1 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
6 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7 Chung-Ju Wu <jasonwucj@gmail.com>
9 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
10 * config/nds32/nds32-md-auxiliary.c
11 (symbolic_reference_mentioned_p): New.
12 (nds32_legitimize_ict_address): New.
13 (nds32_expand_ict_move): New.
14 (nds32_indirect_call_referenced_p): New.
15 (nds32_symbol_binds_local_p): Delete.
16 (nds32_long_call_p): Modify.
17 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
18 * config/nds32/nds32-protos.h
19 (symbolic_reference_mentioned_p): Declare.
20 (nds32_legitimize_ict_address): Declare.
21 (nds32_expand_ict_move): Declare.
22 (nds32_indirect_call_referenced_p): Declare.
23 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
24 (nds32_relax_group): Use nds32_ict_const_p as condition.
25 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
26 (nds32_asm_file_start): Output ict_model directive in asm code.
27 (nds32_legitimate_address_p): Consider indirect call.
28 (nds32_print_operand): Consider indirect call.
29 (nds32_print_operand_address): Consider indirect call.
30 (nds32_insert_attributes): Handle "indirect_call" attribute.
31 (TARGET_LEGITIMATE_ADDRESS_P): Define.
32 (TARGET_LEGITIMATE_CONSTANT_P): Define.
33 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
34 (TARGET_DELEGITIMIZE_ADDRESS): Define.
35 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
36 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
37 (TARGET_ICT_MODEL_SMALL): Define.
38 (TARGET_ICT_MODEL_LARGE): Define.
39 * config/nds32/nds32.md (movsi): Consider ict model.
40 (call, call_value): Consider ict model.
41 (sibcall, sibcall_value): Consider ict model.
42 * config/nds32/nds32.opt (mict-model): New option.
43 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
46 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
47 Monk Chiang <sh.chiang04@gmail.com>
48 Jim Wilson <jimw@sifive.com>
50 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
51 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
52 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
53 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
54 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
55 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
56 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
57 compute save_libcall_adjustment properly.
58 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
59 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
60 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
61 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
62 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
63 (ABI_SPEC): Handle mabi=ilp32e.
64 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
66 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
67 <-march>: Add rv32e as an example.
69 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
72 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
73 (intra_create_variable_infos): Handle C++ constructors.
75 2018-05-18 Martin Liska <mliska@suse.cz>
77 * passes.