love <cel@us.ibm.com>
5913 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5914 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5916 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5917 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5918 P8V_BUILTIN_VEC_VUNSIGNED2.
5920 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5923 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5924 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5926 * lra-constraints.c (curr_insn_transform): Set up
5927 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5929 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5930 * lra-eliminations.c (spill_pseudos): Ditto.
5931 (process_insn_for_elimination): Ditto.
5932 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5933 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5934 LRA_NON_CLOBBERED_ALT.
5936 2018-02-22 Martin Sebor <msebor@redhat.com>
5938 PR tree-optimization/84480
5939 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5940 to maybe_diag_stxncpy_trunc. Call it.
5941 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5942 from gimple_fold_builtin_strcpy. Print inlining stack.
5943 (handle_builtin_stxncpy): Print inlining stack.
5944 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5946 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5949 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5950 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5951 and -fcheck-pointer-bounds are used together.
5952 (indirect_thunk_prefix): New enum.
5953 (indirect_thunk_need_prefix): New function.
5954 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5955 "_nt" instead of "_bnd" for NOTRACK prefix.
5956 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5957 (output_indirect_thunk_function): Likewise.
5959 (ix86_code_end): Update output_indirect_thunk_function calls.
5960 (ix86_output_indirect_branch_via_reg): Replace
5961 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5962 (ix86_output_indirect_branch_via_push): Likewise.
5963 (ix86_output_function_return): Likewise.
5964 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5965 incompatible with -fcf-protection=branch and
5966 -fcheck-pointer-bounds.
5968 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5971 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5972 Change gcc_assert call to output_operand_lossage.
5974 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5976 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5978 2018-02-22 DJ Delorie <dj@redhat.com>
5979 Sebastian Perta <sebastian.perta@renesas.com>
5980 Oleg Endo <olegendo@gcc.gnu.org>
5982 * config/rx/rx.c (rx_rtx_costs): New function.
5983 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5985 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5987 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5989 2018-02-22 Martin Liska <mliska@suse.cz>
5992 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5993 Add "native" as a possible value.
5995 2018-02-22 Martin Liska <mliska@suse.cz>
5998 * config/i386/i386.c (ix86_option_override_internal):
5999 Add "native" as a possible value for -march and -mtune.
6001 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6004 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6005 to all type variants.
6007 PR tree-optimization/84503
6008 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6009 width as info->bitpos + info->bitsize - start.
6010 (merged_store_group::merge_overlapping): Simplify width computation.
6011 (check_no_overlap): New function.
6012 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6013 start + width and last_order of the group, fail if check_no_overlap
6015 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6016 to group if check_no_overlap fails.
6018 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6020 * config/rs6000/altivec.md: Delete contraint arguments to
6021 define_expand, define_split, and define_peephole2, and in
6022 define_insn_and_split if always unused.
6023 * config/rs6000/darwin.md: Ditto.
6024 * config/rs6000/dfp.md: Ditto.
6025 * config/rs6000/rs6000.md: Ditto.
6026 * config/rs6000/sync.md: Ditto.
6027 * config/rs6000/vector.md: Ditto.
6028 * config/rs6000/vsx.md: Ditto.
6030 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6032 * config/rs6000/altivec.md: Write output control strings as braced
6033 blocks instead of double-quoted strings.
6034 * config/rs6000/darwin.md: Ditto.
6035 * config/rs6000/rs6000.md: Ditto.
6036 * config/rs6000/vector.md: Ditto.
6037 * config/rs6000/vsx.md: Ditto.
6039 2018-02-21 Jason Merrill <jason@redhat.com>
6041 PR c++/84314 - ICE with templates and fastcall attribute.
6042 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6044 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6046 * ipa-cp.c (determine_versionability): Fix comment typos.
6048 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6051 * ipa-cp.c (determine_versionability): Do not version functions caling
6054 2018-02-21 Martin Liska <mliska@suse.cz>
6057 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6058 Add "native" as a possible value.
6059 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6060 the macro when native cpu detection is available.
6062 2018-02-21 Martin Liska <mliska@suse.cz>
6065 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6066 Add "native" as a possible value.
6067 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6068 when native cpu detection is available.
6070 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6071 Martin Sebor <msebor@redhat.com>
6073 PR tree-optimization/84478
6074 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6076 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6077 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6078 support which is conservatively correct, for 2 only stay conservative
6079 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6080 argument to the 2 argument get_range_strlen, adjust 6 arg
6081 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6083 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6084 (gimple_fold_builtin_strlen): Pass true as last argument to
6087 2018-02-20 Martin Sebor <msebor@redhat.com>
6090 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6091 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6092 (builtin_memref::builtin_memref): Factor out parts into
6093 set_base_and_offset and call it.
6095 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6098 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6099 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6100 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6101 search at the associated MODE_INT.
6103 2018-02-20 Jeff Law <law@redhat.com>
6106 PR tree-optimization/81592
6108 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6109 for range data rather than using global data.
6110 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6111 range data rather than using global data.
6112 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6113 pass it to children as needed.
6114 (struct directive::fmtresult): Similarly.
6115 (struct directive::set_width): Similarly.
6116 (struct directive::set_precision): Similarly.
6117 (format_integer, format_directive, parse_directive): Similarly.
6118 (format_none): Accept unnamed vr_values parameter.
6119 (format_percent, format_floating, format_character): Similarly.
6120 (format_string, format_plain): Similarly.
6121 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6122 the EVRP range analyzer for range data rather than using global data.
6123 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6124 gimple-ssa-evrp-analyze.h
6125 (class sprintf_dom_walker): Add after_dom_children member function.
6126 Add evrp_range_analyzer member.
6127 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6128 range analyzer as needed.
6129 (sprintf_dom_walker::after_dom_children): New member function.
6130 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6132 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6133 (evrp_range_analyzer::pop_to_marker): Likewise.
6135 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6137 PR tree-optimization/84419
6138 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6139 with the required type if its current type is compatible but
6142 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6145 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6146 after vectorization.
6148 2018-02-20 Martin Liska <mliska@suse.cz>
6151 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6152 possible values if we don't have a hint.
6154 2018-02-20 Martin Liska <mliska@suse.cz>
6158 * final.c (shorten_branches): Build align_tab array with one
6160 * opts.c (finish_options): Add alignment option limit check.
6161 (MAX_CODE_ALIGN): Likewise.
6162 (MAX_CODE_ALIGN_VALUE): Likewise.
6163 * doc/invoke.texi: Document maximum allowed option value for
6164 all -falign-* options.
6166 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6169 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6170 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6171 * var-tracking.c (emit_note_insn_var_location): Remove all references
6172 to NOTE_INSN_CALL_ARG_LOCATION.
6173 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6174 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6175 Use copy_rtx_if_shared.
6176 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6177 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6178 (dwarf2out_var_location): Remove handling of
6179 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6181 * final.c (final_scan_insn): Remove all references to
6182 NOTE_INSN_CALL_ARG_LOCATION.
6183 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6184 before dumping final insns.
6185 * except.c (emit_note_eh_region_end): Remove all references to
6186 NOTE_INSN_CALL_ARG_LOCATION.
6187 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6188 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6189 * config/arc/arc.c (hwloop_optimize): Likewise.
6190 * config/arm/arm.c (create_fix_barrier): Likewise.
6191 * config/s390/s390.c (s390_chunkify_start): Likewise.
6192 * config/sh/sh.c (find_barrier): Likewise.
6193 * config/i386/i386.c (rest_of_insert_endbranch,
6194 ix86_seh_fixup_eh_fallthru): Likewise.
6195 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6196 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6197 * config/frv/frv.c (frv_function_prologue): Likewise.
6198 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6200 (note_outside_basic_block_p): Remove all references to
6201 NOTE_INSN_CALL_ARG_LOCATION.
6202 * gengtype.c (adjust_field_rtx_def): Likewise.
6203 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6205 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6206 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6209 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6212 PR tree-optimization/84452
6213 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6214 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6217 2018-02-19 Martin Liska <mliska@suse.cz>
6220 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6222 2018-02-19 Martin Liska <mliska@suse.cz>
6223 Richard Sandiford <richard.sandiford@linaro.org>
6225 PR tree-optimization/82491
6226 * gimple-fold.c (get_base_constructor): Make earlier bail out
6229 2018-02-19 Carl Love <cel@us.ibm.com>
6231 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6232 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6234 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6235 P8V_BUILTIN_VEC_NEG.
6237 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6239 * config/rl78/rl78.md (movdf): New define expand.
6241 2018-02-19 Martin Liska <mliska@suse.cz>
6244 * doc/invoke.texi: Fix typo.
6245 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6247 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6249 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6250 handle rs6000_single_float and rs6000_double_float specially for
6253 2018-02-16 Jeff Law <law@redhat.com>
6255 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6258 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6261 * ipa-inline.c (inline_small_functions): Fix a typo.
6263 2018-02-16 Nathan Sidwell <nathan@acm.org>
6265 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6267 2018-02-16 Carl Love <cel@us.ibm.com>
6269 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6270 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6271 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6272 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6273 expansion to P8V_BUILTIN_VEC_FLOAT2.
6275 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6277 PR rtl-optimization/70023
6278 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6279 src_regno into account.
6281 2018-02-16 Carl Love <cel@us.ibm.com>
6283 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6284 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6285 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6286 * config/rs6000/rs6000.c: Remove case statements for
6287 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6288 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6289 and P9V_BUILTIN_VEC_VINSERT4B.
6290 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6291 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6292 * config/rs6000/vsx.md:
6293 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6296 2018-02-16 Carl Love <cel@us.ibm.com>
6298 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6300 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6302 * config/rs6000/rs6000-c.c: Add the definitions for
6303 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6304 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6305 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6306 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6307 definition for insert4b and define insn *insert3b_internal.
6308 * doc/extend.texi: Add documentation for vec_extract4b.
6310 2018-02-16 Nathan Sidwell <nathan@acm.org>
6312 * doc/extend.texi (Backwards Compatibility): Mention friend
6313 injection. Note for-scope is deprecated.
6314 * doc/invoke.texi (-ffriend-injection): Deprecate.
6316 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6318 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6319 that moved to I2, also allow destinations that are a paradoxical
6320 subreg (instead of a normal reg).
6322 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6325 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6328 2018-02-16 Richard Biener <rguenther@suse.de>
6330 PR tree-optimization/84037
6331 PR tree-optimization/84016
6333 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6334 Adjust vec_construct for the fact we need additional higher latency
6335 128bit inserts for AVX256 and AVX512 vector builds.
6336 (ix86_add_stmt_cost): Scale vector construction cost for
6339 2018-02-16 Richard Biener <rguenther@suse.de>
6341 PR tree-optimization/84417
6342 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6343 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6344 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6346 2018-02-16 Martin Liska <mliska@suse.cz>
6349 * internal-fn.def (ASAN_CHECK): Set proper flags.
6350 (ASAN_MARK): Likewise.
6352 2018-02-16 Julia Koval <julia.koval@intel.com>
6354 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6355 from PTA_CANNONLAKE.
6357 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6360 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6361 Use ++iter rather than iter++ for std::list iterators.
6362 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6363 defer deleting them until all nodes in the forest are processed. Do
6364 free even leaf nodes. Change to_process into auto_vec.
6367 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6368 * vec.h (vec_default_construct): Use memset instead of placement new
6369 if BROKEN_VALUE_INITIALIZATION is defined.
6370 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6371 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6374 PR rtl-optimization/83723
6375 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6376 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6377 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6379 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6381 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6383 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6385 PR rtl-optimization/81443
6386 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6387 from inner REGs to paradoxical SUBREGs.
6389 2018-02-16 Richard Biener <rguenther@suse.de>
6391 PR tree-optimization/84399
6392 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6393 For operands we can analyze at their definition make sure we can
6394 analyze them at each use as well.
6396 2018-02-16 Richard Biener <rguenther@suse.de>
6398 PR tree-optimization/84190
6399 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6400 volatile accesses if the decl isn't volatile.
6402 2018-02-15 Jason Merrill <jason@redhat.com>
6404 PR c++/84314 - ICE with templates and fastcall attribute.
6405 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6406 TYPE_CANONICAL on an existing type.
6408 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6410 PR tree-optimization/84383
6411 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6412 dstoff nor call operand_equal_p if dstbase is NULL.
6414 PR tree-optimization/84334
6415 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6416 also a CONSTANT_CLASS_P, punt.
6418 2018-02-14 Jim Wilson <jimw@sifive.com>
6420 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6421 first SMALL_OPERAND check. New local min_second_step. Move assert
6422 to where locals are set. Add TARGET_RVC support.
6423 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6425 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6427 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6429 2018-02-14 Martin Sebor <msebor@redhat.com>
6431 PR tree-optimization/83698
6432 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6433 arrays constrain the offset range to their bounds.
6434 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6435 (builtin_access::overlap): Avoid setting the size of overlap if it's
6437 (maybe_diag_overlap): Also consider arrays when deciding what values
6438 of offsets to include in diagnostics.
6440 2018-02-14 Martin Sebor <msebor@redhat.com>
6443 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6444 that correspond to the kind of a declaration.
6446 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6449 * config/pa/pa.md: Load address of PIC label using the linkage table
6450 if the label is nonlocal.
6452 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6454 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6455 warning message if user requests -maltivec=be.
6456 * doc/invoke.texi: Document deprecation of -maltivec=be.
6458 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6461 * config/rs6000/rs6000-c.c: Update definitions for
6462 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6463 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6465 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6468 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6469 add _get_ssp intrinsics. Remove argument from
6470 __builtin_ia32_rdssp[d|q].
6471 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6472 * config/i386/i386-builtin.def: Remove argument from
6473 __builtin_ia32_rdssp[d|q].
6474 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6475 ix86_expand_special_args_builtin for _rdssp[d|q].
6476 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6477 Clear register before usage.
6478 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6479 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6481 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6483 PR tree-optimization/84357
6484 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6485 operand 1 of an ARRAY_REF too.
6487 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6490 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6491 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6493 (set_of_reg): New struct.
6494 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6495 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6496 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6498 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6499 Split into bitclr, bitset, bitinvert patterns if appropriate.
6500 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6501 use rx_fuse_in_memory_bitop.
6502 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6503 to named insn, correct maximum insn length.
6505 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6508 * machmode.def: Define a complex mode for PARTIAL_INT.
6509 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6511 * doc/rtl.texi: Document CSPImode.
6512 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6514 (msp430_hard_regno_nregs_with_padding): Likewise.
6516 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6519 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6521 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6523 PR rtl-optimization/84169
6524 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6525 we generated a parallel as new i3 and we split that to new i2 and i3
6526 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6527 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6528 those to i2, not i1. Partially rewrite this scan code.
6530 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6533 * stor-layout.c (place_field): For variable length fields, adjust
6534 offset_align afterwards not just based on the field's alignment,
6535 but also on the size.
6538 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6539 of exps and logs in the use_exp2 case.
6541 2018-02-13 Jeff Law <law@redhat.com>
6543 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6546 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6549 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6553 * common.opt (gas-loc-support, gas-locview-support): New.
6554 (ginline-points, ginternal-reset-location-views): New.
6555 * doc/invoke.texi: Document them. Use @itemx where intended.
6556 (gvariable-location-views): Adjust.
6557 * target.def (reset_location_view): New.
6558 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6559 (TARGET_RESET_LOCATION_VIEW): New.
6560 * doc/tm.texi: Rebuilt.
6561 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6562 (dwarf2out_default_as_locview_support): New.
6563 (output_asm_line_debug_info): Use option variables.
6564 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6565 (output_loc_list): Likewise.
6566 (add_high_low_attributes): Check option variables.
6567 Don't output entry view attribute in strict mode.
6568 (gen_inlined_subroutine_die): Check option variables.
6569 (dwarf2out_inline_entry): Likewise.
6570 (init_sections_and_labels): Likewise.
6571 (dwarf2out_early_finish): Likewise.
6572 (maybe_reset_location_view): New, from...
6573 (dwarf2out_var_location): ... here. Call it.
6574 * debug.h (dwarf2out_default_as_loc_support): Declare.
6575 (dwarf2out_default_as_locview_support): Declare.
6576 * hooks.c (hook_int_rtx_insn_0): New.
6577 * hooks.h (hook_int_rtx_insn_0): Declare.
6578 * toplev.c (process_options): Take -gas-loc-support and
6579 -gas-locview-support from dwarf2out. Enable
6580 -gvariable-location-views by default only with locview
6581 assembler support. Enable -ginternal-reset-location-views by
6582 default only if the target defines the corresponding hook.
6583 Enable -ginline-points by default if location views are
6584 enabled; force it disabled if statement frontiers are
6586 * tree-inline.c (expand_call_inline): Check option variables.
6587 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6589 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6591 PR tree-optimization/84321
6592 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6593 handling. Also check whether the anti-range contains any values
6594 that satisfy the mask; switch to a VR_RANGE if not.
6596 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6599 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6601 2018-02-13 Martin Jambor <mjambor@suse.cz>
6604 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6605 of call statements, also set location of a load to a temporary.
6607 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6609 * config/rl78/rl78.c (add_vector_labels): New function.
6610 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6611 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6612 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6613 which checks that no arguments are passed.
6614 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6615 * doc/extend.texi: Documentation for the new attribute.
6617 2018-02-13 Andreas Schwab <schwab@suse.de>
6619 * config/riscv/linux.h (CPP_SPEC): Define.
6621 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6624 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6625 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6626 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6627 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6628 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6629 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6630 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6631 temporarily for AES and PCLMUL builtins.
6633 PR tree-optimization/84339
6634 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6635 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6639 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6640 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6641 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6643 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6645 * omp-simd-clone.h: New file.
6646 * omp-simd-clone.c: Include omp-simd-clone.h.
6647 (expand_simd_clones): No longer static.
6648 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6649 cgraph.h and omp-simd-clone.h.
6650 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6651 (vect_recog_widen_shift_pattern): Formatting fix.
6652 (vect_pattern_recog_1): Don't check optab for calls.
6655 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6656 operands[2] into a REG before using gen_lowpart on it.
6658 2018-02-12 Jeff Law <law@redhat.com>
6661 * config/sh/sh.c (find_barrier): Consider a sibling call
6664 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6665 successfully back substituting a reg.
6667 2018-02-12 Richard Biener <rguenther@suse.de>
6669 PR tree-optimization/84037
6670 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6671 parameter, move visited init to caller.
6672 (vect_slp_analyze_operations): Separate cost from validity
6673 check, initialize visited once for all instances.
6674 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6676 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6678 (vect_model_promotion_demotion_cost): Likewise.
6679 (vectorizable_bswap): Guard cost modeling with !slp_node
6680 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6682 (vectorizable_call): Likewise.
6683 (vectorizable_conversion): Likewise.
6684 (vectorizable_assignment): Likewise.
6685 (vectorizable_shift): Likewise.
6686 (vectorizable_operation): Likewise.
6687 (vectorizable_store): Likewise.
6688 (vectorizable_load): Likewise.
6689 (vectorizable_condition): Likewise.
6690 (vectorizable_comparison): Likewise.
6692 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6695 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6696 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6697 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6699 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6702 * params.def (inline-min-speedup): Increase from 8 to 15.
6703 (max-inline-insns-auto): Decrease from 40 to 30.
6704 * ipa-split.c (consider_split): Add some buffer for function to
6705 be considered inlining candidate.
6706 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6709 2018-02-12 Richard Biener <rguenther@suse.de>
6711 PR tree-optimization/84037
6712 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6713 matched stmts if we cannot swap the non-matched ones.
6715 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6717 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6718 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6719 _mm_maskz_scalef_round_ss): New intrinsics.
6720 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6721 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6722 __builtin_ia32_scalefss_round): Remove.
6723 (__builtin_ia32_scalefsd_mask_round,
6724 __builtin_ia32_scalefss_mask_round): New intrinsics.
6725 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6726 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6727 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6728 "<round_constraint>")): Changed to ...
6729 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6730 "<round_scalar_constraint>")): ... this.
6731 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6732 %0, %1, %2<round_op3>}"): Changed to ...
6733 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6734 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6735 %2<round_scalar_mask_op3>}"): ... this.
6736 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6738 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6740 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6741 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6742 (_mm_maskz_sqrt_round_ss): New intrinsics.
6743 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6744 (__builtin_ia32_sqrtsd_mask_round)
6745 (__builtin_ia32_sqrtss_mask_round): New builtins.
6746 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6747 (__builtin_ia32_sqrtss_round): Remove.
6748 (__builtin_ia32_sqrtsd_mask_round)
6749 (__builtin_ia32_sqrtss_mask_round): New builtins.
6750 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6751 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6752 ((match_operand:VF_128 1 "vector_operand"
6753 "xBm,<round_constraint>")): Changed to ...
6754 ((match_operand:VF_128 1 "vector_operand"
6755 "xBm,<round_scalar_constraint>")): ... this.
6756 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6757 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6758 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6759 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6760 %<iptr>1<round_scalar_mask_op3>}): ... this.
6761 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6762 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6764 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6767 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6768 Cast vec_cmpeq result to correct type.
6769 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6770 Cast vec_cmpgt result to correct type.
6772 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6774 * final.c (final_scan_insn_1): Renamed from...
6775 (final_scan_insn): ... this. New wrapper, to recover
6776 seen from the outermost call in recursive ones.
6777 * config/sparc/sparc.c (output_return): Drop seen from call.
6778 (output_sibcall): Likewise.
6779 * config/visium/visium.c (output_branch): Likewise.
6781 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6783 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6786 2018-02-10 Alan Modra <amodra@gmail.com>
6789 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6790 Specify LR as an input.
6792 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6795 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6796 remove_member_access_dummy_vars): New functions.
6797 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6798 lower_omp_1, execute_lower_omp): Use them.
6800 PR rtl-optimization/84308
6801 * shrink-wrap.c (spread_components): Release todo vector.
6803 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6805 PR rtl-optimization/57193
6806 * ira-color.c (struct allocno_color_data): Add member
6807 conflict_allocno_hard_prefs.
6808 (update_conflict_allocno_hard_prefs): New.
6809 (bucket_allocno_compare_func): Add a preference based on
6810 conflict_allocno_hard_prefs.
6811 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6812 (color_allocnos): Remove a dead code. Initiate
6813 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6815 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6818 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6819 constraint from =wa to wa. Avoid a subreg on the output operand,
6820 instead use a pseudo and subreg it in a move.
6821 (p9_xxbrd_<mode>): Changed to ...
6822 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6823 (p9_xxbrd_v2df): New expander.
6824 (p9_xxbrw_<mode>): Changed to ...
6825 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6826 (p9_xxbrw_v4sf): New expander.
6828 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6830 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6832 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6835 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6836 multiply in 32-bit mode.
6837 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6838 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6841 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6843 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6844 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6845 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6846 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6848 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6851 * dwarf2out.c (is_trivial_indirect_ref): New function.
6852 (dwarf2out_late_global_decl): Do not generate a location
6853 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6854 and that are not defined in the current unit.
6856 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6858 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6859 instead of a libcall for UNORDERED.
6861 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6864 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6865 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6867 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6870 * config/s390/s390.c (s390_set_current_function): Invoke
6871 s390_indirect_branch_settings also if fndecl didn't change.
6873 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6875 * config/rs6000/rs6000.md (blockage): Set length to zero.
6877 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6879 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6881 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6884 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6885 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6889 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6890 PARALLEL incoming that failed vt_get_decl_and_offset check.
6893 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6894 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6896 (get_variable_section): For decls in named .bss* sections pass true as
6897 second argument to bss_initializer_p.
6899 2018-02-09 Marek Polacek <polacek@redhat.com>
6900 Jakub Jelinek <jakub@redhat.com>
6903 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6904 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6905 Sync some changes from cxx_fold_indirect_ref.
6907 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6909 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6911 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6912 (BLOCK_INLINE_ENTRY_LABEL): New.
6913 (dwarf2out_var_location): Disregard inline entry markers.
6914 (inline_entry_data): New struct.
6915 (inline_entry_data_hasher): New hashtable type.
6916 (inline_entry_data_hasher::hash): New.
6917 (inline_entry_data_hasher::equal): New.
6918 (inline_entry_data_table): New variable.
6919 (add_high_low_attributes): Add DW_AT_entry_pc and
6920 DW_AT_GNU_entry_view attributes if a pending entry is found
6921 in inline_entry_data_table. Add old entry_pc attribute only
6922 if debug nonbinding markers are disabled.
6923 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6924 markers are enabled.
6925 (block_within_block_p, dwarf2out_inline_entry): New.
6926 (dwarf2out_finish): Check that no entries remained in
6927 inline_entry_data_table.
6928 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6929 (final_scan_insn, notice_source_line): Likewise.
6930 (rest_of_clean_state): Skip inline entry markers.
6931 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6933 * gimple.c (gimple_build_debug_inline_entry): New.
6934 * gimple.h (enum gimple_debug_subcode): Add
6935 GIMPLE_DEBUG_INLINE_ENTRY.
6936 (gimple_build_debug_inline_entry): Declare.
6937 (gimple_debug_inline_entry_p): New.
6938 (gimple_debug_nonbind_marker_p): Adjust.
6939 * insn-notes.def (INLINE_ENTRY): New.
6940 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6941 inline entry marker notes.
6942 (print_insn): Likewise.
6943 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6944 (INSN_DEBUG_MARKER_KIND): Likewise.
6945 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6946 * tree-inline.c (expand_call_inline): Build and insert
6947 debug_inline_entry stmt.
6948 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6949 inline entry blocks early, if nonbind markers are enabled.
6950 (dump_scope_block): Dump fragment info.
6951 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6952 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6953 (gimple_build_debug_inline_entry): New.
6954 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6955 Enable/disable inline entry points too.
6956 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6957 (DEBUG_INSN): Describe inline entry markers.
6959 * common.opt (gvariable-location-views): New.
6960 (gvariable-location-views=incompat5): New.
6961 * config.in: Rebuilt.
6962 * configure: Rebuilt.
6963 * configure.ac: Test assembler for view support.
6964 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6965 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6966 * dwarf2out.c (var_loc_view): New typedef.
6967 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6968 (dwarf2out_locviews_in_attribute): New.
6969 (dwarf2out_locviews_in_loclist): New.
6970 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6971 (enum dw_line_info_opcode): Add LI_adv_address.
6972 (struct dw_line_info_table): Add view.
6973 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6974 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6975 (zero_view_p): New variable.
6976 (ZERO_VIEW_P): New macro.
6977 (output_asm_line_debug_info): New.
6978 (struct var_loc_node): Add view.
6979 (add_AT_view_list, AT_loc_list): New.
6980 (add_var_loc_to_decl): Add view param. Test it against last.
6981 (new_loc_list): Add view params. Record them.
6982 (AT_loc_list_ptr): Handle loc and view lists.
6983 (view_list_to_loc_list_val_node): New.
6984 (print_dw_val): Handle dw_val_class_view_list.
6985 (size_of_die): Likewise.
6986 (value_format): Likewise.
6987 (loc_list_has_views): New.
6988 (gen_llsym): Set vl_symbol too.
6989 (maybe_gen_llsym, skip_loc_list_entry): New.
6990 (dwarf2out_maybe_output_loclist_view_pair): New.
6991 (output_loc_list): Output view list or entries too.
6992 (output_view_list_offset): New.
6993 (output_die): Handle dw_val_class_view_list.
6994 (output_dwarf_version): New.
6995 (output_compilation_unit_header): Use it.
6996 (output_skeleton_debug_sections): Likewise.
6997 (output_rnglists, output_line_info): Likewise.
6998 (output_pubnames, output_aranges): Update version comments.
6999 (output_one_line_info_table): Output view numbers in asm comments.
7000 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7001 Call maybe_gen_llsym.
7002 (loc_list_from_tree_1): Adjust.
7003 (add_AT_location_description): Create view list attribute if
7004 needed, check it's absent otherwise.
7005 (convert_cfa_to_fb_loc_list): Adjust.
7006 (maybe_emit_file): Call output_asm_line_debug_info for test.
7007 (dwarf2out_var_location): Reset views as needed. Precompute
7008 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7009 attribute. Set view.
7010 (new_line_info_table): Reset next view.
7011 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7012 (dwarf2out_source_line): Likewise. Output view resets and labels to
7013 the assembler, or select appropriate line info opcodes.
7014 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7015 (optimize_string_length): Catch it. Adjust.
7016 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7017 dw_val_class_view_list, and remove it if no longer needed.
7018 (hash_loc_list): Hash view numbers.
7019 (loc_list_hasher::equal): Compare them.
7020 (optimize_location_lists): Check whether a view list symbol is
7021 needed, and whether the locview attribute is present, and
7022 whether they match. Remove the locview attribute if no longer
7024 (index_location_lists): Call skip_loc_list_entry for test.
7025 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7026 Use output_dwarf_version.
7027 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7028 (struct dw_val_node): Add val_view_list.
7029 * final.c (SEEN_NEXT_VIEW): New.
7030 (set_next_view_needed): New.
7031 (clear_next_view_needed): New.
7032 (maybe_output_next_view): New.
7033 (final_start_function): Rename to...
7034 (final_start_function_1): ... this. Take pointer to FIRST,
7035 add SEEN parameter. Emit param bindings in the initial view.
7036 (final_start_function): Reintroduce SEEN-less interface.
7037 (final): Rename to...
7038 (final_1): ... this. Take SEEN parameter. Output final pending
7039 next view at the end.
7040 (final): Reintroduce seen-less interface.
7041 (final_scan_insn): Output pending next view before switching
7042 sections or ending a block. Mark the next view as needed when
7043 outputting variable locations. Notify debug backend of section
7044 changes, and of location view changes.
7045 (rest_of_handle_final): Adjust.
7046 * toplev.c (process_options): Autodetect value for debug variable
7047 location views option. Warn on incompat5 without -gdwarf-5.
7048 * doc/invoke.texi (gvariable-location-views): New.
7049 (gvariable-location-views=incompat5): New.
7050 (gno-variable-location-views): New.
7052 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7054 PR tree-optimization/84136
7055 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7056 that the result of find_edge is non-NULL.
7058 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7061 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7062 storing integer register in SImode. Fix cost of 256 and 512
7063 byte aligned SSE register store.
7065 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7067 * config/i386/i386.c (ix86_multiplication_cost): Fix
7068 multiplication cost for TARGET_AVX512DQ.
7070 2018-02-08 Marek Polacek <polacek@redhat.com>
7072 PR tree-optimization/84238
7073 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7076 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7078 PR tree-optimization/84265
7079 * tree-vect-stmts.c (vectorizable_store): Don't treat
7080 VMAT_CONTIGUOUS accesses as grouped.
7081 (vectorizable_load): Likewise.
7083 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7085 PR tree-optimization/81635
7086 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7087 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7088 (test_round_for_mask): New functions.
7089 (wide_int_cc_tests): Call test_round_for_mask.
7090 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7091 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7092 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7093 range returned by get_range_info.
7095 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7098 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7099 * symtab.c: Include builtins.h
7100 (symtab_node::output_to_lto_symbol_table_p): Move here
7101 from lto-streamer-out.c:output_symbol_p.
7102 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7103 (output_symbol_p): Move all logic to symtab.c
7104 (produce_symtab): Update.
7106 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7108 * config/s390/s390-opts.h (enum indirect_branch): Define.
7109 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7110 (s390_indirect_branch_via_thunk)
7111 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7112 (enum s390_indirect_branch_type): Define.
7113 * config/s390/s390.c (struct s390_frame_layout, struct
7114 machine_function): Remove.
7115 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7116 (indirect_branch_table_label_no, indirect_branch_table_name):
7118 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7119 (enum s390_indirect_branch_option): Define.
7120 (s390_return_addr_from_memory): New function.
7121 (s390_handle_string_attribute): New function.
7122 (s390_attribute_table): Add new attribute handler.
7123 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7124 (s390_indirect_branch_via_thunk): New function.
7125 (s390_indirect_branch_via_inline_thunk): New function.
7126 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7127 sibling call optimization for non z10 compiles.
7128 (s390_emit_call): Force indirect branch target to be a single
7129 register. Add r1 clobber for non-z10 compiles.
7130 (s390_emit_epilogue): Emit return jump via return_use expander.
7131 (s390_reorg): Handle JUMP_INSNs as execute targets.
7132 (s390_option_override_internal): Perform validity checks for the
7133 new command line options.
7134 (s390_indirect_branch_attrvalue): New function.
7135 (s390_indirect_branch_settings): New function.
7136 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7137 (s390_output_indirect_thunk_function): New function.
7138 (s390_code_end): Implement target hook.
7139 (s390_case_values_threshold): Implement target hook.
7140 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7142 * config/s390/s390.h (struct s390_frame_layout)
7143 (struct machine_function): Move here from s390.c.
7144 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7145 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7146 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7147 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7148 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7149 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7150 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7151 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7152 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7153 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7154 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7155 (mnemonic attribute): Add values which aren't recognized
7157 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7158 pattern for branch conversion. Fix mnemonic attribute.
7159 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7160 indirect branch via thunk if requested.
7161 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7162 ("*indirect_jump"): Disable for branch conversion using out of
7164 ("indirect_jump_via_thunk<mode>_z10")
7165 ("indirect_jump_via_thunk<mode>")
7166 ("indirect_jump_via_inlinethunk<mode>_z10")
7167 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7168 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7169 ("casesi_jump_via_inlinethunk<mode>_z10")
7170 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7171 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7172 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7173 ("*indirect2_jump"): Disable for branch conversion.
7174 ("casesi_jump"): Turn into expander and expand patterns for branch
7176 ("return_use"): New expander.
7177 ("*return"): Emit return via thunk and rename it to ...
7178 ("*return<mode>"): ... this one.
7179 * config/s390/s390.opt: Add new options and and enum for the
7182 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7184 * lra-constraints.c (match_reload): Unconditionally use
7185 gen_lowpart_SUBREG, rather than selecting between that
7186 and equivalent gen_rtx_SUBREG code.
7188 2018-02-08 Richard Biener <rguenther@suse.de>
7190 PR tree-optimization/84233
7191 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7192 changed flag instead of boguously re-using phi_inserted.
7194 2018-02-08 Martin Jambor <mjambor@suse.cz>
7196 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7197 static local variables.
7199 2018-02-08 Richard Biener <rguenther@suse.de>
7201 PR tree-optimization/84278
7202 * tree-vect-stmts.c (vectorizable_store): When looking for
7203 smaller vector types to perform grouped strided loads/stores
7204 make sure the mode is supported by the target.
7205 (vectorizable_load): Likewise.
7207 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7209 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7210 Increase LDP/STP opportunities by adding adjacent callee-saves.
7212 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7214 PR rtl-optimization/84068
7215 PR rtl-optimization/83459
7216 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7218 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7220 PR tree-optimization/84224
7221 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7222 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7225 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7228 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7229 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7230 count, remove one USE.
7232 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7234 * doc/install.texi (Configuration): Document the
7235 --with-long-double-format={ibm,ieee} PowerPC configuration
7239 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7240 Convert from define_expand to be define_insn_and_split. Rework
7241 float/double/_Float128 conversions to QI/HI/SImode to work with
7242 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7243 conversions to QI/HImode types did a store and then a load to
7244 truncate the value. For conversions to VSX registers, don't split
7245 the insn, instead emit the code directly. Use the code iterator
7246 any_fix to combine signed and unsigned conversions.
7247 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7248 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7249 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7250 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7251 (fix_<mode>di2_hw): Likewise.
7252 (fixuns_<mode>di2_hw): Likewise.
7253 (fix_<mode>si2_hw): Likewise.
7254 (fixuns_<mode>si2_hw): Likewise.
7255 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7256 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7257 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7258 fix<uns>_trunc<SFDF:mode>si2_p8.
7259 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7261 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7262 (fix<uns>_<mode>_mem): Likewise.
7263 (fctiw<u>z_<mode>_mem): Likewise.
7264 (fix<uns>_<mode>_mem): Likewise.
7265 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7266 the register allocator from doing a direct move to the GPRs to do
7267 a store, and instead use the ISA 3.0 store byte/half-word from
7268 vector register instruction. For IEEE 128-bit floating point,
7269 also optimize stores of 32-bit ints.
7270 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7272 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7274 * genextract.c (push_pathstr_operand): New function to support
7276 (walk_rtx): Call push_pathstr_operand.
7277 (print_path): Support [a-zA-Z].
7279 2018-02-07 Richard Biener <rguenther@suse.de>
7281 PR tree-optimization/84037
7282 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7283 (cse_and_gimplify_to_preheader): Declare.
7284 (vect_get_place_in_interleaving_chain): Likewise.
7285 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7287 (_loop_vec_info::~_loop_vec_info): Delete it.
7288 (cse_and_gimplify_to_preheader): New function.
7289 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7290 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7291 (vectorizable_load): Likewise. For grouped stores always base
7292 the IV on the first element.
7293 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7294 condition before gimplifying.
7296 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7298 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7299 *DIV_EXPR and *MOD_EXPR.
7301 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7304 * config/i386/i386.c (ix86_option_override_internal): Mask out
7305 the CF_SET bit when checking -fcf-protection.
7307 2018-02-07 Tom de Vries <tom@codesourcery.com>
7310 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7313 2018-02-07 Richard Biener <rguenther@suse.de>
7315 PR tree-optimization/84204
7316 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7319 PR tree-optimization/84205
7320 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7321 special-case isl_ast_op_zdiv_r.
7323 PR tree-optimization/84223
7324 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7325 Only add conditions from within the region.
7326 (gather_bbs::after_dom_children): Adjust.
7328 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7331 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7332 * config/avr/avr.md: Only post-reload split REG-REG moves if
7333 either register is GENERAL_REG_P.
7335 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7337 PR tree-optimization/84235
7338 * tree-ssa-scopedtables.c
7339 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7340 if the subtraction is performed in floating point type where NaNs are
7341 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7342 build 1. Formatting fix.
7344 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7347 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7348 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7349 and skip it regardless of bb boundaries. Use CALL_P macro,
7350 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7351 unnecessarily, formatting fix.
7353 2018-02-06 Michael Collison <michael.collison@arm.com>
7355 * config/arm/thumb2.md:
7356 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7357 (*thumb_mov_notscc): Ditto.
7359 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7362 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7363 unsigned_fix, not "s".
7365 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7367 * configure.ac (gcc_fn_eh_frame_ro): New function.
7368 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7369 correct .eh_frame permissions.
7370 * configure: Regenerate.
7372 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7374 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7377 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7379 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7380 Display warning message for -mno-speculate-indirect-jumps.
7382 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7384 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7386 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7388 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7390 PR tree-optimization/84225
7391 * tree-eh.c (find_trapping_overflow): Only call
7392 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7394 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7397 * config/i386/i386.c: Reimplement the check of possible options
7398 -mibt/-mshstk conbination. Change error messages.
7399 * doc/invoke.texi: Fix a typo: remove extra '='.
7401 2018-02-06 Marek Polacek <polacek@redhat.com>
7403 PR tree-optimization/84228
7404 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7406 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7409 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7410 emitted arch directives.
7411 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7412 __ARM_FEATURE_COPROC before changing architectures.
7414 2018-02-06 Richard Biener <rguenther@suse.de>
7416 * config/i386/i386.c (print_reg): Fix typo.
7417 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7419 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7421 * configure: Regenerate.
7423 2018-02-05 Martin Sebor <msebor@redhat.com>
7425 PR tree-optimization/83369
7426 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7429 2018-02-05 Martin Liska <mliska@suse.cz>
7431 * doc/invoke.texi: Cherry-pick upstream r323995.
7433 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7435 * ira.c (ira_init_register_move_cost): Adjust comment.
7437 2018-02-05 Martin Liska <mliska@suse.cz>
7439 PR gcov-profile/84137
7440 * doc/gcov.texi: Fix typo in documentation.
7442 2018-02-05 Martin Liska <mliska@suse.cz>
7444 PR gcov-profile/83879
7445 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7446 using dlopen functionality.
7448 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7450 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7451 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7452 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7453 _mm_maskz_range_round_ss): New intrinsics.
7454 (__builtin_ia32_rangesd128_round)
7455 (__builtin_ia32_rangess128_round): Remove.
7456 (__builtin_ia32_rangesd128_mask_round,
7457 __builtin_ia32_rangess128_mask_round): New builtins.
7458 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7459 __builtin_ia32_rangess128_round): Remove.
7460 (__builtin_ia32_rangesd128_mask_round,
7461 __builtin_ia32_rangess128_mask_round): New builtins.
7462 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7463 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7464 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7465 "<round_saeonly_constraint>")): Changed to ...
7466 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7467 "<round_saeonly_scalar_constraint>")): ... this.
7468 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7469 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7470 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7471 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7472 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7474 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7476 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7478 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7479 Remove all values except native, 8540 and 8548.
7481 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7483 * config/i386/i386.c (ix86_output_function_return): Pass
7484 INVALID_REGNUM, instead of -1, as invalid register number to
7485 indirect_thunk_name and output_indirect_thunk.
7487 2018-02-02 Julia Koval <julia.koval@intel.com>
7489 * config.gcc: Add -march=icelake.
7490 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7491 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7492 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7493 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7494 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7495 (processor_target_table): Add icelake.
7496 (ix86_option_override_internal): Handle new PTAs.
7497 (get_builtin_code_for_version): Handle icelake.
7498 (M_INTEL_COREI7_ICELAKE): New.
7499 (fold_builtin_cpu): Handle icelake.
7500 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7501 * doc/invoke.texi: Add -march=icelake.
7503 2018-02-02 Julia Koval <julia.koval@intel.com>
7505 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7506 to wide_int_bitmask.
7507 * wide-int-bitmask.h: New.
7509 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7512 * config/i386/i386.md: Replace Pmode with word_mode in
7513 builtin_setjmp_setup and builtin_longjmp to support x32.
7515 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7519 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7521 (rs6000_supported_cpu_names): New static variable.
7522 (linux_cpu_translation_table): Likewise.
7523 (elf_platform) <cpu>: Define new static variable and use it.
7524 Translate kernel AT_PLATFORM name to canonical name if needed.
7525 Error if platform name is unknown.
7527 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7530 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7532 2018-02-01 Jeff Law <law@redhat.com>
7535 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7536 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7537 the scratch if RELEASE_VIA_POP is false.
7538 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7539 If we have to save a temporary register, decrement SIZE appropriately.
7540 Pass new arguments to release_scratch_register_on_entry.
7541 (ix86_adjust_stack_and_probe): Likewise.
7542 (ix86_emit_probe_stack_range): Pass new arguments to
7543 release_scratch_register_on_entry.
7545 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7547 PR rtl-optimization/84157
7548 * combine.c (change_zero_ext): Use REG_P predicate in
7549 front of HARD_REGISTER_P predicate.
7551 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7553 * config/avr/avr.c (avr_option_override): Move disabling of
7554 -fdelete-null-pointer-checks to...
7555 * common/config/avr/avr-common.c (avr_option_optimization_table):
7558 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7560 PR tree-optimization/81635
7561 * tree-data-ref.c (split_constant_offset_1): For types that
7562 wrap on overflow, try to use range info to prove that wrapping
7565 2018-02-01 Renlin Li <renlin.li@arm.com>
7568 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7570 (aarch64_register_move_cost): Likewise.
7571 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7573 (REG_CLASS_NAMES): Likewise.
7574 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7575 TAILCALL_ADDR_REGS. Remove IP registers.
7576 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7578 2018-02-01 Richard Biener <rguenther@suse.de>
7580 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7581 for specifying RPO order and allow NULL for that.
7582 * domwalk.c (dom_walker::dom_walker): Likewise.
7583 (dom_walker::walk): Handle NULL RPO order.
7584 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7586 (rewrite_update_dom_walker): Likewise.
7587 (mark_def_dom_walker): Likewise.
7589 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7591 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7592 (aarch64_maybe_expand_sve_subreg_move): Declare.
7593 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7594 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7596 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7597 that are semantically a reverse operation.
7598 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7599 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7600 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7602 (aarch64_can_change_mode_class): For big-endian, forbid changes
7603 between two SVE modes if they have different element sizes.
7605 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7607 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7608 the TImode handling for big-endian targets.
7610 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7612 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7613 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7615 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7616 Remove BSWAP handing for big-endian targets and use the form of
7617 LD1RQ appropariate for the mode.
7619 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7621 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7622 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7625 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7628 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7629 check for operands that need to go through aarch64_sve_reload_be.
7631 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7633 PR tree-optimization/81661
7634 PR tree-optimization/84117
7635 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7636 * tree-eh.c: Include gimplify.h.
7637 (find_trapping_overflow, replace_trapping_overflow,
7638 rewrite_to_non_trapping_overflow): New functions.
7639 * tree-vect-loop.c: Include tree-eh.h.
7640 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7641 * tree-data-ref.c: Include tree-eh.h.
7642 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7644 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7646 PR rtl-optimization/84123
7647 * combine.c (change_zero_ext): Check if hard register satisfies
7648 can_change_dest_mode before calling gen_lowpart_SUBREG.
7650 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7653 * ira.c (ira_init_register_move_cost): Remove assert.
7655 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7657 PR rtl-optimization/84071
7658 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7659 * doc/tm.texi: Regenerate.
7661 2018-01-31 Richard Biener <rguenther@suse.de>
7663 PR tree-optimization/84132
7664 * tree-data-ref.c (analyze_miv_subscript): Properly
7665 check whether evolution_function_is_affine_multivariate_p
7666 before calling gcd_of_steps_may_divide_p.
7668 2018-01-31 Julia Koval <julia.koval@intel.com>
7671 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7672 * config/i386/i386.md (rdpid_rex64) New.
7673 (rdpid): Make 32bit only.
7675 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7678 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7679 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7681 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7684 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7686 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7688 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7690 PR rtl-optimization/84071
7691 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7692 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7694 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7696 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7697 (arc_attribute_table): Add 'aux' attribute.
7698 (arc_in_small_data_p): Consider aux like variables.
7699 (arc_is_aux_reg_p): New function.
7700 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7701 (arc_get_aux_arg): New function.
7702 (prepare_move_operands): Handle aux-register access.
7703 (arc_handle_aux_attribute): New function.
7704 * doc/extend.texi (ARC Variable attributes): Add subsection.
7706 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7708 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7709 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7710 (arc_attribute_table): Add 'uncached' attribute.
7711 (arc_print_operand): Print '.di' flag for uncached memory
7713 (arc_in_small_data_p): Do not consider for small data the uncached
7715 (arc_is_uncached_mem_p): New function.
7716 * config/arc/predicates.md (compact_store_memory_operand): Check
7717 for uncached memory accesses.
7718 (nonvol_nonimm_operand): Likewise.
7719 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7721 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7724 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7725 falign-loops=): Add Optimization flag.
7727 2018-01-30 Jeff Law <law@redhat.com>
7730 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7731 INT_REGISTERS_SAVED. Check it prior to calling
7732 get_scratch_register_on_entry.
7733 (ix86_adjust_stack_and_probe): Similarly.
7734 (ix86_emit_probe_stack_range): Similarly.
7735 (ix86_expand_prologue): Corresponding changes.
7737 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7740 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7741 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7743 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7746 * lra-constraints.c (curr_insn_transform): Process AND in the
7749 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7751 PR rtl-optimization/83986
7752 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7753 dependence against last_pending_memory_flush in addition to
7756 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7758 PR tree-optimization/81611
7759 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7762 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7765 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7768 2018-01-30 Richard Biener <rguenther@suse.de>
7769 Jakub Jelinek <jakub@redhat.com>
7771 PR tree-optimization/84111
7772 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7773 inner loops added during recursion, as they don't have up-to-date
7776 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7779 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7780 (can_inline_edge_by_limits_p): ... here.
7781 (can_early_inline_edge_p, check_callers,
7782 update_caller_keys, update_callee_keys, recursive_inlining,
7783 add_new_edges_to_heap, speculation_useful_p,
7784 inline_small_functions,
7785 inline_small_functions, flatten_function,
7786 inline_to_all_callers_1): Update.
7788 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7790 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7793 2018-01-30 Richard Biener <rguenther@suse.de>
7795 PR tree-optimization/83008
7796 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7797 invariant and constant vector uses in stmts when they need
7800 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7803 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7804 * configure: Regenerate.
7806 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7808 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7810 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7811 Use gen_rtx_REG rather than gen_lowpart.
7813 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7815 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7816 rather than 0 when creating partial subregs.
7818 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7820 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7823 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7826 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7827 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7828 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7829 flags. This restores the settings used before the 2017-07-24.
7830 Turning off pre increment/decrement/modify allows IVOPTS to
7831 optimize DF/SF loops where the index is an int.
7833 2018-01-29 Richard Biener <rguenther@suse.de>
7834 Kelvin Nilsen <kelvin@gcc.gnu.org>
7837 * tree-vect-stmts.c (vectorizable_call): Don't call
7838 targetm.vectorize_builtin_md_vectorized_function if callee is
7841 2018-01-22 Carl Love <cel@us.ibm.com>
7843 * doc/extend.tex: Fix typo in second arg in
7844 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7846 2018-01-29 Richard Biener <rguenther@suse.de>
7848 PR tree-optimization/84086
7849 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7850 (flush_ssaname_freelist): When SSA names were released reset
7851 the SCEV hash table.
7853 2018-01-29 Richard Biener <rguenther@suse.de>
7855 PR tree-optimization/84057
7856 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7857 removed paths when removing edges.
7859 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7861 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7862 -mfunction-return=@var{choice}.
7864 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7867 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7869 (layout::print_source_line): Likewise.
7870 (test_get_line_width_without_trailing_whitespace): Add test cases.
7872 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7875 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7878 2018-01-26 Jim Wilson <jimw@sifive.com>
7880 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7882 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7885 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7887 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7888 and CMP + SUB-immediate -> SUBS.
7890 2018-01-26 Martin Sebor <msebor@redhat.com>
7892 PR tree-optimization/83896
7893 * tree-ssa-strlen.c (get_string_len): Rename...
7894 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7895 Avoid assuming length is constant.
7896 (handle_char_store): Use HOST_WIDE_INT for string length.
7898 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7901 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7902 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7904 2018-01-26 Richard Biener <rguenther@suse.de>
7906 PR rtl-optimization/84003
7907 * dse.c (record_store): Only record redundant stores when
7908 the earlier store aliases at least all accesses the later one does.
7910 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7912 PR rtl-optimization/83985
7913 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7914 REG_CFA_RESTORE insns.
7915 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7916 REG_CFA_RESTORE insns here.
7919 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7920 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7922 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7924 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7925 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7926 (arc_init): Likewise.
7927 (arc_override_options): Likewise.
7928 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7930 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7932 * config/arc/arc.h (TARGET_DBNZ): Define.
7933 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7934 properly set the tune attribute.
7935 (dbnz): Use TARGET_DBNZ guard.
7936 * config/arc/arc.opt (mtune): Add core3 option.
7938 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7940 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7941 recognize new pic like addresses.
7942 (arc_delegitimize_address): Clean up.
7944 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7946 * config/arc/arc-arches.def: Option mrf16 valid for all
7948 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7949 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7950 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7951 * config/arc/arc-tables.opt: Regenerate.
7952 * config/arc/arc.c (arc_conditional_register_usage): Handle
7953 reduced register file case.
7954 (arc_file_start): Set must have build attributes.
7955 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7957 * config/arc/arc.opt (mrf16): Add new option.
7958 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7959 * config/arc/genmultilib.awk: Handle new mrf16 option.
7960 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7961 * config/arc/t-multilib: Regenerate.
7962 * doc/invoke.texi (ARC Options): Document mrf16 option.
7964 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7966 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7967 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7968 (arc_attribute_table): Add 'secure_call' attribute.
7969 (arc_print_operand): Print secure call operand.
7970 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7972 (arc_is_secure_call_p): New function. * config/arc/arc.md
7973 (call_i): Add support for sjli instruction.
7974 (call_value_i): Likewise.
7975 * config/arc/constraints.md (Csc): New constraint.
7977 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7978 John Eric Martin <John.Martin@emmicro-us.com>
7980 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7981 * config/arc/arc.c (_arc_jli_section): New struct.
7982 (arc_jli_section): New type.
7983 (rc_jli_sections): New static variable.
7984 (arc_handle_jli_attribute): New function.
7985 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7986 (arc_file_end): New function.
7987 (TARGET_ASM_FILE_END): Define.
7988 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7989 (arc_add_jli_section): New function.
7990 (jli_call_scan): Likewise.
7991 (arc_reorg): Call jli_call_scan.
7992 (arc_output_addsi): Remove 'S' from printing asm operand.
7993 (arc_is_jli_call_p): New function.
7994 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7996 (movhi_insn): Likewise.
7997 (movsi_insn): Likewise.
7998 (movsi_set_cc_insn): Likewise.
7999 (loadqi_update): Likewise.
8000 (load_zeroextendqisi_update): Likewise.
8001 (load_signextendqisi_update): Likewise.
8002 (loadhi_update): Likewise.
8003 (load_zeroextendhisi_update): Likewise.
8004 (load_signextendhisi_update): Likewise.
8005 (loadsi_update): Likewise.
8006 (loadsf_update): Likewise.
8007 (movsicc_insn): Likewise.
8008 (bset_insn): Likewise.
8009 (bxor_insn): Likewise.
8010 (bclr_insn): Likewise.
8011 (bmsk_insn): Likewise.
8012 (bicsi3_insn): Likewise.
8013 (cmpsi_cc_c_insn): Likewise.
8014 (movsi_ne): Likewise.
8015 (movsi_cond_exec): Likewise.
8016 (clrsbsi2): Likewise.
8027 (call_i): Remove 'S' asm letter, add jli instruction.
8028 (call_value_i): Likewise.
8029 * config/arc/arc.op (mjli-always): New option.
8030 * config/arc/constraints.md (Cji): New constraint.
8031 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8033 (subsf3_fpx): Likewise.
8034 (mulsf3_fpx): Likewise.
8035 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8037 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8039 * doc/invoke.texi (ARC): Document mjli-always option.
8041 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8043 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8044 avoid addition with 0 and use incw and decw where possible.
8046 2018-01-26 Richard Biener <rguenther@suse.de>
8048 PR tree-optimization/81082
8049 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8050 association if it requires casting to unsigned.
8051 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8052 from fold_plusminus_mult_expr to catch important cases late when
8053 range info is available.
8055 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8057 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8058 * configure.ac (hidden_linkonce): New test.
8059 * configure: Regenerate.
8060 * config.in: Regenerate.
8062 2018-01-26 Julia Koval <julia.koval@intel.com>
8064 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8065 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8066 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8067 _mm_mask_bitshuffle_epi64_mask): Fix type.
8068 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8069 USI_FTYPE_V4DI_V4DI_USI): Remove.
8070 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8071 __builtin_ia32_vpshufbitqmb256_mask,
8072 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8073 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8074 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8076 2018-01-26 Alan Modra <amodra@gmail.com>
8079 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8080 UNSPEC_VBPERMQ. Sort other unspecs.
8082 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8084 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8086 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8089 * predict.c (drop_profile): Do not push/pop cfun; update also
8091 (handle_missing_profiles): Fix logic looking for zero profiles.
8093 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8096 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8097 on functions with #pragma omp declare simd or functions with simd
8099 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8100 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8101 Remove trailing \n from warning_at calls.
8103 2018-01-25 Tom de Vries <tom@codesourcery.com>
8106 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8107 for neutered workers.
8109 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8112 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8113 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8115 2018-01-24 Jeff Law <law@redhat.com>
8118 * i386.c (get_probe_interval): Move to earlier point.
8119 (ix86_compute_frame_layout): If -fstack-clash-protection and
8120 the frame is larger than the probe interval, then use pushes
8121 to save registers rather than reg->mem moves.
8122 (ix86_expand_prologue): Remove conditional for int_registers_saved
8125 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8128 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8129 min/max for never referenced object.
8131 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8134 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8136 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8137 attributes from DECL_ATTRIBUTES (decl) without affecting
8138 DECL_ATTRIBUTES (current_function_decl).
8139 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8140 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8142 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8144 PR tree-optimization/83979
8145 * fold-const.c (fold_comparison): Use constant_boolean_node
8146 instead of boolean_{true,false}_node.
8148 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8150 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8153 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8155 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8156 Simplify the clause that sets the length attribute.
8157 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8158 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8159 clause that sets the length attribute.
8160 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8162 2018-01-24 Tom de Vries <tom@codesourcery.com>
8165 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8166 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8167 Add strict parameter.
8168 (prevent_branch_around_nothing): Insert dummy insn between branch to
8169 label and label with no ptx insn inbetween.
8170 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8172 2018-01-24 Tom de Vries <tom@codesourcery.com>
8175 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8176 for neutered threads in warp.
8177 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8179 2018-01-24 Richard Biener <rguenther@suse.de>
8181 PR tree-optimization/83176
8182 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8185 2018-01-24 Richard Biener <rguenther@suse.de>
8187 PR tree-optimization/82819
8188 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8189 code generating pluses that are no-ops in the target precision.
8191 2018-01-24 Richard Biener <rguenther@suse.de>
8194 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8196 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8198 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8199 to merge probabilities.
8200 * predict.c (probably_never_executed): Also mark as cold functions
8201 with global 0 profile and guessed local profile.
8202 * profile-count.c (profile_probability::combine_with_count): New
8204 * profile-count.h (profile_probability::operator*,
8205 profile_probability::operator*=, profile_probability::operator/,
8206 profile_probability::operator/=): Reduce precision to adjusted
8207 and set value to guessed on contradictory divisions.
8208 (profile_probability::combine_with_freq): Remove.
8209 (profile_probability::combine_wiht_count): Declare.
8210 (profile_count::force_nonzero):: Set to adjusted.
8211 (profile_count::probability_in):: Set quality to adjusted.
8212 * tree-ssa-tail-merge.c (replace_block_by): Use
8215 2018-01-23 Andrew Waterman <andrew@sifive.com>
8216 Jim Wilson <jimw@sifive.com>
8218 * config/riscv/riscv.c (riscv_stack_boundary): New.
8219 (riscv_option_override): Set riscv_stack_boundary. Handle
8220 riscv_preferred_stack_boundary_arg.
8221 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8222 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8223 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8224 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8225 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8226 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8228 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8231 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8232 of struct ix86_frame.
8233 (ix86_expand_epilogue): Likewise. Add a local variable for
8234 the reg_save_offset field in struct ix86_frame.
8236 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8238 PR tree-optimization/82604
8239 * tree-loop-distribution.c (enum partition_kind): New enum item
8240 PKIND_PARTIAL_MEMSET.
8241 (partition_builtin_p): Support above new enum item.
8242 (generate_code_for_partition): Ditto.
8243 (compute_access_range): Differentiate cases that equality can be
8244 proven at all loops, the innermost loops or no loops.
8245 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8246 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8247 (finalize_partitions, distribute_loop): Don't fuse partition of
8248 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8249 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8252 2018-01-23 Martin Liska <mliska@suse.cz>
8254 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8255 order to ignore the predictor.
8256 (PRED_POLYMORPHIC_CALL): Likewise.
8257 (PRED_RECURSIVE_CALL): Likewise.
8259 2018-01-23 Martin Liska <mliska@suse.cz>
8261 * tree-profile.c (tree_profiling): Print function header to
8262 aware reader which function we are working on.
8263 * value-prof.c (gimple_find_values_to_profile): Do not print
8264 not interesting value histograms.
8266 2018-01-23 Martin Liska <mliska@suse.cz>
8268 * profile-count.h (enum profile_quality): Add
8269 profile_uninitialized as the first value. Do not number values
8270 as they are zero based.
8271 (profile_count::verify): Update sanity check.
8272 (profile_probability::verify): Likewise.
8274 2018-01-23 Nathan Sidwell <nathan@acm.org>
8276 * doc/invoke.texi (ffor-scope): Deprecate.
8278 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8280 PR tree-optimization/83510
8281 * domwalk.c (set_all_edges_as_executable): New function.
8282 (dom_walker::dom_walker): Convert bool param
8283 "skip_unreachable_blocks" to enum reachability. Move setup of
8284 edge flags to set_all_edges_as_executable and only do it when
8285 reachability is REACHABLE_BLOCKS.
8286 * domwalk.h (enum dom_walker::reachability): New enum.
8287 (dom_walker::dom_walker): Convert bool param
8288 "skip_unreachable_blocks" to enum reachability.
8289 (set_all_edges_as_executable): New decl.
8290 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8291 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8293 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8294 but converting true to REACHABLE_BLOCKS.
8295 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8297 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8298 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8299 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8301 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8302 if check_all_array_refs will be called.
8304 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8306 * tree.c (selftest::test_location_wrappers): Add more test
8309 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8311 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8312 (selftest::test_bit_in_range): Likewise.
8314 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8317 * doc/sourcebuild.texi (vect_float): Say that the selector
8318 only describes the situation when -funsafe-math-optimizations is on.
8319 (vect_float_strict): Document.
8321 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8323 PR tree-optimization/83965
8324 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8325 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8326 instead of checking only for a reduction.
8327 (vect_recog_widen_sum_pattern): Likewise.
8329 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8331 * predict.c (probably_never_executed): Only use precise profile info.
8332 (compute_function_frequency): Skip after inlining hack since we now
8333 have quality checking.
8335 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8337 * profile-count.h (profile_probability::very_unlikely,
8338 profile_probability::unlikely, profile_probability::even): Set
8339 precision to guessed.
8341 2018-01-23 Richard Biener <rguenther@suse.de>
8343 PR tree-optimization/83963
8344 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8345 Properly terminate dominator walk when crossing the exit edge not
8346 when visiting its source block.
8348 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8351 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8352 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8354 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8356 PR tree-optimization/83957
8357 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8358 semicolon after for body surrounded by braces.
8360 PR tree-optimization/83081
8361 * profile-count.h (profile_probability::split): New method.
8362 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8363 Use profile_probability::split.
8364 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8365 when splitting a single conditional jump into 2.
8367 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8369 PR tree-optimization/69452
8370 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8373 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8375 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8376 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8377 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8379 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8381 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8382 * config/rl78/rl78.md (movdi): New define_expand.
8383 * config/rl78/rl78.c (rl78_split_movdi): New function.
8385 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8388 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8390 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8391 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8392 128-bit to produce an UNSPEC move to get the double word with the
8393 signbit and then a shift directly to do signbit.
8394 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8395 implementation with a new version that just does either a direct
8396 move or a regular move. Move memory interface to separate insns.
8397 Move insns so they are next to the expander.
8398 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8399 with signbit move. Split big and little endian case.
8400 (signbit<mode>2_dm_mem_le): Likewise.
8401 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8402 (signbit<mode>2_dm2): Likewise.
8404 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8406 * config/rl78/rl78.md (anddi3): New define_expand.
8408 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8410 * config/rl78/rl78.md (umindi3): New define_expand.
8412 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8414 * config/rl78/rl78.md (smindi3): New define_expand.
8416 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8418 * config/rl78/rl78.md (smaxdi3): New define_expand.
8420 2018-01-22 Carl Love <cel@us.ibm.com>
8422 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8423 LVX_V1TI): Add macro expansion.
8424 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8425 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8426 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8427 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8428 Change check to determine if the instruction is a byte reversing
8429 entry. Fix typo in comment.
8430 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8431 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8432 Add def_builtin calls for new builtins.
8433 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8434 Add define_insn expansion.
8436 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8438 * config/rl78/rl78.md (umaxdi3): New define_expand.
8440 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8442 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8443 for non-QImode registers.
8445 2018-01-22 Richard Biener <rguenther@suse.de>
8447 PR tree-optimization/83963
8448 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8449 including the loop exit block.
8450 (scop_detection::merge_sese): Likewise.
8451 (scop_detection::add_scop): Do it here instead.
8453 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8455 * doc/sourcebuild.texi (arm_softfloat): Document.
8457 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8460 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8461 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8462 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8464 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8465 David Edelsohn <dje.gcc@gmail.com>
8468 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8469 Change "crset eq" to "crset 2".
8470 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8471 (*call_indirect_aix<mode>_nospec): Likewise.
8472 (*call_value_indirect_aix<mode>_nospec): Likewise.
8473 (*call_indirect_elfv2<mode>_nospec): Likewise.
8474 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8475 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8476 change assembly output from . to $.
8477 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8478 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8479 (*tablejump<mode>_internal1_nospec): Likewise.
8481 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8484 * config/sh/sh_optimize_sett_clrt.cc:
8485 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8487 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8489 PR tree-optimization/83940
8490 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8491 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8492 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8493 use it to pass back the definition type.
8494 (vect_check_store_rhs): Likewise rhs_dt_out.
8495 (vect_build_gather_load_calls): Add a mask_dt argument and use
8496 it instead of a call to vect_is_simple_use.
8497 (vectorizable_store): Update calls to vect_check_load_store_mask
8498 and vect_check_store_rhs. Use the dt returned by the latter instead
8499 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8500 instead of calls to vect_is_simple_use. Pass the scalar rather
8501 than the vector operand to vect_is_simple_use when handling
8502 second and subsequent copies of an rhs value.
8503 (vectorizable_load): Update calls to vect_check_load_store_mask
8504 and vect_build_gather_load_calls. Use the cached mask_dt and
8505 gs_info.offset_dt instead of calls to vect_is_simple_use.
8507 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8510 * tree-emutls.c: Include gimplify.h.
8511 (lower_emutls_2): New function.
8512 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8513 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8514 it before further processing.
8517 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8518 UINTVAL (trueop1) instead of INTVAL (op1).
8520 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8524 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8525 INCOMING_FRAME_SP_OFFSET if not defined.
8526 (scan_trace): Add ENTRY argument. If true and
8527 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8528 emit a note to adjust the CFA offset.
8529 (create_cfi_notes): Adjust scan_trace callers.
8530 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8531 INCOMING_FRAME_SP_OFFSET in the CIE.
8532 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8533 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8535 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8536 * doc/tm.texi: Regenerated.
8538 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8540 PR rtl-optimization/83147
8541 * lra-constraints.c (remove_inheritance_pseudos): Use
8542 lra_substitute_pseudo_within_insn.
8544 2018-01-19 Tom de Vries <tom@codesourcery.com>
8545 Cesar Philippidis <cesar@codesourcery.com>
8548 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8550 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8553 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8554 spaces for function labels.
8556 2018-01-19 Martin Liska <mliska@suse.cz>
8558 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8559 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8560 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8561 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8562 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8563 (PRED_CONST_RETURN): Change from 69 to 65.
8564 (PRED_NULL_RETURN): Change from 91 to 71.
8565 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8566 (PRED_LOOP_GUARD): Change from 66 to 73.
8568 2018-01-19 Martin Liska <mliska@suse.cz>
8570 * predict.c (predict_insn_def): Add new assert.
8571 (struct branch_predictor): Change type to signed integer.
8572 (test_prediction_value_range): Amend test to cover
8574 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8575 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8576 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8577 (PRED_LOOP_IV_COMPARE): Likewise.
8578 * predict.h (PROB_UNINITIALIZED): Define new constant.
8580 2018-01-19 Martin Liska <mliska@suse.cz>
8582 * predict.c (dump_prediction): Add new format for
8583 analyze_brprob.py script which is enabled with -details
8585 * profile-count.h (precise_p): New function.
8587 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8589 PR tree-optimization/83922
8590 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8591 there are no statements that need masking.
8592 (vect_active_double_reduction_p): New function.
8593 (vect_analyze_loop_operations): Use it when handling phis that
8594 are not in the loop header.
8596 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8598 PR tree-optimization/83914
8599 * tree-vect-loop.c (vectorizable_induction): Don't convert
8600 init_expr or apply the peeling adjustment for inductions
8601 that are nested within the vectorized loop.
8603 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8605 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8608 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8612 * function.h (gimplify_parameters): Add gimple_seq * argument.
8613 * function.c: Include gimple.h and options.h.
8614 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8615 for the added local temporaries if needed.
8616 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8617 if there are any parameter cleanups, wrap whole body into a
8618 try/finally with the cleanups.
8620 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8623 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8624 Use GET_MODE_CLASS for scalar floating point.
8626 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8630 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8631 Fix call of call_cgraph_insertion_hooks.
8633 2018-01-18 Martin Sebor <msebor@redhat.com>
8635 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8637 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8640 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8643 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8646 * common.opt: (-ffile-prefix-map): New option.
8647 * opts.c (common_handle_option): Defer it.
8648 * opts-global.c (handle_common_deferred_options): Handle it.
8649 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8650 * file-prefix-map.h: New file.
8651 (remap_debug_filename, add_debug_prefix_map): ...here.
8652 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8653 * final.c (debug_prefix_map, add_debug_prefix_map
8654 remap_debug_filename): Move to...
8655 * file-prefix-map.c: New file.
8656 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8657 generalize, get rid of alloca(), use strrchr() instead of strchr().
8658 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8659 Implement in terms of add_prefix_map().
8660 (remap_macro_filename, remap_debug_filename): Implement in term of
8662 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8663 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8664 * dbxout.c: Include file-prefix-map.h.
8665 * varasm.c: Likewise.
8666 * vmsdbgout.c: Likewise.
8667 * xcoffout.c: Likewise.
8668 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8669 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8670 * doc/invoke.texi (-ffile-prefix-map): Document.
8671 (-fdebug-prefix-map): Update description.
8673 2018-01-18 Martin Liska <mliska@suse.cz>
8675 * config/i386/i386.c (indirect_thunk_name): Document that also
8677 (output_indirect_thunk): Document why both instructions
8678 (pause and lfence) are generated.
8680 2018-01-18 Richard Biener <rguenther@suse.de>
8682 PR tree-optimization/83887
8683 * graphite-scop-detection.c
8684 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8685 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8686 (scop_detection::merge_sese): Re-implement with a flood-fill
8687 algorithm that properly finds a SESE region if it exists.
8689 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8692 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8693 pointer_diff optimizations use view_convert instead of convert.
8695 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8697 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8698 Generate different code for -mno-speculate-indirect-jumps.
8699 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8700 (*call_indirect_aix<mode>): Disable for
8701 -mno-speculate-indirect-jumps.
8702 (*call_indirect_aix<mode>_nospec): New define_insn.
8703 (*call_value_indirect_aix<mode>): Disable for
8704 -mno-speculate-indirect-jumps.
8705 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8706 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8707 -mno-speculate-indirect-jumps.
8708 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8710 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8712 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8713 long double type, set the flags for noting the default long double
8714 type, even if we don't pass or return a long double type.
8716 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8719 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8722 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8724 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8725 support for merge[hl].
8726 (fold_mergehl_helper): New helper function.
8727 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8728 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8729 (altivec_vmrglw_direct): Add xxmrglw insn.
8731 2018-01-17 Andrew Waterman <andrew@sifive.com>
8733 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8734 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8736 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8739 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8740 call the lto_location_cache before reading the
8741 DECL_SOURCE_LOCATION of the types.
8743 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8744 Richard Sandiford <richard.sandiford@linaro.org>
8746 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8747 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8748 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8749 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8750 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8752 * config/aarch64/constraints.md (aarch64_movti_operand):
8754 * config/aarch64/predicates.md (Uti): Add new constraint.
8756 2018-01-17 Carl Love <cel@us.ibm.com>
8758 * config/rs6000/vsx.md (define_expand xl_len_r,
8759 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8760 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8762 (define_expand, define_insn): Move the shift left from the
8763 define_insn to the define_expand for lxvl and stxvl instructions.
8764 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8765 and XL_LEN_R definitions to PURE.
8767 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8769 * config/i386/i386.c (indirect_thunk_name): Declare regno
8770 as unsigned int. Compare regno with INVALID_REGNUM.
8771 (output_indirect_thunk): Ditto.
8772 (output_indirect_thunk_function): Ditto.
8773 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8774 in the call to output_indirect_thunk_function.
8776 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8779 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8780 rather than the size of inner_type to determine the stack slot size
8781 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8783 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8786 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8789 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8791 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8792 endian Linux systems to optionally enable multilibs for selecting
8793 the long double type if the user configured an explicit type.
8794 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8795 have no long double multilibs if not defined.
8796 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8797 warn if the user used -mabi={ieee,ibm}longdouble and we built
8798 multilibs for long double.
8799 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8800 appropriate multilib option.
8801 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8803 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8804 for building long double multilibs.
8805 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8807 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8809 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8812 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8814 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8817 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8820 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8823 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8825 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8826 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8828 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8830 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8831 different rtl trees depending on TARGET_64BIT.
8832 (rs6000_gen_lvx): Likewise.
8834 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8836 * config/visium/visium.md (nop): Tweak comment.
8837 (hazard_nop): Likewise.
8839 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8841 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8842 -mspeculate-indirect-jumps.
8843 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8844 for -mno-speculate-indirect-jumps.
8845 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8846 (*call_value_indirect_elfv2<mode>): Disable for
8847 -mno-speculate-indirect-jumps.
8848 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8849 (indirect_jump): Emit different RTL for
8850 -mno-speculate-indirect-jumps.
8851 (*indirect_jump<mode>): Disable for
8852 -mno-speculate-indirect-jumps.
8853 (*indirect_jump<mode>_nospec): New define_insn.
8854 (tablejump): Emit different RTL for
8855 -mno-speculate-indirect-jumps.
8856 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8857 (tablejumpsi_nospec): New define_expand.
8858 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8859 (tablejumpdi_nospec): New define_expand.
8860 (*tablejump<mode>_internal1): Disable for
8861 -mno-speculate-indirect-jumps.
8862 (*tablejump<mode>_internal1_nospec): New define_insn.
8863 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8866 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8868 * caller-save.c (insert_save): Drop unnecessary parameter. All
8871 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8872 Richard Biener <rguenth@suse.de>
8875 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8876 return early, inline manually is_gimple_sizepos. Make sure if we
8877 call gimplify_expr we don't end up with a gimple constant.
8878 * tree.c (variably_modified_type_p): Don't return true for
8879 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8880 * gimplify.h (is_gimple_sizepos): Remove.
8882 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8884 PR tree-optimization/83857
8885 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8886 vectorizable_live_operation for pure SLP statements.
8887 (vectorizable_live_operation): Handle PHIs.
8889 2018-01-16 Richard Biener <rguenther@suse.de>
8891 PR tree-optimization/83867
8892 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8893 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8895 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8898 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8899 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8900 If off is not INTEGER_CST, issue a may not be aligned warning
8901 rather than isn't aligned. Use isn%'t rather than isn't.
8902 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8904 <case MULT_EXPR>: Improve the case when bottom and one of the
8905 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8906 operand, in that case check if the other operand is multiple of
8907 bottom divided by the INTEGER_CST operand.
8909 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8912 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8913 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8914 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8915 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8916 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8917 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8918 (pa_function_arg_size): New function.
8920 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8922 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8923 in a separate statement.
8925 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8927 PR tree-optimization/83847
8928 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8929 group gathers and scatters.
8931 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8933 PR rtl-optimization/86620
8934 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8936 PR rtl-optimization/83213
8937 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8938 to last if both are JUMP_INSNs.
8940 PR tree-optimization/83843
8941 * gimple-ssa-store-merging.c
8942 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8943 store_immediate_info for bswap/nop orig_stores.
8945 2018-01-15 Andrew Waterman <andrew@sifive.com>
8947 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8949 <UDIV>: Increase cost if !TARGET_DIV.
8951 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8953 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8954 (define_attr "cr_logical_3op"): New.
8955 (cceq_ior_compare): Adjust.
8956 (cceq_ior_compare_complement): Adjust.
8957 (*cceq_rev_compare): Adjust.
8958 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8959 (is_cracked_insn): Adjust.
8960 (insn_must_be_first_in_group): Adjust.
8961 * config/rs6000/40x.md: Adjust.
8962 * config/rs6000/440.md: Adjust.
8963 * config/rs6000/476.md: Adjust.
8964 * config/rs6000/601.md: Adjust.
8965 * config/rs6000/603.md: Adjust.
8966 * config/rs6000/6xx.md: Adjust.
8967 * config/rs6000/7450.md: Adjust.
8968 * config/rs6000/7xx.md: Adjust.
8969 * config/rs6000/8540.md: Adjust.
8970 * config/rs6000/cell.md: Adjust.
8971 * config/rs6000/e300c2c3.md: Adjust.
8972 * config/rs6000/e500mc.md: Adjust.
8973 * config/rs6000/e500mc64.md: Adjust.
8974 * config/rs6000/e5500.md: Adjust.
8975 * config/rs6000/e6500.md: Adjust.
8976 * config/rs6000/mpc.md: Adjust.
8977 * config/rs6000/power4.md: Adjust.
8978 * config/rs6000/power5.md: Adjust.
8979 * config/rs6000/power6.md: Adjust.
8980 * config/rs6000/power7.md: Adjust.
8981 * config/rs6000/power8.md: Adjust.
8982 * config/rs6000/power9.md: Adjust.
8983 * config/rs6000/rs64.md: Adjust.
8984 * config/rs6000/titan.md: Adjust.
8986 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8988 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8989 ix86_indirect_branch_register logic.
8991 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8993 * config/i386/constraints.md (Bs): Update
8994 ix86_indirect_branch_register check. Don't check
8995 ix86_indirect_branch_register with GOT_memory_operand.
8997 * config/i386/predicates.md (GOT_memory_operand): Don't check
8998 ix86_indirect_branch_register here.
8999 (GOT32_symbol_operand): Likewise.
9001 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9003 * config/i386/predicates.md (constant_call_address_operand):
9004 Rewrite ix86_indirect_branch_register logic.
9005 (sibcall_insn_operand): Likewise.
9007 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9009 * config/i386/constraints.md (Bs): Replace
9010 ix86_indirect_branch_thunk_register with
9011 ix86_indirect_branch_register.
9013 * config/i386/i386.md (indirect_jump): Likewise.
9014 (tablejump): Likewise.
9015 (*sibcall_memory): Likewise.
9016 (*sibcall_value_memory): Likewise.
9017 Peepholes of indirect call and jump via memory: Likewise.
9018 * config/i386/i386.opt: Likewise.
9019 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9020 (GOT_memory_operand): Likewise.
9021 (call_insn_operand): Likewise.
9022 (sibcall_insn_operand): Likewise.
9023 (GOT32_symbol_operand): Likewise.
9025 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9028 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9029 type rather than type addr's type points to.
9030 (expand_omp_atomic_mutex): Likewise.
9031 (expand_omp_atomic): Likewise.
9033 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9036 * config/i386/i386.c (output_indirect_thunk_function): Use
9037 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9038 for __x86_return_thunk.
9040 2018-01-15 Richard Biener <rguenther@suse.de>
9043 * expmed.c (extract_bit_field_1): Fix typo.
9045 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9048 * config/arm/iterators.md (VF): New mode iterator.
9049 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9050 Remove integer-related logic from pattern.
9051 (neon_vabd<mode>_3): Likewise.
9053 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9056 * common.opt (fstrict-overflow): No longer an alias.
9057 (fwrapv-pointer): New option.
9058 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9059 also for pointer types based on flag_wrapv_pointer.
9060 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9061 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9062 opts->x_flag_wrapv got set.
9063 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9064 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9065 POINTER_TYPE_OVERFLOW_UNDEFINED.
9066 * match.pd: Likewise in address comparison pattern.
9067 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9069 2018-01-15 Richard Biener <rguenther@suse.de>
9072 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9073 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9074 Reset type names to their identifier if their TYPE_DECL doesn't
9075 have linkage (and thus is used for ODR and devirt).
9076 (save_debug_info_for_decl): Remove.
9077 (save_debug_info_for_type): Likewise.
9078 (add_tree_to_fld_list): Adjust.
9079 * tree-pretty-print.c (dump_generic_node): Make dumping of
9080 type names more robust.
9082 2018-01-15 Richard Biener <rguenther@suse.de>
9084 * BASE-VER: Bump to 8.0.1.
9086 2018-01-14 Martin Sebor <msebor@redhat.com>
9089 * builtins.c (check_access): Avoid warning when the no-warning bit
9092 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9094 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9095 * ira-color (allocno_hard_regs_compare): Likewise.
9097 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9100 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9101 Use .pushsection/.popsection.
9103 2018-01-14 Martin Sebor <msebor@redhat.com>
9106 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9108 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9110 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9111 entry from extra_headers.
9112 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9113 extra_headers, make the list bitwise identical to the i?86-*-* one.
9115 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9117 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9118 -mcmodel=large with -mindirect-branch=thunk,
9119 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9120 -mfunction-return=thunk-extern.
9121 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9122 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9123 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9125 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9127 * config/i386/i386.c (print_reg): Print the name of the full
9128 integer register without '%'.
9129 (ix86_print_operand): Handle 'V'.
9130 * doc/extend.texi: Document 'V' modifier.
9132 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9134 * config/i386/constraints.md (Bs): Disallow memory operand for
9135 -mindirect-branch-register.
9137 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9138 (GOT_memory_operand): Likewise.
9139 (call_insn_operand): Likewise.
9140 (sibcall_insn_operand): Likewise.
9141 (GOT32_symbol_operand): Likewise.
9142 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9143 for -mindirect-branch-register.
9144 (tablejump): Likewise.
9145 (*sibcall_memory): Likewise.
9146 (*sibcall_value_memory): Likewise.
9147 Disallow peepholes of indirect call and jump via memory for
9148 -mindirect-branch-register.
9149 (*call_pop): Replace m with Bw.
9150 (*call_value_pop): Likewise.
9151 (*sibcall_pop_memory): Replace m with Bs.
9152 * config/i386/i386.opt (mindirect-branch-register): New option.
9153 * doc/invoke.texi: Document -mindirect-branch-register option.
9155 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9157 * config/i386/i386-protos.h (ix86_output_function_return): New.
9158 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9159 set function_return_type.
9160 (indirect_thunk_name): Add ret_p to indicate thunk for function
9162 (output_indirect_thunk_function): Pass false to
9163 indirect_thunk_name.
9164 (ix86_output_indirect_branch_via_reg): Likewise.
9165 (ix86_output_indirect_branch_via_push): Likewise.
9166 (output_indirect_thunk_function): Create alias for function
9167 return thunk if regno < 0.
9168 (ix86_output_function_return): New function.
9169 (ix86_handle_fndecl_attribute): Handle function_return.
9170 (ix86_attribute_table): Add function_return.
9171 * config/i386/i386.h (machine_function): Add
9172 function_return_type.
9173 * config/i386/i386.md (simple_return_internal): Use
9174 ix86_output_function_return.
9175 (simple_return_internal_long): Likewise.
9176 * config/i386/i386.opt (mfunction-return=): New option.
9177 (indirect_branch): Mention -mfunction-return=.
9178 * doc/extend.texi: Document function_return function attribute.
9179 * doc/invoke.texi: Document -mfunction-return= option.
9181 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9183 * config/i386/i386-opts.h (indirect_branch): New.
9184 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9185 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9186 with local indirect jump when converting indirect call and jump.
9187 (ix86_set_indirect_branch_type): New.
9188 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9189 (indirectlabelno): New.
9190 (indirect_thunk_needed): Likewise.
9191 (indirect_thunk_bnd_needed): Likewise.
9192 (indirect_thunks_used): Likewise.
9193 (indirect_thunks_bnd_used): Likewise.
9194 (INDIRECT_LABEL): Likewise.
9195 (indirect_thunk_name): Likewise.
9196 (output_indirect_thunk): Likewise.
9197 (output_indirect_thunk_function): Likewise.
9198 (ix86_output_indirect_branch_via_reg): Likewise.
9199 (ix86_output_indirect_branch_via_push): Likewise.
9200 (ix86_output_indirect_branch): Likewise.
9201 (ix86_output_indirect_jmp): Likewise.
9202 (ix86_code_end): Call output_indirect_thunk_function if needed.
9203 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9205 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9206 (ix86_attribute_table): Add indirect_branch.
9207 * config/i386/i386.h (machine_function): Add indirect_branch_type
9208 and has_local_indirect_jump.
9209 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9211 (tablejump): Likewise.
9212 (*indirect_jump): Use ix86_output_indirect_jmp.
9213 (*tablejump_1): Likewise.
9214 (simple_return_indirect_internal): Likewise.
9215 * config/i386/i386.opt (mindirect-branch=): New option.
9216 (indirect_branch): New.
9219 (thunk-inline): Likewise.
9220 (thunk-extern): Likewise.
9221 * doc/extend.texi: Document indirect_branch function attribute.
9222 * doc/invoke.texi: Document -mindirect-branch= option.
9224 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9227 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9229 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9231 * ipa-inline.c (want_inline_small_function_p): Return false if
9232 inlining has already failed with CIF_FINAL_ERROR.
9233 (update_caller_keys): Call want_inline_small_function_p before
9235 (update_callee_keys): Likewise.
9237 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9239 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9241 (rs6000_quadword_masked_address_p): Likewise.
9242 (quad_aligned_load_p): Likewise.
9243 (quad_aligned_store_p): Likewise.
9244 (const_load_sequence_p): Add comment to describe the outer-most loop.
9245 (mimic_memory_attributes_and_flags): New function.
9246 (rs6000_gen_stvx): Likewise.
9247 (replace_swapped_aligned_store): Likewise.
9248 (rs6000_gen_lvx): Likewise.
9249 (replace_swapped_aligned_load): Likewise.
9250 (replace_swapped_load_constant): Capitalize argument name in
9251 comment describing this function.
9252 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9253 and stores that access quad-word aligned addresses and replace
9254 with stvx or lvx instructions when appropriate.
9255 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9256 New function prototype.
9257 (rs6000_quadword_masked_address_p): Likewise.
9258 (rs6000_gen_lvx): Likewise.
9259 (rs6000_gen_stvx): Likewise.
9260 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9261 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9262 when memory address is aligned.
9263 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9264 this split to select lvx instruction when memory address is aligned.
9265 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9266 instruction when memory address is aligned.
9267 (*vsx_le_perm_load_v16qi): Likewise.
9268 (four unnamed splitters): Modify to select the stvx instruction
9269 when memory is aligned.
9271 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9273 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9274 which appears in the queue multiple times.
9276 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9277 Alan Hayward <alan.hayward@arm.com>
9278 David Sherwood <david.sherwood@arm.com>
9280 * tree-vectorizer.h (vec_lower_bound): New structure.
9281 (_loop_vec_info): Add check_nonzero and lower_bounds.
9282 (LOOP_VINFO_CHECK_NONZERO): New macro.
9283 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9284 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9285 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9286 fields. Make seg_len the distance travelled, not including the
9288 (dr_direction_indicator): Declare.
9289 (dr_zero_step_indicator): Likewise.
9290 (dr_known_forward_stride_p): Likewise.
9291 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9293 (runtime_alias_check_p): Allow runtime alias checks with
9295 (operator ==): Compare access_size and align.
9296 (prune_runtime_alias_test_list): Rework for new distinction between
9297 the access_size and seg_len.
9298 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9300 (get_segment_min_max): New function.
9301 (create_intersect_range_checks): Use it.
9302 (dr_step_indicator): New function.
9303 (dr_direction_indicator): Likewise.
9304 (dr_zero_step_indicator): Likewise.
9305 (dr_known_forward_stride_p): Likewise.
9306 * tree-loop-distribution.c (data_ref_segment_size): Return
9307 DR_STEP * (niters - 1).
9308 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9310 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9311 (vect_preserves_scalar_order_p): New function, split out from...
9312 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9313 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9314 (vect_vfa_access_size): New function.
9315 (vect_vfa_align): Likewise.
9316 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9317 (dump_lower_bound): New function.
9318 (vect_check_lower_bound): Likewise.
9319 (vect_small_gap_p): Likewise.
9320 (vectorizable_with_step_bound_p): Likewise.
9321 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9322 depencies if the vectorization factor is 1. Convert the checks
9323 for nonzero steps into checks on the bounds of DR_STEP. Try using
9324 a bunds check for variable steps if the minimum required step is
9325 relatively small. Update calls to the dr_with_seg_len
9326 constructor and to vect_compile_time_alias.
9327 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9329 (vect_loop_versioning): Call it.
9330 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9332 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9334 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9335 Alan Hayward <alan.hayward@arm.com>
9336 David Sherwood <david.sherwood@arm.com>
9338 * doc/sourcebuild.texi (vect_scatter_store): Document.
9339 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9341 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9343 * genopinit.c (main): Add supports_vec_scatter_store and
9344 supports_vec_scatter_store_cached to target_optabs.
9345 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9346 IFN_MASK_SCATTER_STORE.
9347 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9349 * internal-fn.h (internal_store_fn_p): Declare.
9350 (internal_fn_stored_value_index): Likewise.
9351 * internal-fn.c (scatter_store_direct): New macro.
9352 (expand_scatter_store_optab_fn): New function.
9353 (direct_scatter_store_optab_supported_p): New macro.
9354 (internal_store_fn_p): New function.
9355 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9356 IFN_MASK_SCATTER_STORE.
9357 (internal_fn_mask_index): Likewise.
9358 (internal_fn_stored_value_index): New function.
9359 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9361 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9362 * optabs-query.c (supports_vec_scatter_store_p): New function.
9363 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9364 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9365 true for scatter stores.
9366 (vect_gather_scatter_fn_p): Handle scatter stores too.
9367 (vect_check_gather_scatter): Consider using scatter stores if
9368 supports_vec_scatter_store_p.
9369 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9371 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9372 internal_fn_stored_value_index.
9373 (check_load_store_masking): Handle scatter stores too.
9374 (vect_get_store_rhs): Make public.
9375 (vectorizable_call): Use internal_store_fn_p.
9376 (vectorizable_store): Handle scatter store internal functions.
9377 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9378 when deciding whether the end of the group has been reached.
9379 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9380 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9381 (mask_scatter_store<mode>): New insns.
9383 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9384 Alan Hayward <alan.hayward@arm.com>
9385 David Sherwood <david.sherwood@arm.com>
9387 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9388 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9389 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9391 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9392 Use vect_truncate_gather_scatter_offset if we can't treat the
9393 operation as a normal gather load or scatter store.
9394 (get_group_load_store_type): Take the gather_scatter_info
9395 as argument. Try using a gather load or scatter store for
9396 single-element groups.
9397 (get_load_store_type): Update calls to get_group_load_store_type
9398 and vect_use_strided_gather_scatters_p.
9400 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9401 Alan Hayward <alan.hayward@arm.com>
9402 David Sherwood <david.sherwood@arm.com>
9404 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9405 optional tree argument.
9406 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9408 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9409 but continue to use the current value as a fallback.
9410 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9411 to compare the updates.
9412 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9413 (get_load_store_type): Use it when handling a strided access.
9414 (vect_get_strided_load_store_ops): New function.
9415 (vect_get_data_ptr_increment): Likewise.
9416 (vectorizable_load): Handle strided gather loads. Always pass
9417 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9419 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9420 Alan Hayward <alan.hayward@arm.com>
9421 David Sherwood <david.sherwood@arm.com>
9423 * doc/md.texi (gather_load@var{m}): Document.
9424 (mask_gather_load@var{m}): Likewise.
9425 * genopinit.c (main): Add supports_vec_gather_load and
9426 supports_vec_gather_load_cached to target_optabs.
9427 * optabs-tree.c (init_tree_optimization_optabs): Use
9428 ggc_cleared_alloc to allocate target_optabs.
9429 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9430 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9432 * internal-fn.h (internal_load_fn_p): Declare.
9433 (internal_gather_scatter_fn_p): Likewise.
9434 (internal_fn_mask_index): Likewise.
9435 (internal_gather_scatter_fn_supported_p): Likewise.
9436 * internal-fn.c (gather_load_direct): New macro.
9437 (expand_gather_load_optab_fn): New function.
9438 (direct_gather_load_optab_supported_p): New macro.
9439 (direct_internal_fn_optab): New function.
9440 (internal_load_fn_p): Likewise.
9441 (internal_gather_scatter_fn_p): Likewise.
9442 (internal_fn_mask_index): Likewise.
9443 (internal_gather_scatter_fn_supported_p): Likewise.
9444 * optabs-query.c (supports_at_least_one_mode_p): New function.
9445 (supports_vec_gather_load_p): Likewise.
9446 * optabs-query.h (supports_vec_gather_load_p): Declare.
9447 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9448 and memory_type field.
9449 (NUM_PATTERNS): Bump to 15.
9450 * tree-vect-data-refs.c: Include internal-fn.h.
9451 (vect_gather_scatter_fn_p): New function.
9452 (vect_describe_gather_scatter_call): Likewise.
9453 (vect_check_gather_scatter): Try using internal functions for
9454 gather loads. Recognize existing calls to a gather load function.
9455 (vect_analyze_data_refs): Consider using gather loads if
9456 supports_vec_gather_load_p.
9457 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9458 (vect_get_gather_scatter_offset_type): Likewise.
9459 (vect_convert_mask_for_vectype): Likewise.
9460 (vect_add_conversion_to_patterm): Likewise.
9461 (vect_try_gather_scatter_pattern): Likewise.
9462 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9463 (vect_vect_recog_func_ptrs): Add it.
9464 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9465 internal_fn_mask_index and internal_gather_scatter_fn_p.
9466 (check_load_store_masking): Take the gather_scatter_info as an
9467 argument and handle gather loads.
9468 (vect_get_gather_scatter_ops): New function.
9469 (vectorizable_call): Check internal_load_fn_p.
9470 (vectorizable_load): Likewise. Handle gather load internal
9472 (vectorizable_store): Update call to check_load_store_masking.
9473 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9474 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9475 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9476 (aarch64_gather_scale_operand_d): New predicates.
9477 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9478 (mask_gather_load<mode>): New insns.
9480 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9481 Alan Hayward <alan.hayward@arm.com>
9482 David Sherwood <david.sherwood@arm.com>
9484 * optabs.def (fold_left_plus_optab): New optab.
9485 * doc/md.texi (fold_left_plus_@var{m}): Document.
9486 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9487 * internal-fn.c (fold_left_direct): Define.
9488 (expand_fold_left_optab_fn): Likewise.
9489 (direct_fold_left_optab_supported_p): Likewise.
9490 * fold-const-call.c (fold_const_fold_left): New function.
9491 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9492 * tree-parloops.c (valid_reduction_p): New function.
9493 (gather_scalar_reductions): Use it.
9494 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9495 (vect_finish_replace_stmt): Declare.
9496 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9497 (needs_fold_left_reduction_p): New function, split out from...
9498 (vect_is_simple_reduction): ...here. Accept reductions that
9499 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9500 (vect_force_simple_reduction): Also store the reduction type in
9501 the assignment's STMT_VINFO_REDUC_TYPE.
9502 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9503 (merge_with_identity): New function.
9504 (vect_expand_fold_left): Likewise.
9505 (vectorize_fold_left_reduction): Likewise.
9506 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9507 scalar phi in place for it. Check for target support and reject
9508 cases that would reassociate the operation. Defer the transform
9509 phase to vectorize_fold_left_reduction.
9510 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9511 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9512 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9514 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9516 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9517 call to ifc_temp_var.
9519 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9520 Alan Hayward <alan.hayward@arm.com>
9521 David Sherwood <david.sherwood@arm.com>
9523 * target.def (legitimize_address_displacement): Take the original
9524 offset as a poly_int.
9525 * targhooks.h (default_legitimize_address_displacement): Update
9527 * targhooks.c (default_legitimize_address_displacement): Likewise.
9528 * doc/tm.texi: Regenerate.
9529 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9530 as an argument, moving assert of ad->disp == ad->disp_term to...
9531 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9532 Try calling targetm.legitimize_address_displacement before expanding
9533 the address rather than afterwards, and adjust for the new interface.
9534 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9535 Match the new hook interface. Handle SVE addresses.
9536 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9539 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9541 * Makefile.in (OBJS): Add early-remat.o.
9542 * target.def (select_early_remat_modes): New hook.
9543 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9544 * doc/tm.texi: Regenerate.
9545 * targhooks.h (default_select_early_remat_modes): Declare.
9546 * targhooks.c (default_select_early_remat_modes): New function.
9547 * timevar.def (TV_EARLY_REMAT): New timevar.
9548 * passes.def (pass_early_remat): New pass.
9549 * tree-pass.h (make_pass_early_remat): Declare.
9550 * early-remat.c: New file.
9551 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9553 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9555 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9556 Alan Hayward <alan.hayward@arm.com>
9557 David Sherwood <david.sherwood@arm.com>
9559 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9560 vfm1 with a bound_epilog parameter.
9561 (vect_do_peeling): Update calls accordingly, and move the prologue
9562 call earlier in the function. Treat the base bound_epilog as 0 for
9563 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9564 this base when peeling for gaps.
9565 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9566 with fully-masked loops.
9567 (vect_estimate_min_profitable_iters): Handle the single peeled
9568 iteration in that case.
9570 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9571 Alan Hayward <alan.hayward@arm.com>
9572 David Sherwood <david.sherwood@arm.com>
9574 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9575 single-element interleaving even if the size is not a power of 2.
9576 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9577 accesses for single-element interleaving if the group size is
9580 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9581 Alan Hayward <alan.hayward@arm.com>
9582 David Sherwood <david.sherwood@arm.com>
9584 * doc/md.texi (fold_extract_last_@var{m}): Document.
9585 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9586 * optabs.def (fold_extract_last_optab): New optab.
9587 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9588 * internal-fn.c (fold_extract_direct): New macro.
9589 (expand_fold_extract_optab_fn): Likewise.
9590 (direct_fold_extract_optab_supported_p): Likewise.
9591 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9592 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9593 EXTRACT_LAST_REDUCTION.
9594 (get_initial_def_for_reduction): Do not create an initial vector
9595 for EXTRACT_LAST_REDUCTION reductions.
9596 (vectorizable_reduction): Leave the scalar phi in place for
9597 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9598 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9599 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9600 transform phase to vectorizable_condition.
9601 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9603 (vect_finish_stmt_generation): ...here.
9604 (vect_finish_replace_stmt): New function.
9605 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9606 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9608 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9610 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9611 Alan Hayward <alan.hayward@arm.com>
9612 David Sherwood <david.sherwood@arm.com>
9614 * doc/md.texi (extract_last_@var{m}): Document.
9615 * optabs.def (extract_last_optab): New optab.
9616 * internal-fn.def (EXTRACT_LAST): New internal function.
9617 * internal-fn.c (cond_unary_direct): New macro.
9618 (expand_cond_unary_optab_fn): Likewise.
9619 (direct_cond_unary_optab_supported_p): Likewise.
9620 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9621 loops using EXTRACT_LAST.
9622 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9623 (extract_last_<mode>): ...this optab.
9624 (vec_extract<mode><Vel>): Update accordingly.
9626 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9627 Alan Hayward <alan.hayward@arm.com>
9628 David Sherwood <david.sherwood@arm.com>
9630 * target.def (empty_mask_is_expensive): New hook.
9631 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9632 * doc/tm.texi: Regenerate.
9633 * targhooks.h (default_empty_mask_is_expensive): Declare.
9634 * targhooks.c (default_empty_mask_is_expensive): New function.
9635 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9636 if the target says that empty masks are expensive.
9637 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9639 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9641 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9642 Alan Hayward <alan.hayward@arm.com>
9643 David Sherwood <david.sherwood@arm.com>
9645 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9646 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9647 (vect_use_loop_mask_for_alignment_p): New function.
9648 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9649 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9650 niters_skip argument. Make sure that the first niters_skip elements
9651 of the first iteration are inactive.
9652 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9653 Update call to vect_set_loop_masks_directly.
9654 (get_misalign_in_elems): New function, split out from...
9655 (vect_gen_prolog_loop_niters): ...here.
9656 (vect_update_init_of_dr): Take a code argument that specifies whether
9657 the adjustment should be added or subtracted.
9658 (vect_update_init_of_drs): Likewise.
9659 (vect_prepare_for_masked_peels): New function.
9660 (vect_do_peeling): Skip prologue peeling if we're using a mask
9661 instead. Update call to vect_update_inits_of_drs.
9662 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9664 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9665 alignment. Do not include the number of peeled iterations in
9666 the minimum threshold in that case.
9667 (vectorizable_induction): Adjust the start value down by
9668 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9669 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9670 Take the number of skipped iterations into account when calculating
9672 * tree-vect-stmts.c (vect_gen_while_not): New function.
9674 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9675 Alan Hayward <alan.hayward@arm.com>
9676 David Sherwood <david.sherwood@arm.com>
9678 * doc/sourcebuild.texi (vect_fully_masked): Document.
9679 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9681 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9683 (vect_analyze_loop_2): ...here. Don't check the vectorization
9684 factor against the number of loop iterations if the loop is
9687 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9688 Alan Hayward <alan.hayward@arm.com>
9689 David Sherwood <david.sherwood@arm.com>
9691 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9692 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9693 (dump_groups): Update accordingly.
9694 (iv_use::mem_type): New member variable.
9695 (address_p): New function.
9696 (record_use): Add a mem_type argument and initialize the new
9698 (record_group_use): Add a mem_type argument. Use address_p.
9699 Remove obsolete null checks of base_object. Update call to record_use.
9700 (find_interesting_uses_op): Update call to record_group_use.
9701 (find_interesting_uses_cond): Likewise.
9702 (find_interesting_uses_address): Likewise.
9703 (get_mem_type_for_internal_fn): New function.
9704 (find_address_like_use): Likewise.
9705 (find_interesting_uses_stmt): Try find_address_like_use before
9706 calling find_interesting_uses_op.
9707 (addr_offset_valid_p): Use the iv mem_type field as the type
9708 of the addressed memory.
9709 (add_autoinc_candidates): Likewise.
9710 (get_address_cost): Likewise.
9711 (split_small_address_groups_p): Use address_p.
9712 (split_address_groups): Likewise.
9713 (add_iv_candidate_for_use): Likewise.
9714 (autoinc_possible_for_pair): Likewise.
9715 (rewrite_groups): Likewise.
9716 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9717 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9718 (get_alias_ptr_type_for_ptr_address): New function.
9719 (rewrite_use_address): Rewrite address uses in calls that were
9720 identified by find_address_like_use.
9722 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9723 Alan Hayward <alan.hayward@arm.com>
9724 David Sherwood <david.sherwood@arm.com>
9726 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9728 * gimple-expr.h (is_gimple_addressable: Likewise.
9729 * gimple-expr.c (is_gimple_address): Likewise.
9730 * internal-fn.c (expand_call_mem_ref): New function.
9731 (expand_mask_load_optab_fn): Use it.
9732 (expand_mask_store_optab_fn): Likewise.
9734 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9735 Alan Hayward <alan.hayward@arm.com>
9736 David Sherwood <david.sherwood@arm.com>
9738 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9739 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9740 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9741 (cond_umax@var{mode}): Document.
9742 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9743 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9744 (cond_umin_optab, cond_umax_optab): New optabs.
9745 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9746 (COND_IOR, COND_XOR): New internal functions.
9747 * internal-fn.h (get_conditional_internal_fn): Declare.
9748 * internal-fn.c (cond_binary_direct): New macro.
9749 (expand_cond_binary_optab_fn): Likewise.
9750 (direct_cond_binary_optab_supported_p): Likewise.
9751 (get_conditional_internal_fn): New function.
9752 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9753 Cope with reduction statements that are vectorized as calls rather
9755 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9756 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9757 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9758 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9759 (UNSPEC_COND_EOR): New unspecs.
9760 (optab): Add mappings for them.
9761 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9762 (sve_int_op, sve_fp_op): New int attributes.
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 * optabs.def (while_ult_optab): New optab.
9769 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9770 * internal-fn.def (WHILE_ULT): New internal function.
9771 * internal-fn.h (direct_internal_fn_supported_p): New override
9772 that takes two types as argument.
9773 * internal-fn.c (while_direct): New macro.
9774 (expand_while_optab_fn): New function.
9775 (convert_optab_supported_p): Likewise.
9776 (direct_while_optab_supported_p): New macro.
9777 * wide-int.h (wi::udiv_ceil): New function.
9778 * tree-vectorizer.h (rgroup_masks): New structure.
9779 (vec_loop_masks): New typedef.
9780 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9782 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9783 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9784 (vect_max_vf): New function.
9785 (slpeel_make_loop_iterate_ntimes): Delete.
9786 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9787 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9788 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9789 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9790 internal-fn.h, stor-layout.h and optabs-query.h.
9791 (vect_set_loop_mask): New function.
9792 (add_preheader_seq): Likewise.
9793 (add_header_seq): Likewise.
9794 (interleave_supported_p): Likewise.
9795 (vect_maybe_permute_loop_masks): Likewise.
9796 (vect_set_loop_masks_directly): Likewise.
9797 (vect_set_loop_condition_masked): Likewise.
9798 (vect_set_loop_condition_unmasked): New function, split out from
9799 slpeel_make_loop_iterate_ntimes.
9800 (slpeel_make_loop_iterate_ntimes): Rename to..
9801 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9802 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9803 (vect_do_peeling): Update call accordingly.
9804 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9806 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9807 mask_compare_type, can_fully_mask_p and fully_masked_p.
9808 (release_vec_loop_masks): New function.
9809 (_loop_vec_info): Use it to free the loop masks.
9810 (can_produce_all_loop_masks_p): New function.
9811 (vect_get_max_nscalars_per_iter): Likewise.
9812 (vect_verify_full_masking): Likewise.
9813 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9814 retries, and free the mask rgroups before retrying. Check loop-wide
9815 reasons for disallowing fully-masked loops. Make the final decision
9816 about whether use a fully-masked loop or not.
9817 (vect_estimate_min_profitable_iters): Do not assume that peeling
9818 for the number of iterations will be needed for fully-masked loops.
9819 (vectorizable_reduction): Disable fully-masked loops.
9820 (vectorizable_live_operation): Likewise.
9821 (vect_halve_mask_nunits): New function.
9822 (vect_double_mask_nunits): Likewise.
9823 (vect_record_loop_mask): Likewise.
9824 (vect_get_loop_mask): Likewise.
9825 (vect_transform_loop): Handle the case in which the final loop
9826 iteration might handle a partial vector. Call vect_set_loop_condition
9827 instead of slpeel_make_loop_iterate_ntimes.
9828 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9829 (check_load_store_masking): New function.
9830 (prepare_load_store_mask): Likewise.
9831 (vectorizable_store): Handle fully-masked loops.
9832 (vectorizable_load): Likewise.
9833 (supportable_widening_operation): Use vect_halve_mask_nunits for
9835 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9836 (vect_gen_while): New function.
9837 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9838 (aarch64_uqdec<mode>): New insn.
9840 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9841 Alan Hayward <alan.hayward@arm.com>
9842 David Sherwood <david.sherwood@arm.com>
9844 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9845 (reduc_xor_scal_optab): New optabs.
9846 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9847 (reduc_xor_scal_@var{m}): Document.
9848 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9849 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9851 * fold-const-call.c (fold_const_call): Handle them.
9852 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9853 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9854 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9855 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9856 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9857 (UNSPEC_XORV): New unspecs.
9858 (optab): Add entries for them.
9859 (BITWISEV): New int iterator.
9860 (bit_reduc_op): New int attributes.
9862 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9863 Alan Hayward <alan.hayward@arm.com>
9864 David Sherwood <david.sherwood@arm.com>
9866 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9867 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9868 * optabs.def (vec_shl_insert_optab): New optab.
9869 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9870 (duplicate_and_interleave): Likewise.
9871 * tree-vect-loop.c: Include internal-fn.h.
9872 (neutral_op_for_slp_reduction): New function, split out from
9873 get_initial_defs_for_reduction.
9874 (get_initial_def_for_reduction): Handle option 2 for variable-length
9875 vectors by loading the neutral value into a vector and then shifting
9876 the initial value into element 0.
9877 (get_initial_defs_for_reduction): Replace the code argument with
9878 the neutral value calculated by neutral_op_for_slp_reduction.
9879 Use gimple_build_vector for constant-length vectors.
9880 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9881 but the first group_size elements have a neutral value.
9882 Use duplicate_and_interleave otherwise.
9883 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9884 Update call to get_initial_defs_for_reduction. Handle SLP
9885 reductions for variable-length vectors by creating one vector
9886 result for each scalar result, with the elements associated
9887 with other scalar results stubbed out with the neutral value.
9888 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9889 Require IFN_VEC_SHL_INSERT for double reductions on
9890 variable-length vectors, or SLP reductions that have
9891 a neutral value. Require can_duplicate_and_interleave_p
9892 support for variable-length unchained SLP reductions if there
9893 is no neutral value, such as for MIN/MAX reductions. Also require
9894 the number of vector elements to be a multiple of the number of
9895 SLP statements when doing variable-length unchained SLP reductions.
9896 Update call to vect_create_epilog_for_reduction.
9897 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9898 and remove initial values.
9899 (duplicate_and_interleave): Make public.
9900 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9901 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9903 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9904 Alan Hayward <alan.hayward@arm.com>
9905 David Sherwood <david.sherwood@arm.com>
9907 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9908 (can_duplicate_and_interleave_p): New function.
9909 (vect_get_and_check_slp_defs): Take the vector of statements
9910 rather than just the current one. Remove excess parentheses.
9911 Restriction rejectinon of vect_constant_def and vect_external_def
9912 for variable-length vectors to boolean types, or types for which
9913 can_duplicate_and_interleave_p is false.
9914 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9915 (duplicate_and_interleave): New function.
9916 (vect_get_constant_vectors): Use gimple_build_vector for
9917 constant-length vectors and suitable variable-length constant
9918 vectors. Use duplicate_and_interleave for other variable-length
9919 vectors. Don't defer the update when inserting new statements.
9921 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9922 Alan Hayward <alan.hayward@arm.com>
9923 David Sherwood <david.sherwood@arm.com>
9925 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9926 min_profitable_iters doesn't go negative.
9928 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9929 Alan Hayward <alan.hayward@arm.com>
9930 David Sherwood <david.sherwood@arm.com>
9932 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9933 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9934 * optabs.def (vec_mask_load_lanes_optab): New optab.
9935 (vec_mask_store_lanes_optab): Likewise.
9936 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9937 (MASK_STORE_LANES): Likewise.
9938 * internal-fn.c (mask_load_lanes_direct): New macro.
9939 (mask_store_lanes_direct): Likewise.
9940 (expand_mask_load_optab_fn): Handle masked operations.
9941 (expand_mask_load_lanes_optab_fn): New macro.
9942 (expand_mask_store_optab_fn): Handle masked operations.
9943 (expand_mask_store_lanes_optab_fn): New macro.
9944 (direct_mask_load_lanes_optab_supported_p): Likewise.
9945 (direct_mask_store_lanes_optab_supported_p): Likewise.
9946 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9948 (vect_load_lanes_supported): Likewise.
9949 * tree-vect-data-refs.c (strip_conversion): New function.
9950 (can_group_stmts_p): Likewise.
9951 (vect_analyze_data_ref_accesses): Use it instead of checking
9952 for a pair of assignments.
9953 (vect_store_lanes_supported): Take a masked_p parameter.
9954 (vect_load_lanes_supported): Likewise.
9955 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9956 vect_store_lanes_supported and vect_load_lanes_supported.
9957 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9958 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9959 parameter. Don't allow gaps for masked accesses.
9960 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9961 and vect_load_lanes_supported.
9962 (get_load_store_type): Take a masked_p parameter and update
9963 call to get_group_load_store_type.
9964 (vectorizable_store): Update call to get_load_store_type.
9965 Handle IFN_MASK_STORE_LANES.
9966 (vectorizable_load): Update call to get_load_store_type.
9967 Handle IFN_MASK_LOAD_LANES.
9969 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9970 Alan Hayward <alan.hayward@arm.com>
9971 David Sherwood <david.sherwood@arm.com>
9973 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9975 * config/aarch64/aarch64-protos.h
9976 (aarch64_sve_struct_memory_operand_p): Declare.
9977 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9978 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9979 (VPRED, vpred): Handle SVE structure modes.
9980 * config/aarch64/constraints.md (Utx): New constraint.
9981 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9982 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9983 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9984 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9985 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9986 structure modes. Split into pieces after RA.
9987 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9988 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9990 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9991 SVE structure modes.
9992 (aarch64_classify_address): Likewise.
9993 (sizetochar): Move earlier in file.
9994 (aarch64_print_operand): Handle SVE register lists.
9995 (aarch64_array_mode): New function.
9996 (aarch64_sve_struct_memory_operand_p): Likewise.
9997 (TARGET_ARRAY_MODE): Redefine.
9999 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10000 Alan Hayward <alan.hayward@arm.com>
10001 David Sherwood <david.sherwood@arm.com>
10003 * target.def (array_mode): New target hook.
10004 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10005 * doc/tm.texi: Regenerate.
10006 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10007 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10008 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10009 targetm.array_mode.
10010 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10013 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10014 Alan Hayward <alan.hayward@arm.com>
10015 David Sherwood <david.sherwood@arm.com>
10017 * fold-const.c (fold_binary_loc): Check the argument types
10018 rather than the result type when testing for a vector operation.
10020 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10022 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10023 * doc/tm.texi: Regenerate.
10025 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10026 Alan Hayward <alan.hayward@arm.com>
10027 David Sherwood <david.sherwood@arm.com>
10029 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10030 (sve): Document new AArch64 extension.
10031 * doc/md.texi (w): Extend the description of the AArch64
10032 constraint to include SVE vectors.
10033 (Upl, Upa): Document new AArch64 predicate constraints.
10034 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10036 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10037 (msve-vector-bits=): New option.
10038 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10039 SVE when these are disabled.
10040 (sve): New extension.
10041 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10042 modes. Adjust their number of units based on aarch64_sve_vg.
10043 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10044 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10045 aarch64_addr_query_type.
10046 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10047 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10048 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10049 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10050 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10051 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10052 (aarch64_simd_imm_zero_p): Delete.
10053 (aarch64_check_zero_based_sve_index_immediate): Declare.
10054 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10055 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10056 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10057 (aarch64_sve_float_mul_immediate_p): Likewise.
10058 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10059 rather than an rtx.
10060 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10061 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10062 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10063 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10064 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10065 (aarch64_regmode_natural_size): Likewise.
10066 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10067 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10069 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10070 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10071 for VG and the SVE predicate registers.
10072 (V_ALIASES): Add a "z"-prefixed alias.
10073 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10074 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10075 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10076 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10077 (REG_CLASS_NAMES): Add entries for them.
10078 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10079 and the predicate registers.
10080 (aarch64_sve_vg): Declare.
10081 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10082 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10083 (REGMODE_NATURAL_SIZE): Define.
10084 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10086 * config/aarch64/aarch64.c: Include cfgrtl.h.
10087 (simd_immediate_info): Add a constructor for series vectors,
10088 and an associated step field.
10089 (aarch64_sve_vg): New variable.
10090 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10091 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10092 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10093 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10094 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10095 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10096 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10097 (aarch64_get_mask_mode): New functions.
10098 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10099 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10100 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10101 predicate modes and predicate registers. Explicitly restrict
10102 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10103 to store a vector mode if it is recognized by
10104 aarch64_classify_vector_mode.
10105 (aarch64_regmode_natural_size): New function.
10106 (aarch64_hard_regno_caller_save_mode): Return the original mode
10108 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10109 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10110 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10111 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10113 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10114 does not overlap dest if the function is frame-related. Handle
10116 (aarch64_split_add_offset): New function.
10117 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10118 them aarch64_add_offset.
10119 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10120 and update call to aarch64_sub_sp.
10121 (aarch64_add_cfa_expression): New function.
10122 (aarch64_expand_prologue): Pass extra temporary registers to the
10123 functions above. Handle the case in which we need to emit new
10124 DW_CFA_expressions for registers that were originally saved
10125 relative to the stack pointer, but now have to be expressed
10126 relative to the frame pointer.
10127 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10129 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10130 IP0 and IP1 values for SVE frames.
10131 (aarch64_expand_vec_series): New function.
10132 (aarch64_expand_sve_widened_duplicate): Likewise.
10133 (aarch64_expand_sve_const_vector): Likewise.
10134 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10135 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10136 into the register, rather than emitting a SET directly.
10137 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10138 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10139 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10140 (offset_9bit_signed_scaled_p): New functions.
10141 (aarch64_replicate_bitmask_imm): New function.
10142 (aarch64_bitmask_imm): Use it.
10143 (aarch64_cannot_force_const_mem): Reject expressions involving
10144 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10145 (aarch64_classify_index): Handle SVE indices, by requiring
10146 a plain register index with a scale that matches the element size.
10147 (aarch64_classify_address): Handle SVE addresses. Assert that
10148 the mode of the address is VOIDmode or an integer mode.
10149 Update call to aarch64_classify_symbol.
10150 (aarch64_classify_symbolic_expression): Update call to
10151 aarch64_classify_symbol.
10152 (aarch64_const_vec_all_in_range_p): New function.
10153 (aarch64_print_vector_float_operand): Likewise.
10154 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10155 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10156 and the FP immediates 1.0 and 0.5.
10157 (aarch64_print_address_internal): Handle SVE addresses.
10158 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10159 (aarch64_regno_regclass): Handle predicate registers.
10160 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10162 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10163 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10164 (aarch64_convert_sve_vector_bits): New function.
10165 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10166 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10167 rather than an rtx.
10168 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10169 Handle SVE vector and predicate modes. Accept VL-based constants
10170 that need only one temporary register, and VL offsets that require
10171 no temporary registers.
10172 (aarch64_conditional_register_usage): Mark the predicate registers
10173 as fixed if SVE isn't available.
10174 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10175 Return true for SVE vector and predicate modes.
10176 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10177 rather than an unsigned int. Handle SVE modes.
10178 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10180 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10182 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10183 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10184 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10185 (aarch64_sve_float_mul_immediate_p): New functions.
10186 (aarch64_sve_valid_immediate): New function.
10187 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10188 Explicitly reject structure modes. Check for INDEX constants.
10189 Handle PTRUE and PFALSE constants.
10190 (aarch64_check_zero_based_sve_index_immediate): New function.
10191 (aarch64_simd_imm_zero_p): Delete.
10192 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10193 vector modes. Accept constants in the range of CNT[BHWD].
10194 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10195 ask for an Advanced SIMD mode.
10196 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10197 (aarch64_simd_vector_alignment): Handle SVE predicates.
10198 (aarch64_vectorize_preferred_vector_alignment): New function.
10199 (aarch64_simd_vector_alignment_reachable): Use it instead of
10201 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10202 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10204 (MAX_VECT_LEN): Delete.
10205 (expand_vec_perm_d): Add a vec_flags field.
10206 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10207 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10208 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10210 (aarch64_evpc_rev): Rename to...
10211 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10212 (aarch64_evpc_rev_global): New function.
10213 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10214 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10216 (aarch64_evpc_sve_tbl): New function.
10217 (aarch64_expand_vec_perm_const_1): Update after rename of
10218 aarch64_evpc_rev. Handle SVE permutes too, trying
10219 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10220 than aarch64_evpc_tbl.
10221 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10222 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10223 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10224 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10225 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10226 (aarch64_expand_sve_vcond): New functions.
10227 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10228 of aarch64_vector_mode_p.
10229 (aarch64_dwarf_poly_indeterminate_value): New function.
10230 (aarch64_compute_pressure_classes): Likewise.
10231 (aarch64_can_change_mode_class): Likewise.
10232 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10233 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10234 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10235 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10236 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10237 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10238 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10239 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10241 (Dn, Dl, Dr): Accept const as well as const_vector.
10242 (Dz): Likewise. Compare against CONST0_RTX.
10243 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10244 of "vector" where appropriate.
10245 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10246 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10247 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10248 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10249 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10250 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10251 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10252 (v_int_equiv): Extend to SVE modes.
10253 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10255 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10256 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10257 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10258 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10259 (SVE_COND_FP_CMP): New int iterators.
10260 (perm_hilo): Handle the new unpack unspecs.
10261 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10263 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10264 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10265 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10266 (aarch64_equality_operator, aarch64_constant_vector_operand)
10267 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10268 (aarch64_sve_nonimmediate_operand): Likewise.
10269 (aarch64_sve_general_operand): Likewise.
10270 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10271 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10272 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10273 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10274 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10275 (aarch64_sve_float_arith_immediate): Likewise.
10276 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10277 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10278 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10279 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10280 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10281 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10282 (aarch64_sve_float_arith_operand): Likewise.
10283 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10284 (aarch64_sve_float_mul_operand): Likewise.
10285 (aarch64_sve_vec_perm_operand): Likewise.
10286 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10287 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10288 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10289 as well as const_vector.
10290 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10291 in file. Use CONST0_RTX and CONSTM1_RTX.
10292 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10293 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10294 Use aarch64_simd_imm_zero.
10295 * config/aarch64/aarch64-sve.md: New file.
10296 * config/aarch64/aarch64.md: Include it.
10297 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10298 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10299 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10300 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10301 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10302 (sve): New attribute.
10303 (enabled): Disable instructions with the sve attribute unless
10305 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10306 aarch64_expand_mov_immediate.
10307 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10308 CNT[BHSD] immediates.
10309 (movti): Split CONST_POLY_INT moves into two halves.
10310 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10311 Split additions that need a temporary here if the destination
10312 is the stack pointer.
10313 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10314 (*add<mode>3_poly_1): New instruction.
10315 (set_clobber_cc): New expander.
10317 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10319 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10320 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10321 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10322 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10323 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10324 Change innermode from fixed_mode_size to machine_mode.
10325 (simplify_subreg): Update call accordingly. Handle a constant-sized
10326 subreg of a variable-length CONST_VECTOR.
10328 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10329 Alan Hayward <alan.hayward@arm.com>
10330 David Sherwood <david.sherwood@arm.com>
10332 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10333 (add_offset_to_base): New function, split out from...
10334 (create_mem_ref): ...here. When handling a scale other than 1,
10335 check first whether the address is valid without the offset.
10336 Add it into the base if so, leaving the index and scale as-is.
10338 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10341 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10342 fold_for_warn before checking if arg2 is INTEGER_CST.
10344 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10346 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10347 (store_multiple_operation): Delete.
10348 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10349 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10350 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10351 guarded by TARGET_STRING.
10352 (rs6000_output_load_multiple): Delete.
10353 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10354 OPTION_MASK_STRING / TARGET_STRING handling.
10355 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10356 (const rs6000_opt_masks) <"string">: Change mask to 0.
10357 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10358 (MASK_STRING): Delete.
10359 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10361 (load_multiple): Delete.
10368 (store_multiple): Delete.
10375 (movmemsi_8reg): Delete.
10376 (corresponding unnamed define_insn): Delete.
10377 (movmemsi_6reg): Delete.
10378 (corresponding unnamed define_insn): Delete.
10379 (movmemsi_4reg): Delete.
10380 (corresponding unnamed define_insn): Delete.
10381 (movmemsi_2reg): Delete.
10382 (corresponding unnamed define_insn): Delete.
10383 (movmemsi_1reg): Delete.
10384 (corresponding unnamed define_insn): Delete.
10385 * config/rs6000/rs6000.opt (mno-string): New.
10386 (mstring): Replace by deprecation warning stub.
10387 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10389 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10391 * regrename.c (regrename_do_replace): If replacing the same
10392 reg multiple times, try to reuse last created gen_raw_REG.
10395 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10396 main to workaround a bug in GDB.
10398 2018-01-12 Tom de Vries <tom@codesourcery.com>
10401 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10403 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10405 PR rtl-optimization/80481
10406 * ira-color.c (get_cap_member): New function.
10407 (allocnos_conflict_by_live_ranges_p): Use it.
10408 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10409 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10411 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10414 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10415 (*saddl_se_1): Ditto.
10416 (*ssubsi_1): Ditto.
10417 (*ssubl_se_1): Ditto.
10419 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10421 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10422 rather than wi::to_widest for DR_INITs.
10423 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10424 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10425 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10427 (vect_analyze_group_access_1): Note that here.
10429 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10431 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10432 polynomial type sizes.
10434 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10436 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10437 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10438 (gimple_add_tmp_var): Likewise.
10440 2018-01-12 Martin Liska <mliska@suse.cz>
10442 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10443 (gimple_alloc_sizes): Likewise.
10444 (dump_gimple_statistics): Use PRIu64 in printf format.
10445 * gimple.h: Change uint64_t to int.
10447 2018-01-12 Martin Liska <mliska@suse.cz>
10449 * tree-core.h: Use uint64_t instead of int.
10450 * tree.c (tree_node_counts): Likewise.
10451 (tree_node_sizes): Likewise.
10452 (dump_tree_statistics): Use PRIu64 in printf format.
10454 2018-01-12 Martin Liska <mliska@suse.cz>
10456 * Makefile.in: As qsort_chk is implemented in vec.c, add
10457 vec.o to linkage of gencfn-macros.
10458 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10459 passing the info to record_node_allocation_statistics.
10460 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10462 * ggc-common.c (struct ggc_usage): Add operator== and use
10463 it in operator< and compare function.
10464 * mem-stats.h (struct mem_usage): Likewise.
10465 * vec.c (struct vec_usage): Remove operator< and compare
10466 function. Can be simply inherited.
10468 2018-01-12 Martin Jambor <mjambor@suse.cz>
10471 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10472 * tree-ssa-math-opts.c: Include domwalk.h.
10473 (convert_mult_to_fma_1): New function.
10474 (fma_transformation_info): New type.
10475 (fma_deferring_state): Likewise.
10476 (cancel_fma_deferring): New function.
10477 (result_of_phi): Likewise.
10478 (last_fma_candidate_feeds_initial_phi): Likewise.
10479 (convert_mult_to_fma): Added deferring logic, split actual
10480 transformation to convert_mult_to_fma_1.
10481 (math_opts_dom_walker): New type.
10482 (math_opts_dom_walker::after_dom_children): New method, body moved
10483 here from pass_optimize_widening_mul::execute, added deferring logic
10485 (pass_optimize_widening_mul::execute): Moved most of code to
10486 math_opts_dom_walker::after_dom_children.
10487 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10488 * config/i386/i386.c (ix86_option_override_internal): Added
10489 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10491 2018-01-12 Richard Biener <rguenther@suse.de>
10494 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10495 inline instance vars.
10497 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10500 * config/rx/rx.c (rx_is_restricted_memory_address):
10501 Handle SUBREG case.
10503 2018-01-12 Richard Biener <rguenther@suse.de>
10505 PR tree-optimization/80846
10506 * target.def (split_reduction): New target hook.
10507 * targhooks.c (default_split_reduction): New function.
10508 * targhooks.h (default_split_reduction): Declare.
10509 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10510 target requests first reduce vectors by combining low and high
10512 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10513 (get_vectype_for_scalar_type_and_size): Export.
10514 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10515 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10516 * doc/tm.texi: Regenerate.
10517 * config/i386/i386.c (ix86_split_reduction): Implement
10518 TARGET_VECTORIZE_SPLIT_REDUCTION.
10520 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10523 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10524 in PIC mode except for TARGET_VXWORKS_RTP.
10525 * config/sparc/sparc.c: Include cfgrtl.h.
10526 (TARGET_INIT_PIC_REG): Define.
10527 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10528 (sparc_pic_register_p): New predicate.
10529 (sparc_legitimate_address_p): Use it.
10530 (sparc_legitimize_pic_address): Likewise.
10531 (sparc_delegitimize_address): Likewise.
10532 (sparc_mode_dependent_address_p): Likewise.
10533 (gen_load_pcrel_sym): Remove 4th parameter.
10534 (load_got_register): Adjust call to above. Remove obsolete stuff.
10535 (sparc_expand_prologue): Do not call load_got_register here.
10536 (sparc_flat_expand_prologue): Likewise.
10537 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10538 (sparc_use_pseudo_pic_reg): New function.
10539 (sparc_init_pic_reg): Likewise.
10540 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10541 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10543 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10545 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10546 Add item for branch_cost.
10548 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10550 PR rtl-optimization/83565
10551 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10552 not extend the result to a larger mode for rotate operations.
10553 (num_sign_bit_copies1): Likewise.
10555 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10558 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10560 Use values-Xc.o for -pedantic.
10561 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10563 2018-01-12 Martin Liska <mliska@suse.cz>
10566 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10568 (possible_polymorphic_call_targets): Use it.
10569 (ipa_devirt): Likewise.
10571 2018-01-12 Martin Liska <mliska@suse.cz>
10573 * profile-count.h (enum profile_quality): Use 0 as invalid
10574 enum value of profile_quality.
10576 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10578 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10579 -mext-string options.
10581 2018-01-12 Richard Biener <rguenther@suse.de>
10583 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10584 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10585 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10587 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10589 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10591 * configure.ac (--with-long-double-format): Add support for the
10592 configuration option to change the default long double format on
10594 * config.gcc (powerpc*-linux*-*): Likewise.
10595 * configure: Regenerate.
10596 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10597 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10598 used without modification.
10600 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10602 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10603 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10604 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10605 MISC_BUILTIN_SPEC_BARRIER.
10606 (rs6000_init_builtins): Likewise.
10607 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10609 (speculation_barrier): New define_insn.
10610 * doc/extend.texi: Document __builtin_speculation_barrier.
10612 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10615 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10616 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10617 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10619 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10620 integral modes instead of "ss" and "sd".
10621 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10622 vectors with 32-bit and 64-bit elements.
10623 (vecdupssescalarmodesuffix): New mode attribute.
10624 (vec_dup<mode>): Use it.
10626 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10629 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10630 frame if argument is passed on stack.
10632 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10635 * ree.c (combine_reaching_defs): Optimize also
10636 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10637 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10639 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10641 PR middle-end/83189
10642 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10644 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10646 PR middle-end/83718
10647 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10648 after they are computed.
10650 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10652 PR tree-optimization/83695
10653 * gimple-loop-linterchange.cc
10654 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10655 reset cached scev information after interchange.
10656 (pass_linterchange::execute): Remove call to scev_reset_htab.
10658 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10660 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10661 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10662 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10663 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10664 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10665 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10666 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10667 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10668 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10669 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10670 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10671 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10672 (V_lane_reg): Likewise.
10673 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10675 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10676 (vfmal_lane_low<mode>_intrinsic,
10677 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10678 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10679 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10680 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10681 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10682 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10684 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10686 * config/arm/arm-cpus.in (fp16fml): New feature.
10687 (ALL_SIMD): Add fp16fml.
10688 (armv8.2-a): Add fp16fml as an option.
10689 (armv8.3-a): Likewise.
10690 (armv8.4-a): Add fp16fml as part of fp16.
10691 * config/arm/arm.h (TARGET_FP16FML): Define.
10692 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10694 * config/arm/arm-modes.def (V2HF): Define.
10695 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10696 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10697 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10698 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10699 vfmsl_low, vfmsl_high): New set of builtins.
10700 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10701 (vfml_op): New code attribute.
10702 (VFMLHALVES): New int iterator.
10703 (VFML, VFMLSEL): New mode attributes.
10704 (V_reg): Define mapping for V2HF.
10705 (V_hi, V_lo): New mode attributes.
10706 (VF_constraint): Likewise.
10707 (vfml_half, vfml_half_selector): New int attributes.
10708 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10710 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10711 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10713 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10714 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10715 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10716 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10718 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10719 Document new effective target and option set.
10721 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10723 * config/arm/arm-cpus.in (armv8_4): New feature.
10724 (ARMv8_4a): New fgroup.
10725 (armv8.4-a): New arch.
10726 * config/arm/arm-tables.opt: Regenerate.
10727 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10728 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10729 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10730 Add matching rules for -march=armv8.4-a and extensions.
10731 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10733 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10736 * config/rx/rx.md (BW): New mode attribute.
10737 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10739 2018-01-11 Richard Biener <rguenther@suse.de>
10741 PR tree-optimization/83435
10742 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10743 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10744 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10746 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10747 Alan Hayward <alan.hayward@arm.com>
10748 David Sherwood <david.sherwood@arm.com>
10750 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10752 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10753 (aarch64_print_address_internal): Use it to check for a zero offset.
10755 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10756 Alan Hayward <alan.hayward@arm.com>
10757 David Sherwood <david.sherwood@arm.com>
10759 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10760 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10761 Return a poly_int64 rather than a HOST_WIDE_INT.
10762 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10763 rather than a HOST_WIDE_INT.
10764 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10765 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10766 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10767 final_offset from HOST_WIDE_INT to poly_int64.
10768 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10769 to_constant when getting the number of units in an Advanced SIMD
10771 (aarch64_builtin_vectorized_function): Check for a constant number
10773 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10775 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10776 attribute instead of GET_MODE_NUNITS.
10777 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10778 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10779 GET_MODE_SIZE for fixed-size registers.
10780 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10781 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10782 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10783 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10784 (aarch64_print_operand, aarch64_print_address_internal)
10785 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10786 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10787 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10788 Handle polynomial GET_MODE_SIZE.
10789 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10790 wider than SImode without modification.
10791 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10792 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10793 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10794 passing and returning SVE modes.
10795 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10796 rather than GEN_INT.
10797 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10798 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10799 (aarch64_allocate_and_probe_stack_space): Likewise.
10800 (aarch64_layout_frame): Cope with polynomial offsets.
10801 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10802 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10803 polynomial offsets.
10804 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10805 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10806 poly_int64 rather than a HOST_WIDE_INT.
10807 (aarch64_get_separate_components, aarch64_process_components)
10808 (aarch64_expand_prologue, aarch64_expand_epilogue)
10809 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10810 (aarch64_anchor_offset): New function, split out from...
10811 (aarch64_legitimize_address): ...here.
10812 (aarch64_builtin_vectorization_cost): Handle polynomial
10813 TYPE_VECTOR_SUBPARTS.
10814 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10816 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10817 number of elements from the PARALLEL rather than the mode.
10818 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10819 rather than GET_MODE_BITSIZE.
10820 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10821 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10822 (aarch64_expand_vec_perm_const_1): Handle polynomial
10823 d->perm.length () and d->perm elements.
10824 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10825 Apply to_constant to d->perm elements.
10826 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10827 polynomial CONST_VECTOR_NUNITS.
10828 (aarch64_move_pointer): Take amount as a poly_int64 rather
10830 (aarch64_progress_pointer): Avoid temporary variable.
10831 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10832 the mode attribute instead of GET_MODE.
10834 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10835 Alan Hayward <alan.hayward@arm.com>
10836 David Sherwood <david.sherwood@arm.com>
10838 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10839 x exists before using it.
10840 (aarch64_add_constant_internal): Rename to...
10841 (aarch64_add_offset_1): ...this. Replace regnum with separate
10842 src and dest rtxes. Handle the case in which they're different,
10843 including when the offset is zero. Replace scratchreg with an rtx.
10844 Use 2 additions if there is no spare register into which we can
10845 move a 16-bit constant.
10846 (aarch64_add_constant): Delete.
10847 (aarch64_add_offset): Replace reg with separate src and dest
10848 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10849 Use aarch64_add_offset_1.
10850 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10851 an rtx rather than an int. Take the delta as a poly_int64
10852 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10853 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10854 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10855 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10856 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10857 and aarch64_add_sp.
10858 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10859 aarch64_add_constant.
10861 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10863 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10864 Use scalar_float_mode.
10866 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10868 * config/aarch64/aarch64-simd.md
10869 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10870 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10871 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10872 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10873 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10874 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10875 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10876 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10877 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10878 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10880 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10883 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10884 targ_options->x_arm_arch_string is non NULL.
10886 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10888 * config/aarch64/aarch64.h
10889 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10891 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10894 * expmed.c (emit_store_flag_force): Swap if const op0
10895 and change VOIDmode to mode of op0.
10897 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10899 PR rtl-optimization/83761
10900 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10901 than bytes to mode_for_size.
10903 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10905 PR middle-end/83189
10906 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10907 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10910 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10912 PR middle-end/83575
10913 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10914 when in layout mode.
10915 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10916 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10919 2018-01-10 Michael Collison <michael.collison@arm.com>
10921 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10922 * config/aarch64/aarch64-option-extension.def: Add
10923 AARCH64_OPT_EXTENSION of 'fp16fml'.
10924 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10925 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10926 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10927 * config/aarch64/constraints.md (Ui7): New constraint.
10928 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10929 (VFMLA_SEL_W): Ditto.
10932 (VFMLA16_LOW): New int iterator.
10933 (VFMLA16_HIGH): Ditto.
10934 (UNSPEC_FMLAL): New unspec.
10935 (UNSPEC_FMLSL): Ditto.
10936 (UNSPEC_FMLAL2): Ditto.
10937 (UNSPEC_FMLSL2): Ditto.
10938 (f16mac): New code attribute.
10939 * config/aarch64/aarch64-simd-builtins.def
10940 (aarch64_fmlal_lowv2sf): Ditto.
10941 (aarch64_fmlsl_lowv2sf): Ditto.
10942 (aarch64_fmlalq_lowv4sf): Ditto.
10943 (aarch64_fmlslq_lowv4sf): Ditto.
10944 (aarch64_fmlal_highv2sf): Ditto.
10945 (aarch64_fmlsl_highv2sf): Ditto.
10946 (aarch64_fmlalq_highv4sf): Ditto.
10947 (aarch64_fmlslq_highv4sf): Ditto.
10948 (aarch64_fmlal_lane_lowv2sf): Ditto.
10949 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10950 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10951 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10952 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10953 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10954 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10955 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10956 (aarch64_fmlal_lane_highv2sf): Ditto.
10957 (aarch64_fmlsl_lane_highv2sf): Ditto.
10958 (aarch64_fmlal_laneq_highv2sf): Ditto.
10959 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10960 (aarch64_fmlalq_lane_highv4sf): Ditto.
10961 (aarch64_fmlsl_lane_highv4sf): Ditto.
10962 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10963 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10964 * config/aarch64/aarch64-simd.md:
10965 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10966 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10967 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10968 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10969 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10970 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10971 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10972 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10973 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10974 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10975 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10976 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10977 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10978 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10979 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10980 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10981 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10982 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10983 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10984 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10985 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10986 (vfmlsl_low_u32): Ditto.
10987 (vfmlalq_low_u32): Ditto.
10988 (vfmlslq_low_u32): Ditto.
10989 (vfmlal_high_u32): Ditto.
10990 (vfmlsl_high_u32): Ditto.
10991 (vfmlalq_high_u32): Ditto.
10992 (vfmlslq_high_u32): Ditto.
10993 (vfmlal_lane_low_u32): Ditto.
10994 (vfmlsl_lane_low_u32): Ditto.
10995 (vfmlal_laneq_low_u32): Ditto.
10996 (vfmlsl_laneq_low_u32): Ditto.
10997 (vfmlalq_lane_low_u32): Ditto.
10998 (vfmlslq_lane_low_u32): Ditto.
10999 (vfmlalq_laneq_low_u32): Ditto.
11000 (vfmlslq_laneq_low_u32): Ditto.
11001 (vfmlal_lane_high_u32): Ditto.
11002 (vfmlsl_lane_high_u32): Ditto.
11003 (vfmlal_laneq_high_u32): Ditto.
11004 (vfmlsl_laneq_high_u32): Ditto.
11005 (vfmlalq_lane_high_u32): Ditto.
11006 (vfmlslq_lane_high_u32): Ditto.
11007 (vfmlalq_laneq_high_u32): Ditto.
11008 (vfmlslq_laneq_high_u32): Ditto.
11009 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11010 (AARCH64_FL_FOR_ARCH8_4): New.
11011 (AARCH64_ISA_F16FML): New ISA flag.
11012 (TARGET_F16FML): New feature flag for fp16fml.
11013 (doc/invoke.texi): Document new fp16fml option.
11015 2018-01-10 Michael Collison <michael.collison@arm.com>
11017 * config/aarch64/aarch64-builtins.c:
11018 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11019 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11020 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11021 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11022 (AARCH64_ISA_SHA3): New ISA flag.
11023 (TARGET_SHA3): New feature flag for sha3.
11024 * config/aarch64/iterators.md (sha512_op): New int attribute.
11025 (CRYPTO_SHA512): New int iterator.
11026 (UNSPEC_SHA512H): New unspec.
11027 (UNSPEC_SHA512H2): Ditto.
11028 (UNSPEC_SHA512SU0): Ditto.
11029 (UNSPEC_SHA512SU1): Ditto.
11030 * config/aarch64/aarch64-simd-builtins.def
11031 (aarch64_crypto_sha512hqv2di): New builtin.
11032 (aarch64_crypto_sha512h2qv2di): Ditto.
11033 (aarch64_crypto_sha512su0qv2di): Ditto.
11034 (aarch64_crypto_sha512su1qv2di): Ditto.
11035 (aarch64_eor3qv8hi): Ditto.
11036 (aarch64_rax1qv2di): Ditto.
11037 (aarch64_xarqv2di): Ditto.
11038 (aarch64_bcaxqv8hi): Ditto.
11039 * config/aarch64/aarch64-simd.md:
11040 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11041 (aarch64_crypto_sha512su0qv2di): Ditto.
11042 (aarch64_crypto_sha512su1qv2di): Ditto.
11043 (aarch64_eor3qv8hi): Ditto.
11044 (aarch64_rax1qv2di): Ditto.
11045 (aarch64_xarqv2di): Ditto.
11046 (aarch64_bcaxqv8hi): Ditto.
11047 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11048 (vsha512h2q_u64): Ditto.
11049 (vsha512su0q_u64): Ditto.
11050 (vsha512su1q_u64): Ditto.
11051 (veor3q_u16): Ditto.
11052 (vrax1q_u64): Ditto.
11053 (vxarq_u64): Ditto.
11054 (vbcaxq_u16): Ditto.
11055 * config/arm/types.md (crypto_sha512): New type attribute.
11056 (crypto_sha3): Ditto.
11057 (doc/invoke.texi): Document new sha3 option.
11059 2018-01-10 Michael Collison <michael.collison@arm.com>
11061 * config/aarch64/aarch64-builtins.c:
11062 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11063 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11064 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11065 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11066 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11067 (AARCH64_ISA_SM4): New ISA flag.
11068 (TARGET_SM4): New feature flag for sm4.
11069 * config/aarch64/aarch64-simd-builtins.def
11070 (aarch64_sm3ss1qv4si): Ditto.
11071 (aarch64_sm3tt1aq4si): Ditto.
11072 (aarch64_sm3tt1bq4si): Ditto.
11073 (aarch64_sm3tt2aq4si): Ditto.
11074 (aarch64_sm3tt2bq4si): Ditto.
11075 (aarch64_sm3partw1qv4si): Ditto.
11076 (aarch64_sm3partw2qv4si): Ditto.
11077 (aarch64_sm4eqv4si): Ditto.
11078 (aarch64_sm4ekeyqv4si): Ditto.
11079 * config/aarch64/aarch64-simd.md:
11080 (aarch64_sm3ss1qv4si): Ditto.
11081 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11082 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11083 (aarch64_sm4eqv4si): Ditto.
11084 (aarch64_sm4ekeyqv4si): Ditto.
11085 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11086 (sm3part_op): Ditto.
11087 (CRYPTO_SM3TT): Ditto.
11088 (CRYPTO_SM3PART): Ditto.
11089 (UNSPEC_SM3SS1): New unspec.
11090 (UNSPEC_SM3TT1A): Ditto.
11091 (UNSPEC_SM3TT1B): Ditto.
11092 (UNSPEC_SM3TT2A): Ditto.
11093 (UNSPEC_SM3TT2B): Ditto.
11094 (UNSPEC_SM3PARTW1): Ditto.
11095 (UNSPEC_SM3PARTW2): Ditto.
11096 (UNSPEC_SM4E): Ditto.
11097 (UNSPEC_SM4EKEY): Ditto.
11098 * config/aarch64/constraints.md (Ui2): New constraint.
11099 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11100 * config/arm/types.md (crypto_sm3): New type attribute.
11101 (crypto_sm4): Ditto.
11102 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11103 (vsm3tt1aq_u32): Ditto.
11104 (vsm3tt1bq_u32): Ditto.
11105 (vsm3tt2aq_u32): Ditto.
11106 (vsm3tt2bq_u32): Ditto.
11107 (vsm3partw1q_u32): Ditto.
11108 (vsm3partw2q_u32): Ditto.
11109 (vsm4eq_u32): Ditto.
11110 (vsm4ekeyq_u32): Ditto.
11111 (doc/invoke.texi): Document new sm4 option.
11113 2018-01-10 Michael Collison <michael.collison@arm.com>
11115 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11116 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11117 (AARCH64_FL_FOR_ARCH8_4): New.
11118 (AARCH64_FL_V8_4): New flag.
11119 (doc/invoke.texi): Document new armv8.4-a option.
11121 2018-01-10 Michael Collison <michael.collison@arm.com>
11123 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11124 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11125 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11126 * config/aarch64/aarch64-option-extension.def: Add
11127 AARCH64_OPT_EXTENSION of 'sha2'.
11128 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11129 (crypto): Disable sha2 and aes if crypto disabled.
11130 (crypto): Enable aes and sha2 if enabled.
11131 (simd): Disable sha2 and aes if simd disabled.
11132 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11134 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11135 (TARGET_SHA2): New feature flag for sha2.
11136 (TARGET_AES): New feature flag for aes.
11137 * config/aarch64/aarch64-simd.md:
11138 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11139 conditional on TARGET_AES.
11140 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11141 (aarch64_crypto_sha1hsi): Make pattern conditional
11143 (aarch64_crypto_sha1hv4si): Ditto.
11144 (aarch64_be_crypto_sha1hv4si): Ditto.
11145 (aarch64_crypto_sha1su1v4si): Ditto.
11146 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11147 (aarch64_crypto_sha1su0v4si): Ditto.
11148 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11149 (aarch64_crypto_sha256su0v4si): Ditto.
11150 (aarch64_crypto_sha256su1v4si): Ditto.
11151 (doc/invoke.texi): Document new aes and sha2 options.
11153 2018-01-10 Martin Sebor <msebor@redhat.com>
11155 PR tree-optimization/83781
11156 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11159 2018-01-11 Martin Sebor <msebor@gmail.com>
11160 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11162 PR tree-optimization/83501
11163 PR tree-optimization/81703
11165 * tree-ssa-strlen.c (get_string_cst): Rename...
11166 (get_string_len): ...to this. Handle global constants.
11167 (handle_char_store): Adjust.
11169 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11170 Jim Wilson <jimw@sifive.com>
11172 * config/riscv/riscv-protos.h (riscv_output_return): New.
11173 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11174 (riscv_attribute_table, riscv_output_return),
11175 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11176 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11177 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11179 (riscv_expand_prologue): Add early return for naked function.
11180 (riscv_expand_epilogue): Likewise.
11181 (riscv_function_ok_for_sibcall): Return false for naked function.
11182 (riscv_set_current_function): New.
11183 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11184 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11185 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11186 * doc/extend.texi (RISC-V Function Attributes): New.
11188 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11190 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11191 check for 128-bit long double before checking TCmode.
11192 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11193 128-bit long doubles before checking TFmode or TCmode.
11194 (FLOAT128_IBM_P): Likewise.
11196 2018-01-10 Martin Sebor <msebor@redhat.com>
11198 PR tree-optimization/83671
11199 * builtins.c (c_strlen): Unconditionally return zero for the empty
11201 Use -Warray-bounds for warnings.
11202 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11203 for non-constant array indices with COMPONENT_REF, arrays of
11204 arrays, and pointers to arrays.
11205 (gimple_fold_builtin_strlen): Determine and set length range for
11206 non-constant character arrays.
11208 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11210 PR middle-end/81897
11211 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11214 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11216 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11218 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11221 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11222 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11223 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11224 indexed_or_indirect_operand predicate.
11225 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11226 (*vsx_le_perm_load_v8hi): Likewise.
11227 (*vsx_le_perm_load_v16qi): Likewise.
11228 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11229 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11230 (*vsx_le_perm_store_v8hi): Likewise.
11231 (*vsx_le_perm_store_v16qi): Likewise.
11232 (eight unnamed splitters): Likewise.
11234 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11236 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11237 * config/rs6000/emmintrin.h: Likewise.
11238 * config/rs6000/mmintrin.h: Likewise.
11239 * config/rs6000/xmmintrin.h: Likewise.
11241 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11244 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11246 * tree.c (tree_nop_conversion): Return true for location wrapper
11248 (maybe_wrap_with_location): New function.
11249 (selftest::check_strip_nops): New function.
11250 (selftest::test_location_wrappers): New function.
11251 (selftest::tree_c_tests): Call it.
11252 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11253 (maybe_wrap_with_location): New decl.
11254 (EXPR_LOCATION_WRAPPER_P): New macro.
11255 (location_wrapper_p): New inline function.
11256 (tree_strip_any_location_wrapper): New inline function.
11258 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11261 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11262 stack_realign_offset for the largest alignment of stack slot
11264 (ix86_find_max_used_stack_alignment): New function.
11265 (ix86_finalize_stack_frame_flags): Use it. Set
11266 max_used_stack_alignment if we don't realign stack.
11267 * config/i386/i386.h (machine_function): Add
11268 max_used_stack_alignment.
11270 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11272 * config/arm/arm.opt (-mbranch-cost): New option.
11273 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11276 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11279 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11280 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11282 2018-01-10 Richard Biener <rguenther@suse.de>
11285 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11286 early out so it also covers the case where we have a non-NULL
11289 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11291 PR tree-optimization/83753
11292 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11293 for non-strided grouped accesses if the number of elements is 1.
11295 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11298 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11299 * i386.h (TARGET_USE_GATHER): Define.
11300 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11302 2018-01-10 Martin Liska <mliska@suse.cz>
11305 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11306 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11308 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11309 CLEANUP_NO_PARTITIONING is not set.
11311 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11313 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11314 for vectors, as a partial revert of r254296.
11315 * rtl.h (const_vec_p): Delete.
11316 (const_vec_duplicate_p): Don't test for vector CONSTs.
11317 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11318 * expmed.c (make_tree): Likewise.
11321 * common.md (E, F): Use CONSTANT_P instead of checking for
11323 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11324 checking for CONST_VECTOR.
11326 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11328 PR middle-end/83575
11329 * predict.c (force_edge_cold): Handle in more sane way edges
11330 with no prediction.
11332 2018-01-09 Carl Love <cel@us.ibm.com>
11334 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11336 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11337 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11338 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11339 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11340 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11341 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11342 * config/rs6000/rs6000-protos.h: Add extern defition for
11343 rs6000_generate_float2_double_code.
11344 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11346 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11347 (float2_v2df): Add define_expand.
11349 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11352 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11353 op_mode in the force_to_mode call.
11355 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11357 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11358 instead of checking each element individually.
11359 (aarch64_evpc_uzp): Likewise.
11360 (aarch64_evpc_zip): Likewise.
11361 (aarch64_evpc_ext): Likewise.
11362 (aarch64_evpc_rev): Likewise.
11363 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11364 instead of checking each element individually. Return true without
11366 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11367 whether all selected elements come from the same input, instead of
11368 checking each element individually. Remove calls to gen_rtx_REG,
11369 start_sequence and end_sequence and instead assert that no rtl is
11372 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11374 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11375 order of HIGH and CONST checks.
11377 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11379 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11380 if the destination isn't an SSA_NAME.
11382 2018-01-09 Richard Biener <rguenther@suse.de>
11384 PR tree-optimization/83668
11385 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11387 (canonicalize_loop_form): ... here, renamed from ...
11388 (canonicalize_loop_closed_ssa_form): ... this and amended to
11389 swap successor edges for loop exit blocks to make us use
11390 the RPO order we need for initial schedule generation.
11392 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11394 PR tree-optimization/64811
11395 * match.pd: When optimizing comparisons with Inf, avoid
11396 introducing or losing exceptions from comparisons with NaN.
11398 2018-01-09 Martin Liska <mliska@suse.cz>
11401 * asan.c (shadow_mem_size): Add gcc_assert.
11403 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11405 Don't save registers in main().
11408 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11409 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11410 * config/avr/avr.c (avr_set_current_function): Don't error if
11411 naked, OS_task or OS_main are specified at the same time.
11412 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11414 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11416 * common/config/avr/avr-common.c (avr_option_optimization_table):
11417 Switch on -mmain-is-OS_task for optimizing compilations.
11419 2018-01-09 Richard Biener <rguenther@suse.de>
11421 PR tree-optimization/83572
11422 * graphite.c: Include cfganal.h.
11423 (graphite_transform_loops): Connect infinite loops to exit
11424 and remove fake edges at the end.
11426 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11428 * ipa-inline.c (edge_badness): Revert accidental checkin.
11430 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11433 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11434 symbols; not inline clones.
11436 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11439 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11440 hard registers. Formatting fixes.
11442 PR preprocessor/83722
11443 * gcc.c (try_generate_repro): Pass
11444 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11445 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11448 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11449 Kito Cheng <kito.cheng@gmail.com>
11451 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11452 (riscv_leaf_function_p): Delete.
11453 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11455 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11457 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11459 (do_ifelse): New function.
11460 (do_isel): New function.
11461 (do_sub3): New function.
11462 (do_add3): New function.
11463 (do_load_mask_compare): New function.
11464 (do_overlap_load_compare): New function.
11465 (expand_compare_loop): New function.
11466 (expand_block_compare): Call expand_compare_loop() when appropriate.
11467 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11468 option description.
11469 (-mblock-compare-inline-loop-limit): New option.
11471 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11474 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11475 Reverse order of second and third operands in first alternative.
11476 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11477 of first and second elements in UNSPEC_VPERMR vector.
11478 (altivec_expand_vec_perm_le): Likewise.
11480 2018-01-08 Jeff Law <law@redhat.com>
11482 PR rtl-optimizatin/81308
11483 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11484 (process_switch): If group_case_labels makes a change, then set
11486 (pass_convert_switch::execute): If a switch is converted, then
11487 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11489 PR rtl-optimization/81308
11490 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11493 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11495 PR target/83663 - Revert r255946
11496 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11497 generation for cases where splatting a value is not useful.
11498 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11499 across a vec_duplicate and a paradoxical subreg forming a vector
11500 mode to a vec_concat.
11502 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11504 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11505 -march=armv8.3-a variants.
11506 * config/arm/t-multilib: Likewise.
11507 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11509 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11511 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11513 (cceq_ior_compare_complement): Give it a name so I can use it, and
11514 change boolean_or_operator predicate to boolean_operator so it can
11515 be used to generate a crand.
11516 (eqne): New code iterator.
11517 (bd/bd_neg): New code_attrs.
11518 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11519 a single define_insn.
11520 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11521 decrement (bdnzt/bdnzf/bdzt/bdzf).
11522 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11523 with the new names of the branch decrement patterns, and added the
11524 names of the branch decrement conditional patterns.
11526 2018-01-08 Richard Biener <rguenther@suse.de>
11528 PR tree-optimization/83563
11529 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11532 2018-01-08 Richard Biener <rguenther@suse.de>
11534 PR middle-end/83713
11535 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11537 2018-01-08 Richard Biener <rguenther@suse.de>
11539 PR tree-optimization/83685
11540 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11541 references to abnormals.
11543 2018-01-08 Richard Biener <rguenther@suse.de>
11546 * dwarf2out.c (output_indirect_strings): Handle empty
11547 skeleton_debug_str_hash.
11548 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11550 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11552 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11553 (emit_store_direct): Likewise.
11554 (arc_trampoline_adjust_address): Likewise.
11555 (arc_asm_trampoline_template): New function.
11556 (arc_initialize_trampoline): Use asm_trampoline_template.
11557 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11558 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11559 * config/arc/arc.md (flush_icache): Delete pattern.
11561 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11563 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11564 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11567 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11570 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11571 by not USED_FOR_TARGET.
11572 (make_pass_resolve_sw_modes): Likewise.
11574 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11576 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11579 2018-01-08 Richard Biener <rguenther@suse.de>
11581 PR middle-end/83580
11582 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11584 2018-01-08 Richard Biener <rguenther@suse.de>
11586 PR middle-end/83517
11587 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11589 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11591 PR middle-end/81897
11592 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11593 basic blocks with a small number of successors.
11594 (convert_control_dep_chain_into_preds): Improve handling of
11596 (dump_predicates): Split apart into...
11597 (dump_pred_chain): ...here...
11598 (dump_pred_info): ...and here.
11599 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11600 (can_chain_union_be_invalidated_p): Improve check for invalidation
11602 (uninit_uses_cannot_happen): Avoid unnecessary if
11603 convert_control_dep_chain_into_preds yielded nothing.
11605 2018-01-06 Martin Sebor <msebor@redhat.com>
11607 PR tree-optimization/83640
11608 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11609 subtracting negative offset from size.
11610 (builtin_access::overlap): Adjust offset bounds of the access to fall
11611 within the size of the object if possible.
11613 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11615 PR rtl-optimization/83699
11616 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11617 extract_bit_field_as_subreg to cases in which the extracted
11618 value is also a vector.
11620 * lra-constraints.c (process_alt_operands): Test for the equivalence
11621 substitutions when detecting a possible reload cycle.
11623 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11626 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11627 by default if flag_selective_schedling{,2}. Formatting fixes.
11629 PR rtl-optimization/83682
11630 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11631 if it has non-VECTOR_MODE element mode.
11632 (vec_duplicate_p): Likewise.
11634 PR middle-end/83694
11635 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11636 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11638 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11641 * config/i386/i386-builtin.def
11642 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11643 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11644 Require also OPTION_MASK_ISA_AVX512F in addition to
11645 OPTION_MASK_ISA_GFNI.
11646 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11647 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11648 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11649 to OPTION_MASK_ISA_GFNI.
11650 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11651 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11652 OPTION_MASK_ISA_AVX512BW.
11653 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11654 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11655 addition to OPTION_MASK_ISA_GFNI.
11656 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11657 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11658 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11659 to OPTION_MASK_ISA_GFNI.
11660 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11661 a requirement for all ISAs rather than any of them with a few
11663 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11665 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11666 bitmasks to be enabled with 3 exceptions, instead of requiring any
11667 enabled ISA with lots of exceptions.
11668 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11669 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11670 Change avx512bw in isa attribute to avx512f.
11671 * config/i386/sgxintrin.h: Add license boilerplate.
11672 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11673 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11674 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11675 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11677 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11678 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11679 temporarily sse2 rather than sse if not enabled already.
11682 * config/i386/sse.md (VI248_VLBW): Rename to ...
11683 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11684 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11685 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11686 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11687 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11688 mode iterator instead of VI248_VLBW.
11690 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11692 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11693 (record_modified): Skip clobbers; add debug output.
11694 (param_change_prob): Use sreal frequencies.
11696 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11698 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11699 punt for user-aligned variables.
11701 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11703 * tree-chrec.c (chrec_contains_symbols): Return true for
11706 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11709 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11710 of (x|y) == x for BICS pattern.
11712 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11714 PR tree-optimization/83605
11715 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11716 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11719 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11721 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11722 * config/epiphany/rtems.h: New file.
11724 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11725 Uros Bizjak <ubizjak@gmail.com>
11728 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11729 QIreg_operand instead of register_operand predicate.
11730 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11731 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11732 comments instead of -fmitigate[-_]rop.
11734 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11737 * cgraphunit.c (symbol_table::compile): Switch to text_section
11738 before calling assembly_start debug hook.
11739 * run-rtl-passes.c (run_rtl_passes): Likewise.
11742 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11744 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11745 range_int_cst_p rather than !symbolic_range_p before calling
11746 extract_range_from_multiplicative_op_1.
11748 2018-01-04 Jeff Law <law@redhat.com>
11750 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11751 redundant test in assertion.
11753 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11755 * doc/rtl.texi: Document machine_mode wrapper classes.
11757 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11759 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11760 using tree_to_uhwi.
11762 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11764 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11765 the VEC_PERM_EXPR fold to fail.
11767 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11770 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11771 to switched_sections.
11773 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11776 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11777 test for d.testing.
11779 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11782 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11783 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11785 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11788 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11789 is BLKmode and bitpos not zero or mode change is needed.
11791 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11794 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11797 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11800 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11801 instead of MULT rtx. Update all corresponding splitters.
11802 (*saddl_se): Ditto.
11803 (*ssub<modesuffix>): Ditto.
11804 (*ssubl_se): Ditto.
11805 (*cmp_sadd_di): Update split patterns.
11806 (*cmp_sadd_si): Ditto.
11807 (*cmp_sadd_sidi): Ditto.
11808 (*cmp_ssub_di): Ditto.
11809 (*cmp_ssub_si): Ditto.
11810 (*cmp_ssub_sidi): Ditto.
11811 * config/alpha/predicates.md (const23_operand): New predicate.
11812 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11813 Look for ASHIFT, not MULT inner operand.
11814 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11816 2018-01-04 Martin Liska <mliska@suse.cz>
11818 PR gcov-profile/83669
11819 * gcov.c (output_intermediate_file): Add version to intermediate
11821 * doc/gcov.texi: Document new field 'version' in intermediate
11822 file format. Fix location of '-k' option of gcov command.
11824 2018-01-04 Martin Liska <mliska@suse.cz>
11827 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11829 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11831 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11833 2018-01-03 Martin Sebor <msebor@redhat.com>
11835 PR tree-optimization/83655
11836 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11837 checking calls with invalid arguments.
11839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11841 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11842 (vectorizable_mask_load_store): Delete.
11843 (vectorizable_call): Return false for masked loads and stores.
11844 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11845 instead of gimple_assign_rhs1.
11846 (vectorizable_load): Handle IFN_MASK_LOAD.
11847 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11851 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11853 (vectorizable_mask_load_store): ...here.
11854 (vectorizable_load): ...and here.
11856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11858 * tree-vect-stmts.c (vect_build_all_ones_mask)
11859 (vect_build_zero_merge_argument): New functions, split out from...
11860 (vectorizable_load): ...here.
11862 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11864 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11866 (vectorizable_mask_load_store): ...here.
11867 (vectorizable_store): ...and here.
11869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11871 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11873 (vectorizable_mask_load_store): ...here.
11875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11877 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11878 (vect_model_store_cost): Take a vec_load_store_type instead of a
11880 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11881 (vect_model_store_cost): Take a vec_load_store_type instead of a
11883 (vectorizable_mask_load_store): Update accordingly.
11884 (vectorizable_store): Likewise.
11885 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11887 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11889 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11890 IFN_MASK_LOAD calls here rather than...
11891 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11894 Alan Hayward <alan.hayward@arm.com>
11895 David Sherwood <david.sherwood@arm.com>
11897 * expmed.c (extract_bit_field_1): For vector extracts,
11898 fall back to extract_bit_field_as_subreg if vec_extract
11901 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11902 Alan Hayward <alan.hayward@arm.com>
11903 David Sherwood <david.sherwood@arm.com>
11905 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11906 they are variable or constant sized.
11907 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11908 slots for constant-sized data.
11910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11911 Alan Hayward <alan.hayward@arm.com>
11912 David Sherwood <david.sherwood@arm.com>
11914 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11915 handling COND_EXPRs with boolean comparisons, try to find a better
11916 basis for the mask type than the boolean itself.
11918 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11920 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11921 is calculated and how it can be overridden.
11922 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11923 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11925 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11928 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11929 Alan Hayward <alan.hayward@arm.com>
11930 David Sherwood <david.sherwood@arm.com>
11932 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11933 Remove the mode argument.
11934 (aarch64_simd_valid_immediate): Remove the mode and inverse
11936 * config/aarch64/iterators.md (bitsize): New iterator.
11937 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11938 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11939 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11940 aarch64_simd_valid_immediate.
11941 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11942 (aarch64_reg_or_bic_imm): Likewise.
11943 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11944 with an insn_type enum and msl with a modifier_type enum.
11945 Replace element_width with a scalar_mode. Change the shift
11946 to unsigned int. Add constructors for scalar_float_mode and
11947 scalar_int_mode elements.
11948 (aarch64_vect_float_const_representable_p): Delete.
11949 (aarch64_can_const_movi_rtx_p)
11950 (aarch64_simd_scalar_immediate_valid_for_move)
11951 (aarch64_simd_make_constant): Update call to
11952 aarch64_simd_valid_immediate.
11953 (aarch64_advsimd_valid_immediate_hs): New function.
11954 (aarch64_advsimd_valid_immediate): Likewise.
11955 (aarch64_simd_valid_immediate): Remove mode and inverse
11956 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11957 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11958 and aarch64_float_const_representable_p on the result.
11959 (aarch64_output_simd_mov_immediate): Remove mode argument.
11960 Update call to aarch64_simd_valid_immediate and use of
11961 simd_immediate_info.
11962 (aarch64_output_scalar_simd_mov_immediate): Update call
11965 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11966 Alan Hayward <alan.hayward@arm.com>
11967 David Sherwood <david.sherwood@arm.com>
11969 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11970 (mode_nunits): Likewise CONST_MODE_NUNITS.
11971 * machmode.def (ADJUST_NUNITS): Document.
11972 * genmodes.c (mode_data::need_nunits_adj): New field.
11973 (blank_mode): Update accordingly.
11974 (adj_nunits): New variable.
11975 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11977 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11978 listed in adj_nunits.
11979 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11980 listed in adj_nunits. Don't emit case statements for such modes.
11981 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11982 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11983 nothing if adj_nunits is nonnull.
11984 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11985 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11986 (emit_mode_fbit): Update use of print_maybe_const_decl.
11987 (emit_move_size): Likewise. Treat the array as non-const
11989 (emit_mode_adjustments): Handle adj_nunits.
11991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11993 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11994 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11995 (VECTOR_MODES): Use it.
11996 (make_vector_modes): Take the prefix as an argument.
11998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11999 Alan Hayward <alan.hayward@arm.com>
12000 David Sherwood <david.sherwood@arm.com>
12002 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12003 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12004 for MODE_VECTOR_BOOL.
12005 * machmode.def (VECTOR_BOOL_MODE): Document.
12006 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12007 (make_vector_bool_mode): New function.
12008 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12010 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12011 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12013 * stor-layout.c (int_mode_for_mode): Likewise.
12014 * tree.c (build_vector_type_for_mode): Likewise.
12015 * varasm.c (output_constant_pool_2): Likewise.
12016 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12017 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12018 for MODE_VECTOR_BOOL.
12019 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12020 of mode class checks.
12021 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12022 instead of a list of mode class checks.
12023 (expand_vector_scalar_condition): Likewise.
12024 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12027 Alan Hayward <alan.hayward@arm.com>
12028 David Sherwood <david.sherwood@arm.com>
12030 * machmode.h (mode_size): Change from unsigned short to
12032 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12033 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12034 or if measurement_type is not polynomial.
12035 (fixed_size_mode::includes_p): Check for constant-sized modes.
12036 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12037 return a poly_uint16 rather than an unsigned short.
12038 (emit_mode_size): Change the type of mode_size from unsigned short
12039 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12040 (emit_mode_adjustments): Cope with polynomial vector sizes.
12041 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12043 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12045 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12046 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12047 * caller-save.c (setup_save_areas): Likewise.
12048 (replace_reg_with_saved_mem): Likewise.
12049 * calls.c (emit_library_call_value_1): Likewise.
12050 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12051 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12052 (gen_lowpart_for_combine): Likewise.
12053 * convert.c (convert_to_integer_1): Likewise.
12054 * cse.c (equiv_constant, cse_insn): Likewise.
12055 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12056 (cselib_subst_to_values): Likewise.
12057 * dce.c (word_dce_process_block): Likewise.
12058 * df-problems.c (df_word_lr_mark_ref): Likewise.
12059 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12060 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12061 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12062 (rtl_for_decl_location): Likewise.
12063 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12064 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12065 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12066 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12067 (expand_expr_real_1): Likewise.
12068 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12069 (pad_below): Likewise.
12070 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12071 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12072 * ira.c (get_subreg_tracking_sizes): Likewise.
12073 * ira-build.c (ira_create_allocno_objects): Likewise.
12074 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12075 (ira_sort_regnos_for_alter_reg): Likewise.
12076 * ira-costs.c (record_operand_costs): Likewise.
12077 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12078 (resolve_simple_move): Likewise.
12079 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12080 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12081 (lra_constraints): Likewise.
12082 (CONST_POOL_OK_P): Reject variable-sized modes.
12083 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12084 (add_pseudo_to_slot, lra_spill): Likewise.
12085 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12086 * optabs-query.c (get_best_extraction_insn): Likewise.
12087 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12088 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12089 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12090 * recog.c (offsettable_address_addr_space_p): Likewise.
12091 * regcprop.c (maybe_mode_change): Likewise.
12092 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12093 * regrename.c (build_def_use): Likewise.
12094 * regstat.c (dump_reg_info): Likewise.
12095 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12096 (find_reloads, find_reloads_subreg_address): Likewise.
12097 * reload1.c (eliminate_regs_1): Likewise.
12098 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12099 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12100 (simplify_binary_operation_1, simplify_subreg): Likewise.
12101 * targhooks.c (default_function_arg_padding): Likewise.
12102 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12103 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12104 (verify_gimple_assign_ternary): Likewise.
12105 * tree-inline.c (estimate_move_cost): Likewise.
12106 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12107 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12108 (get_address_cost_ainc): Likewise.
12109 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12110 (vect_supportable_dr_alignment): Likewise.
12111 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12112 (vectorizable_reduction): Likewise.
12113 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12114 (vectorizable_operation, vectorizable_load): Likewise.
12115 * tree.c (build_same_sized_truth_vector_type): Likewise.
12116 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12117 * var-tracking.c (emit_note_insn_var_location): Likewise.
12118 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12119 (ADDR_VEC_ALIGN): Likewise.
12121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12122 Alan Hayward <alan.hayward@arm.com>
12123 David Sherwood <david.sherwood@arm.com>
12125 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12127 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12128 or if measurement_type is polynomial.
12129 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12130 * combine.c (make_extraction): Likewise.
12131 * dse.c (find_shift_sequence): Likewise.
12132 * dwarf2out.c (mem_loc_descriptor): Likewise.
12133 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12134 (extract_bit_field, extract_low_bits): Likewise.
12135 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12136 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12137 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12138 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12139 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12140 * reload.c (find_reloads): Likewise.
12141 * reload1.c (alter_reg): Likewise.
12142 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12143 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12144 * tree-if-conv.c (predicate_mem_writes): Likewise.
12145 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12146 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12147 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12148 * valtrack.c (dead_debug_insert_temp): Likewise.
12149 * varasm.c (mergeable_constant_section): Likewise.
12150 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12152 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12153 Alan Hayward <alan.hayward@arm.com>
12154 David Sherwood <david.sherwood@arm.com>
12156 * expr.c (expand_assignment): Cope with polynomial mode sizes
12157 when assigning to a CONCAT.
12159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12160 Alan Hayward <alan.hayward@arm.com>
12161 David Sherwood <david.sherwood@arm.com>
12163 * machmode.h (mode_precision): Change from unsigned short to
12165 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12167 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12168 or if measurement_type is not polynomial.
12169 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12170 in which the mode is already known to be a scalar_int_mode.
12171 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12172 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12174 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12175 for GET_MODE_PRECISION.
12176 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12177 for GET_MODE_PRECISION.
12178 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12180 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12181 (expand_field_assignment, make_extraction): Likewise.
12182 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12183 (get_last_value): Likewise.
12184 * convert.c (convert_to_integer_1): Likewise.
12185 * cse.c (cse_insn): Likewise.
12186 * expr.c (expand_expr_real_1): Likewise.
12187 * lra-constraints.c (simplify_operand_subreg): Likewise.
12188 * optabs-query.c (can_atomic_load_p): Likewise.
12189 * optabs.c (expand_atomic_load): Likewise.
12190 (expand_atomic_store): Likewise.
12191 * ree.c (combine_reaching_defs): Likewise.
12192 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12193 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12194 * tree.h (type_has_mode_precision_p): Likewise.
12195 * ubsan.c (instrument_si_overflow): Likewise.
12197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12198 Alan Hayward <alan.hayward@arm.com>
12199 David Sherwood <david.sherwood@arm.com>
12201 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12202 polynomial numbers of units.
12203 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12204 (valid_vector_subparts_p): New function.
12205 (build_vector_type): Remove temporary shim and take the number
12206 of units as a poly_uint64 rather than an int.
12207 (build_opaque_vector_type): Take the number of units as a
12208 poly_uint64 rather than an int.
12209 * tree.c (build_vector_from_ctor): Handle polynomial
12210 TYPE_VECTOR_SUBPARTS.
12211 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12212 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12213 (build_vector_from_val): If the number of units is variable,
12214 use build_vec_duplicate_cst for constant operands and
12215 VEC_DUPLICATE_EXPR otherwise.
12216 (make_vector_type): Remove temporary is_constant ().
12217 (build_vector_type, build_opaque_vector_type): Take the number of
12218 units as a poly_uint64 rather than an int.
12219 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12221 * cfgexpand.c (expand_debug_expr): Likewise.
12222 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12223 (store_constructor, expand_expr_real_1): Likewise.
12224 (const_scalar_mask_from_tree): Likewise.
12225 * fold-const-call.c (fold_const_reduction): Likewise.
12226 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12227 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12228 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12229 (fold_relational_const): Likewise.
12230 (native_interpret_vector): Likewise. Change the size from an
12231 int to an unsigned int.
12232 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12233 TYPE_VECTOR_SUBPARTS.
12234 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12235 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12236 duplicating a non-constant operand into a variable-length vector.
12237 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12238 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12239 * ipa-icf.c (sem_variable::equals): Likewise.
12240 * match.pd: Likewise.
12241 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12242 * print-tree.c (print_node): Likewise.
12243 * stor-layout.c (layout_type): Likewise.
12244 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12245 * tree-cfg.c (verify_gimple_comparison): Likewise.
12246 (verify_gimple_assign_binary): Likewise.
12247 (verify_gimple_assign_ternary): Likewise.
12248 (verify_gimple_assign_single): Likewise.
12249 * tree-pretty-print.c (dump_generic_node): Likewise.
12250 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12251 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12252 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12253 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12254 (vect_shift_permute_load_chain): Likewise.
12255 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12256 (expand_vector_condition, optimize_vector_constructor): Likewise.
12257 (lower_vec_perm, get_compute_type): Likewise.
12258 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12259 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12260 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12261 (vect_recog_mask_conversion_pattern): Likewise.
12262 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12263 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12264 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12265 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12266 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12267 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12268 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12269 (supportable_widening_operation): Likewise.
12270 (supportable_narrowing_operation): Likewise.
12271 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12273 * varasm.c (output_constant): Likewise.
12275 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12276 Alan Hayward <alan.hayward@arm.com>
12277 David Sherwood <david.sherwood@arm.com>
12279 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12280 so that both the length == 3 and length != 3 cases set up their
12281 own permute vectors. Add comments explaining why we know the
12282 number of elements is constant.
12283 (vect_permute_load_chain): Likewise.
12285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12286 Alan Hayward <alan.hayward@arm.com>
12287 David Sherwood <david.sherwood@arm.com>
12289 * machmode.h (mode_nunits): Change from unsigned char to
12291 (ONLY_FIXED_SIZE_MODES): New macro.
12292 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12293 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12294 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12296 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12297 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12298 or if measurement_type is not polynomial.
12299 * genmodes.c (ZERO_COEFFS): New macro.
12300 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12302 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12303 Use ZERO_COEFFS when emitting initializers.
12304 * data-streamer.h (bp_pack_poly_value): New function.
12305 (bp_unpack_poly_value): Likewise.
12306 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12307 for GET_MODE_NUNITS.
12308 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12309 for GET_MODE_NUNITS.
12310 * tree.c (make_vector_type): Remove temporary shim and make
12311 the real function take the number of units as a poly_uint64
12312 rather than an int.
12313 (build_vector_type_for_mode): Handle polynomial nunits.
12314 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12315 * emit-rtl.c (const_vec_series_p_1): Likewise.
12316 (gen_rtx_CONST_VECTOR): Likewise.
12317 * fold-const.c (test_vec_duplicate_folding): Likewise.
12318 * genrecog.c (validate_pattern): Likewise.
12319 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12320 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12321 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12322 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12323 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12324 * rtlanal.c (subreg_get_info): Likewise.
12325 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12326 (vect_grouped_load_supported): Likewise.
12327 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12328 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12329 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12330 (simplify_const_unary_operation, simplify_binary_operation_1)
12331 (simplify_const_binary_operation, simplify_ternary_operation)
12332 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12333 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12334 instead of CONST_VECTOR_NUNITS.
12335 * varasm.c (output_constant_pool_2): Likewise.
12336 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12337 explicit-encoded elements in the XVEC for variable-length vectors.
12339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12341 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12344 Alan Hayward <alan.hayward@arm.com>
12345 David Sherwood <david.sherwood@arm.com>
12347 * coretypes.h (fixed_size_mode): Declare.
12348 (fixed_size_mode_pod): New typedef.
12349 * builtins.h (target_builtins::x_apply_args_mode)
12350 (target_builtins::x_apply_result_mode): Change type to
12351 fixed_size_mode_pod.
12352 * builtins.c (apply_args_size, apply_result_size, result_vector)
12353 (expand_builtin_apply_args_1, expand_builtin_apply)
12354 (expand_builtin_return): Update accordingly.
12356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12358 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12359 * cselib.c (cselib_hash_rtx): Likewise.
12360 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12361 CONST_VECTOR encoding.
12363 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12364 Jeff Law <law@redhat.com>
12367 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12368 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12369 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12370 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12373 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12374 explicitly probe *sp in a noreturn function if there were any callee
12375 register saves or frame pointer is needed.
12377 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12380 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12381 BLKmode for ternary, binary or unary expressions.
12384 * var-tracking.c (delete_vta_debug_insn): New inline function.
12385 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12386 insns from get_insns () to NULL instead of each bb separately.
12387 Use delete_vta_debug_insn. No longer static.
12388 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12389 delete_vta_debug_insns callers.
12390 * rtl.h (delete_vta_debug_insns): Declare.
12391 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12392 instead of variable_tracking_main.
12394 2018-01-03 Martin Sebor <msebor@redhat.com>
12396 PR tree-optimization/83603
12397 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12398 arguments past the endof the argument list in functions declared
12399 without a prototype.
12400 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12401 Avoid checking when arguments are null.
12403 2018-01-03 Martin Sebor <msebor@redhat.com>
12406 * doc/extend.texi (attribute const): Fix a typo.
12407 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12408 issuing -Wsuggest-attribute for void functions.
12410 2018-01-03 Martin Sebor <msebor@redhat.com>
12412 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12413 offset_int::from instead of wide_int::to_shwi.
12414 (maybe_diag_overlap): Remove assertion.
12415 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12416 * gimple-ssa-sprintf.c (format_directive): Same.
12417 (parse_directive): Same.
12418 (sprintf_dom_walker::compute_format_length): Same.
12419 (try_substitute_return_value): Same.
12421 2018-01-03 Jeff Law <law@redhat.com>
12423 PR middle-end/83654
12424 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12425 non-constant residual for zero at runtime and avoid probing in
12426 that case. Reorganize code for trailing problem to mirror handling
12429 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12431 PR tree-optimization/83501
12432 * tree-ssa-strlen.c (get_string_cst): New.
12433 (handle_char_store): Call get_string_cst.
12435 2018-01-03 Martin Liska <mliska@suse.cz>
12437 PR tree-optimization/83593
12438 * tree-ssa-strlen.c: Include tree-cfg.h.
12439 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12440 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12441 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12443 (strlen_dom_walker::before_dom_children): Call
12444 gimple_purge_dead_eh_edges. Dump tranformation with details
12446 (strlen_dom_walker::before_dom_children): Update call by adding
12447 new argument cleanup_eh.
12448 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12450 2018-01-03 Martin Liska <mliska@suse.cz>
12453 * cif-code.def (VARIADIC_THUNK): New enum value.
12454 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12457 2018-01-03 Jan Beulich <jbeulich@suse.com>
12459 * sse.md (mov<mode>_internal): Tighten condition for when to use
12460 vmovdqu<ssescalarsize> for TI and OI modes.
12462 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12464 Update copyright years.
12466 2018-01-03 Martin Liska <mliska@suse.cz>
12469 * ipa-visibility.c (function_and_variable_visibility): Skip
12470 functions with noipa attribure.
12472 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12474 * gcc.c (process_command): Update copyright notice dates.
12475 * gcov-dump.c (print_version): Ditto.
12476 * gcov.c (print_version): Ditto.
12477 * gcov-tool.c (print_version): Ditto.
12478 * gengtype.c (create_file): Ditto.
12479 * doc/cpp.texi: Bump @copying's copyright year.
12480 * doc/cppinternals.texi: Ditto.
12481 * doc/gcc.texi: Ditto.
12482 * doc/gccint.texi: Ditto.
12483 * doc/gcov.texi: Ditto.
12484 * doc/install.texi: Ditto.
12485 * doc/invoke.texi: Ditto.
12487 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12489 * vector-builder.h (vector_builder::m_full_nelts): Change from
12490 unsigned int to poly_uint64.
12491 (vector_builder::full_nelts): Update prototype accordingly.
12492 (vector_builder::new_vector): Likewise.
12493 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12494 (vector_builder::operator ==): Likewise.
12495 (vector_builder::finalize): Likewise.
12496 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12497 Take the number of elements as a poly_uint64 rather than an
12499 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12500 from unsigned int to poly_uint64.
12501 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12502 (vec_perm_indices::new_vector): Likewise.
12503 (vec_perm_indices::length): Likewise.
12504 (vec_perm_indices::nelts_per_input): Likewise.
12505 (vec_perm_indices::input_nelts): Likewise.
12506 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12507 number of elements per input as a poly_uint64 rather than an
12508 unsigned int. Use the original encoding for variable-length
12509 vectors, rather than clamping each individual element.
12510 For the second and subsequent elements in each pattern,
12511 clamp the step and base before clamping their sum.
12512 (vec_perm_indices::series_p): Handle polynomial element counts.
12513 (vec_perm_indices::all_in_range_p): Likewise.
12514 (vec_perm_indices_to_tree): Likewise.
12515 (vec_perm_indices_to_rtx): Likewise.
12516 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12517 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12518 (tree_vector_builder::new_binary_operation): Handle polynomial
12519 element counts. Return false if we need to know the number
12520 of elements at compile time.
12521 * fold-const.c (fold_vec_perm): Punt if the number of elements
12522 isn't known at compile time.
12524 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12526 * vec-perm-indices.h (vec_perm_builder): Change element type
12527 from HOST_WIDE_INT to poly_int64.
12528 (vec_perm_indices::element_type): Update accordingly.
12529 (vec_perm_indices::clamp): Handle polynomial element_types.
12530 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12531 (vec_perm_indices::all_in_range_p): Likewise.
12532 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12534 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12535 polynomial vec_perm_indices element types.
12536 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12537 * fold-const.c (fold_vec_perm): Likewise.
12538 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12539 * tree-vect-generic.c (lower_vec_perm): Likewise.
12540 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12541 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12542 element type to HOST_WIDE_INT.
12544 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12545 Alan Hayward <alan.hayward@arm.com>
12546 David Sherwood <david.sherwood@arm.com>
12548 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12549 rather than an int. Use plus_constant.
12550 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12551 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12554 Alan Hayward <alan.hayward@arm.com>
12555 David Sherwood <david.sherwood@arm.com>
12557 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12558 a HOST_WIDE_INT to a poly_int64.
12560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12561 Alan Hayward <alan.hayward@arm.com>
12562 David Sherwood <david.sherwood@arm.com>
12564 * calls.c (load_register_parameters): Cope with polynomial
12565 mode sizes. Require a constant size for BLKmode parameters
12566 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12567 forces a parameter to be padded at the lsb end in order to
12568 fill a complete number of words, require the parameter size
12569 to be ordered wrt UNITS_PER_WORD.
12571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12572 Alan Hayward <alan.hayward@arm.com>
12573 David Sherwood <david.sherwood@arm.com>
12575 * reload1.c (spill_stack_slot_width): Change element type
12576 from unsigned int to poly_uint64_pod.
12577 (alter_reg): Treat mode sizes as polynomial.
12579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12580 Alan Hayward <alan.hayward@arm.com>
12581 David Sherwood <david.sherwood@arm.com>
12583 * reload.c (complex_word_subreg_p): New function.
12584 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
12591 operands whose sizes aren't ordered.
12592 (match_reload): Refer to this check here.
12594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12595 Alan Hayward <alan.hayward@arm.com>
12596 David Sherwood <david.sherwood@arm.com>
12598 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12599 that the mode size is in the set {1, 2, 4, 8, 16}.
12601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12602 Alan Hayward <alan.hayward@arm.com>
12603 David Sherwood <david.sherwood@arm.com>
12605 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12606 Use plus_constant instead of gen_rtx_PLUS.
12608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12609 Alan Hayward <alan.hayward@arm.com>
12610 David Sherwood <david.sherwood@arm.com>
12612 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12613 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12614 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12615 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12616 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12617 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12618 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12619 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12620 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12621 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12623 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12624 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12625 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12626 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12627 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12628 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12629 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12630 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12631 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12632 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12634 * expr.c (emit_move_resolve_push): Treat the input and result
12635 of PUSH_ROUNDING as a poly_int64.
12636 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12637 (emit_push_insn): Likewise.
12638 * lra-eliminations.c (mark_not_eliminable): Likewise.
12639 * recog.c (push_operand): Likewise.
12640 * reload1.c (elimination_effects): Likewise.
12641 * rtlanal.c (nonzero_bits1): Likewise.
12642 * calls.c (store_one_arg): Likewise. Require the padding to be
12643 known at compile time.
12645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12646 Alan Hayward <alan.hayward@arm.com>
12647 David Sherwood <david.sherwood@arm.com>
12649 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12650 Use plus_constant instead of gen_rtx_PLUS.
12652 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12653 Alan Hayward <alan.hayward@arm.com>
12654 David Sherwood <david.sherwood@arm.com>
12656 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12657 rather than an int.
12659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12660 Alan Hayward <alan.hayward@arm.com>
12661 David Sherwood <david.sherwood@arm.com>
12663 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12664 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12665 via stack temporaries. Treat the mode size as polynomial too.
12667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12668 Alan Hayward <alan.hayward@arm.com>
12669 David Sherwood <david.sherwood@arm.com>
12671 * expr.c (expand_expr_real_2): When handling conversions involving
12672 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12673 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12674 as a poly_uint64 too.
12676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12677 Alan Hayward <alan.hayward@arm.com>
12678 David Sherwood <david.sherwood@arm.com>
12680 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12683 Alan Hayward <alan.hayward@arm.com>
12684 David Sherwood <david.sherwood@arm.com>
12686 * combine.c (can_change_dest_mode): Handle polynomial
12687 REGMODE_NATURAL_SIZE.
12688 * expmed.c (store_bit_field_1): Likewise.
12689 * expr.c (store_constructor): Likewise.
12690 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12691 and polynomial REGMODE_NATURAL_SIZE.
12692 (gen_lowpart_common): Likewise.
12693 * reginfo.c (record_subregs_of_mode): Likewise.
12694 * rtlanal.c (read_modify_subreg_p): Likewise.
12696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12697 Alan Hayward <alan.hayward@arm.com>
12698 David Sherwood <david.sherwood@arm.com>
12700 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12701 numbers of elements.
12703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12704 Alan Hayward <alan.hayward@arm.com>
12705 David Sherwood <david.sherwood@arm.com>
12707 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12714 in a POINTER_PLUS_EXPR.
12716 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12717 Alan Hayward <alan.hayward@arm.com>
12718 David Sherwood <david.sherwood@arm.com>
12720 * omp-simd-clone.c (simd_clone_subparts): New function.
12721 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12722 (ipa_simd_modify_function_body): Likewise.
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 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12729 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12730 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12731 (expand_vector_condition, vector_element): Likewise.
12732 (subparts_gt): New function.
12733 (get_compute_type): Use subparts_gt.
12734 (count_type_subparts): Delete.
12735 (expand_vector_operations_1): Use subparts_gt instead of
12736 count_type_subparts.
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 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12743 (vect_compile_time_alias): ...this new function. Do the calculation
12744 on poly_ints rather than trees.
12745 (vect_prune_runtime_alias_test_list): Update call accordingly.
12747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12748 Alan Hayward <alan.hayward@arm.com>
12749 David Sherwood <david.sherwood@arm.com>
12751 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12753 (vect_schedule_slp_instance): Likewise.
12755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12756 Alan Hayward <alan.hayward@arm.com>
12757 David Sherwood <david.sherwood@arm.com>
12759 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12760 constant and extern definitions for variable-length vectors.
12761 (vect_get_constant_vectors): Note that the number of units
12762 is known to be constant.
12764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12765 Alan Hayward <alan.hayward@arm.com>
12766 David Sherwood <david.sherwood@arm.com>
12768 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12769 of units as polynomial. Choose between WIDE and NARROW based
12772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12773 Alan Hayward <alan.hayward@arm.com>
12774 David Sherwood <david.sherwood@arm.com>
12776 * tree-vect-stmts.c (simd_clone_subparts): New function.
12777 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12780 Alan Hayward <alan.hayward@arm.com>
12781 David Sherwood <david.sherwood@arm.com>
12783 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12784 vectors as polynomial. Use build_index_vector for
12785 IFN_GOMP_SIMD_LANE.
12787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12788 Alan Hayward <alan.hayward@arm.com>
12789 David Sherwood <david.sherwood@arm.com>
12791 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12792 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12793 for variable-length vectors.
12794 (vectorizable_mask_load_store): Treat the number of units as
12795 polynomial, asserting that it is constant if the condition has
12796 already been enforced.
12797 (vectorizable_store, vectorizable_load): Likewise.
12799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12800 Alan Hayward <alan.hayward@arm.com>
12801 David Sherwood <david.sherwood@arm.com>
12803 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12804 of units as polynomial. Punt if we can't tell at compile time
12805 which vector contains the final result.
12807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12808 Alan Hayward <alan.hayward@arm.com>
12809 David Sherwood <david.sherwood@arm.com>
12811 * tree-vect-loop.c (vectorizable_induction): Treat the number
12812 of units as polynomial. Punt on SLP inductions. Use an integer
12813 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12814 cast of such a series for variable-length floating-point
12817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12818 Alan Hayward <alan.hayward@arm.com>
12819 David Sherwood <david.sherwood@arm.com>
12821 * tree.h (build_index_vector): Declare.
12822 * tree.c (build_index_vector): New function.
12823 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12824 of units as polynomial, forcibly converting it to a constant if
12825 vectorizable_reduction has already enforced the condition.
12826 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12827 to create a {1,2,3,...} vector.
12828 (vectorizable_reduction): Treat the number of units as polynomial.
12829 Choose vectype_in based on the largest scalar element size rather
12830 than the smallest number of units. Enforce the restrictions
12833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12834 Alan Hayward <alan.hayward@arm.com>
12835 David Sherwood <david.sherwood@arm.com>
12837 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12838 number of units as polynomial.
12840 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12841 Alan Hayward <alan.hayward@arm.com>
12842 David Sherwood <david.sherwood@arm.com>
12844 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12845 * target.def (autovectorize_vector_sizes): Return the vector sizes
12846 by pointer, using vector_sizes rather than a bitmask.
12847 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12848 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12849 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12851 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12852 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12853 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12854 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12855 * omp-general.c (omp_max_vf): Likewise.
12856 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12857 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12858 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12859 * tree-vect-slp.c (vect_slp_bb): Likewise.
12860 * doc/tm.texi: Regenerate.
12861 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12863 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12864 the vector size as a poly_uint64 rather than an unsigned int.
12865 (current_vector_size): Change from an unsigned int to a poly_uint64.
12866 (get_vectype_for_scalar_type): Update accordingly.
12867 * tree.h (build_truth_vector_type): Take the size and number of
12868 units as a poly_uint64 rather than an unsigned int.
12869 (build_vector_type): Add a temporary overload that takes
12870 the number of units as a poly_uint64 rather than an unsigned int.
12871 * tree.c (make_vector_type): Likewise.
12872 (build_truth_vector_type): Take the number of units as a poly_uint64
12873 rather than an unsigned int.
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 * target.def (get_mask_mode): Take the number of units and length
12880 as poly_uint64s rather than unsigned ints.
12881 * targhooks.h (default_get_mask_mode): Update accordingly.
12882 * targhooks.c (default_get_mask_mode): Likewise.
12883 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12884 * doc/tm.texi: Regenerate.
12886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12887 Alan Hayward <alan.hayward@arm.com>
12888 David Sherwood <david.sherwood@arm.com>
12890 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12891 * omp-general.c (omp_max_vf): Likewise.
12892 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12893 (expand_omp_simd): Handle polynomial safelen.
12894 * omp-low.c (omplow_simd_context): Add a default constructor.
12895 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12896 (lower_rec_simd_input_clauses): Update accordingly.
12897 (lower_rec_input_clauses): Likewise.
12899 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12900 Alan Hayward <alan.hayward@arm.com>
12901 David Sherwood <david.sherwood@arm.com>
12903 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12904 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12905 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12906 (vect_analyze_slp_cost): Likewise.
12907 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12908 (vect_model_load_cost): Likewise.
12910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12911 Alan Hayward <alan.hayward@arm.com>
12912 David Sherwood <david.sherwood@arm.com>
12914 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12915 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12916 from an unsigned int * to a poly_uint64_pod *.
12917 (calculate_unrolling_factor): New function.
12918 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12921 Alan Hayward <alan.hayward@arm.com>
12922 David Sherwood <david.sherwood@arm.com>
12924 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12925 from an unsigned int to a poly_uint64.
12926 (_loop_vec_info::slp_unrolling_factor): Likewise.
12927 (_loop_vec_info::vectorization_factor): Change from an int
12929 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12930 (vect_get_num_vectors): New function.
12931 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12932 (vect_get_num_copies): Use vect_get_num_vectors.
12933 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12934 to an unsigned int *.
12935 (vect_analyze_data_refs): Change min_vf from an int * to a
12937 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12938 than an unsigned HOST_WIDE_INT.
12939 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12940 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12941 to an unsigned int *.
12942 (vect_analyze_data_ref_dependences): Likewise.
12943 (vect_compute_data_ref_alignment): Handle polynomial vf.
12944 (vect_enhance_data_refs_alignment): Likewise.
12945 (vect_prune_runtime_alias_test_list): Likewise.
12946 (vect_shift_permute_load_chain): Likewise.
12947 (vect_supportable_dr_alignment): Likewise.
12948 (dependence_distance_ge_vf): Take the vectorization factor as a
12949 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12950 (vect_analyze_data_refs): Change min_vf from an int * to a
12952 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12953 vfm1 as a poly_uint64 rather than an int. Make the same change
12954 for the returned bound_scalar.
12955 (vect_gen_vector_loop_niters): Handle polynomial vf.
12956 (vect_do_peeling): Likewise. Update call to
12957 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12958 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12960 * tree-vect-loop.c (vect_determine_vectorization_factor)
12961 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12962 (vect_get_known_peeling_cost): Likewise.
12963 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12964 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12965 (vect_transform_loop): Likewise. Use the lowest possible VF when
12966 updating the upper bounds of the loop.
12967 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12968 rather than an int.
12969 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12970 polynomial unroll factors.
12971 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12972 (vect_make_slp_decision): Likewise.
12973 (vect_supported_load_permutation_p): Likewise, and polynomial
12975 (vect_analyze_slp_cost): Handle polynomial vf.
12976 (vect_slp_analyze_node_operations): Likewise.
12977 (vect_slp_analyze_bb_1): Likewise.
12978 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12979 than an unsigned HOST_WIDE_INT.
12980 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12981 (vectorizable_load): Handle polynomial vf.
12982 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12984 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12987 Alan Hayward <alan.hayward@arm.com>
12988 David Sherwood <david.sherwood@arm.com>
12990 * match.pd: Handle bit operations involving three constants
12991 and try to fold one pair.
12993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12995 * tree-vect-loop-manip.c: Include gimple-fold.h.
12996 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12997 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12998 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12999 Add a path that uses a step of VF instead of 1, but disable it
13001 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13002 and niters_no_overflow parameters. Update calls to
13003 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13004 Create a new SSA name if the latter choses to use a ste other
13005 than zero, and return it via niters_vector_mult_vf_var.
13006 * tree-vect-loop.c (vect_transform_loop): Update calls to
13007 vect_do_peeling, vect_gen_vector_loop_niters and
13008 slpeel_make_loop_iterate_ntimes.
13009 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13010 (vect_gen_vector_loop_niters): Update declarations after above changes.
13012 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13014 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13015 128-bit round to integer instructions.
13016 (ceil<mode>2): Likewise.
13017 (btrunc<mode>2): Likewise.
13018 (round<mode>2): Likewise.
13020 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13022 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13023 unaligned VSX load/store on P8/P9.
13024 (expand_block_clear): Allow the use of unaligned VSX
13025 load/store on P8/P9.
13027 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13029 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13031 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13032 swap associated with both a load and a store.
13034 2018-01-02 Andrew Waterman <andrew@sifive.com>
13036 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13037 * config/riscv/riscv.md (clear_cache): Use it.
13039 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13041 * web.c: Remove out-of-date comment.
13043 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13045 * expr.c (fixup_args_size_notes): Check that any existing
13046 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13047 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13048 (emit_single_push_insn): ...here.
13050 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13052 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13053 (const_vector_encoded_nelts): New function.
13054 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13055 (const_vector_int_elt, const_vector_elt): Declare.
13056 * emit-rtl.c (const_vector_int_elt_1): New function.
13057 (const_vector_elt): Likewise.
13058 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13059 of CONST_VECTOR_ELT.
13061 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13063 * expr.c: Include rtx-vector-builder.h.
13064 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13065 directly on the tree encoding.
13066 (const_vector_from_tree): Likewise.
13067 * optabs.c: Include rtx-vector-builder.h.
13068 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13069 sequence of "u" values.
13070 * vec-perm-indices.c: Include rtx-vector-builder.h.
13071 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13072 directly on the vec_perm_indices encoding.
13074 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13076 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13077 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13078 * rtx-vector-builder.h: New file.
13079 * rtx-vector-builder.c: Likewise.
13080 * rtl.h (rtx_def::u2): Add a const_vector field.
13081 (CONST_VECTOR_NPATTERNS): New macro.
13082 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13083 (CONST_VECTOR_DUPLICATE_P): Likewise.
13084 (CONST_VECTOR_STEPPED_P): Likewise.
13085 (CONST_VECTOR_ENCODED_ELT): Likewise.
13086 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13087 (unwrap_const_vec_duplicate): Likewise.
13088 (const_vec_series_p): Check for a non-duplicated vector encoding.
13089 Say that the function only returns true for integer vectors.
13090 * emit-rtl.c: Include rtx-vector-builder.h.
13091 (gen_const_vec_duplicate_1): Delete.
13092 (gen_const_vector): Call gen_const_vec_duplicate instead of
13093 gen_const_vec_duplicate_1.
13094 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13095 (gen_const_vec_duplicate): Use rtx_vector_builder.
13096 (gen_const_vec_series): Likewise.
13097 (gen_rtx_CONST_VECTOR): Likewise.
13098 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13099 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13100 Build a new vector rather than modifying a CONST_VECTOR in-place.
13101 (handle_special_swappables): Update call accordingly.
13102 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13103 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13104 Build a new vector rather than modifying a CONST_VECTOR in-place.
13105 (handle_special_swappables): Update call accordingly.
13107 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13109 * simplify-rtx.c (simplify_const_binary_operation): Use
13110 CONST_VECTOR_ELT instead of XVECEXP.
13112 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13114 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13115 the selector elements to be different from the data elements
13116 if the selector is a VECTOR_CST.
13117 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13118 ssizetype for the selector.
13120 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13122 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13123 before testing each element individually.
13124 * tree-vect-generic.c (lower_vec_perm): Likewise.
13126 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13128 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13129 * selftest-run-tests.c (selftest::run_tests): Call it.
13130 * vector-builder.h (vector_builder::operator ==): New function.
13131 (vector_builder::operator !=): Likewise.
13132 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13133 (vec_perm_indices::all_from_input_p): New function.
13134 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13135 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13136 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13137 instead of reading the VECTOR_CST directly. Detect whether both
13138 vector inputs are the same before constructing the vec_perm_indices,
13139 and update the number of inputs argument accordingly. Use the
13140 utility functions added above. Only construct sel2 if we need to.
13142 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13144 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13145 the broadcast of the low byte.
13146 (expand_mult_highpart): Use an explicit encoding for the permutes.
13147 * optabs-query.c (can_mult_highpart_p): Likewise.
13148 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13149 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13150 (vectorizable_bswap): Likewise.
13151 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13152 explicit encoding for the power-of-2 permutes.
13153 (vect_permute_store_chain): Likewise.
13154 (vect_grouped_load_supported): Likewise.
13155 (vect_permute_load_chain): Likewise.
13157 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13159 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13160 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13161 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13162 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13163 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13164 (vect_gen_perm_mask_any): Likewise.
13166 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13168 * int-vector-builder.h: New file.
13169 * vec-perm-indices.h: Include int-vector-builder.h.
13170 (vec_perm_indices): Redefine as an int_vector_builder.
13171 (auto_vec_perm_indices): Delete.
13172 (vec_perm_builder): Redefine as a stand-alone class.
13173 (vec_perm_indices::vec_perm_indices): New function.
13174 (vec_perm_indices::clamp): Likewise.
13175 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13176 (vec_perm_indices::new_vector): New function.
13177 (vec_perm_indices::new_expanded_vector): Update for new
13178 vec_perm_indices class.
13179 (vec_perm_indices::rotate_inputs): New function.
13180 (vec_perm_indices::all_in_range_p): Operate directly on the
13181 encoded form, without computing elided elements.
13182 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13183 encoding. Update for new vec_perm_indices class.
13184 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13185 the given vec_perm_builder.
13186 (expand_vec_perm_var): Update vec_perm_builder constructor.
13187 (expand_mult_highpart): Use vec_perm_builder instead of
13188 auto_vec_perm_indices.
13189 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13190 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13191 or double series encoding as appropriate.
13192 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13193 vec_perm_indices instead of auto_vec_perm_indices.
13194 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13195 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13196 (vect_permute_store_chain): Likewise.
13197 (vect_grouped_load_supported): Likewise.
13198 (vect_permute_load_chain): Likewise.
13199 (vect_shift_permute_load_chain): Likewise.
13200 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13201 (vect_transform_slp_perm_load): Likewise.
13202 (vect_schedule_slp_instance): Likewise.
13203 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13204 (vectorizable_mask_load_store): Likewise.
13205 (vectorizable_bswap): Likewise.
13206 (vectorizable_store): Likewise.
13207 (vectorizable_load): Likewise.
13208 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13209 vec_perm_indices instead of auto_vec_perm_indices. Use
13210 tree_to_vec_perm_builder to read the vector from a tree.
13211 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13212 vec_perm_builder instead of a vec_perm_indices.
13213 (have_whole_vector_shift): Use vec_perm_builder and
13214 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13215 truncation to calc_vec_perm_mask_for_shift.
13216 (vect_create_epilog_for_reduction): Likewise.
13217 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13218 from auto_vec_perm_indices to vec_perm_indices.
13219 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13220 instead of changing individual elements.
13221 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13222 the vector in d.perm.
13223 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13224 from auto_vec_perm_indices to vec_perm_indices.
13225 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13226 instead of changing individual elements.
13227 (arm_vectorize_vec_perm_const): Use new_vector to install
13228 the vector in d.perm.
13229 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13230 Update vec_perm_builder constructor.
13231 (rs6000_expand_interleave): Likewise.
13232 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13233 (rs6000_expand_interleave): Likewise.
13235 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13237 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13238 to qimode could truncate the indices.
13239 * optabs.c (expand_vec_perm_var): Likewise.
13241 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13243 * Makefile.in (OBJS): Add vec-perm-indices.o.
13244 * vec-perm-indices.h: New file.
13245 * vec-perm-indices.c: Likewise.
13246 * target.h (vec_perm_indices): Replace with a forward class
13248 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13249 * optabs.h: Include vec-perm-indices.h.
13250 (expand_vec_perm): Delete.
13251 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13252 (expand_vec_perm_const): Declare.
13253 * target.def (vec_perm_const_ok): Replace with...
13254 (vec_perm_const): ...this new hook.
13255 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13256 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13257 * doc/tm.texi: Regenerate.
13258 * optabs.def (vec_perm_const): Delete.
13259 * doc/md.texi (vec_perm_const): Likewise.
13260 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13261 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13262 expand_vec_perm for constant permutation vectors. Assert that
13263 the mode of variable permutation vectors is the integer equivalent
13264 of the mode that is being permuted.
13265 * optabs-query.h (selector_fits_mode_p): Declare.
13266 * optabs-query.c: Include vec-perm-indices.h.
13267 (selector_fits_mode_p): New function.
13268 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13269 is defined, instead of checking whether the vec_perm_const_optab
13270 exists. Use targetm.vectorize.vec_perm_const instead of
13271 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13272 fit in the vector mode before using a variable permute.
13273 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13274 vec_perm_indices instead of an rtx.
13275 (expand_vec_perm): Replace with...
13276 (expand_vec_perm_const): ...this new function. Take the selector
13277 as a vec_perm_indices rather than an rtx. Also take the mode of
13278 the selector. Update call to shift_amt_for_vec_perm_mask.
13279 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13280 Use vec_perm_indices::new_expanded_vector to expand the original
13281 selector into bytes. Check whether the indices fit in the vector
13282 mode before using a variable permute.
13283 (expand_vec_perm_var): Make global.
13284 (expand_mult_highpart): Use expand_vec_perm_const.
13285 * fold-const.c: Includes vec-perm-indices.h.
13286 * tree-ssa-forwprop.c: Likewise.
13287 * tree-vect-data-refs.c: Likewise.
13288 * tree-vect-generic.c: Likewise.
13289 * tree-vect-loop.c: Likewise.
13290 * tree-vect-slp.c: Likewise.
13291 * tree-vect-stmts.c: Likewise.
13292 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13294 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13295 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13296 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13297 (aarch64_vectorize_vec_perm_const): ...this new function.
13298 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13299 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13300 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13301 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13302 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13303 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13304 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13306 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13307 check for NEON modes.
13308 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13309 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13310 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13311 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13313 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13314 the old VEC_PERM_CONST conditions.
13315 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13316 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13317 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13318 (ia64_vectorize_vec_perm_const_ok): Merge into...
13319 (ia64_vectorize_vec_perm_const): ...this new function.
13320 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13321 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13322 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13323 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13324 * config/mips/mips.c (mips_expand_vec_perm_const)
13325 (mips_vectorize_vec_perm_const_ok): Merge into...
13326 (mips_vectorize_vec_perm_const): ...this new function.
13327 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13328 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13329 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13330 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13331 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13332 (rs6000_expand_vec_perm_const): Delete.
13333 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13335 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13336 (altivec_expand_vec_perm_const_le): Take each operand individually.
13337 Operate on constant selectors rather than rtxes.
13338 (altivec_expand_vec_perm_const): Likewise. Update call to
13339 altivec_expand_vec_perm_const_le.
13340 (rs6000_expand_vec_perm_const): Delete.
13341 (rs6000_vectorize_vec_perm_const_ok): Delete.
13342 (rs6000_vectorize_vec_perm_const): New function.
13343 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13344 an element count and rtx array.
13345 (rs6000_expand_extract_even): Update call accordingly.
13346 (rs6000_expand_interleave): Likewise.
13347 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13348 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13349 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13350 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13351 (rs6000_expand_vec_perm_const): Delete.
13352 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13353 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13354 (altivec_expand_vec_perm_const_le): Take each operand individually.
13355 Operate on constant selectors rather than rtxes.
13356 (altivec_expand_vec_perm_const): Likewise. Update call to
13357 altivec_expand_vec_perm_const_le.
13358 (rs6000_expand_vec_perm_const): Delete.
13359 (rs6000_vectorize_vec_perm_const_ok): Delete.
13360 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13361 reference to the SPE evmerge intructions.
13362 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13363 an element count and rtx array.
13364 (rs6000_expand_extract_even): Update call accordingly.
13365 (rs6000_expand_interleave): Likewise.
13366 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13367 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13369 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13371 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13373 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13374 vector mode and that that mode matches the mode of the data
13376 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13377 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13378 directly using expand_vec_perm_1 when forcing selectors into
13380 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13382 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13384 * optabs-query.h (can_vec_perm_p): Delete.
13385 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13386 * optabs-query.c (can_vec_perm_p): Split into...
13387 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13388 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13389 particular selector is valid.
13390 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13391 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13392 (vect_grouped_load_supported): Likewise.
13393 (vect_shift_permute_load_chain): Likewise.
13394 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13395 (vect_transform_slp_perm_load): Likewise.
13396 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13397 (vectorizable_bswap): Likewise.
13398 (vect_gen_perm_mask_checked): Likewise.
13399 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13400 implementations of variable permutation vectors into account
13401 when deciding which selector to use.
13402 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13403 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13404 with a false third argument.
13405 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13406 to test whether the constant selector is valid and can_vec_perm_var_p
13407 to test whether a variable selector is valid.
13409 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13411 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13412 * optabs-query.c (can_vec_perm_p): Likewise.
13413 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13414 instead of vec_perm_indices.
13415 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13416 (vect_gen_perm_mask_checked): Likewise,
13417 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13418 (vect_gen_perm_mask_checked): Likewise,
13420 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13422 * optabs-query.h (qimode_for_vec_perm): Declare.
13423 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13424 (qimode_for_vec_perm): ...this new function.
13425 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13427 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13429 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13430 does not have a conditional at the top.
13432 2018-01-02 Richard Biener <rguenther@suse.de>
13434 * ipa-inline.c (big_speedup_p): Fix expression.
13436 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13439 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13442 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13446 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13447 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13448 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13449 cond_taken_branch_cost 3->4.
13451 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13453 PR tree-optimization/83581
13454 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13455 TODO_cleanup_cfg if any changes have been made.
13457 PR middle-end/83608
13458 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13459 convert_modes if target mode has the right side, but different mode
13462 PR middle-end/83609
13463 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13464 last argument when extracting from CONCAT. If either from_real or
13465 from_imag is NULL, use expansion through memory. If result is not
13466 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13467 the parts directly to inner mode, if even that fails, use expansion
13470 PR middle-end/83623
13471 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13472 check for bswap in mode rather than HImode and use that in expand_unop
13475 Copyright (C) 2018 Free Software Foundation, Inc.
13477 Copying and distribution of this file, with or without modification,
13478 are permitted in any medium without royalty provided the copyright
13479 notice and this notice are preserved.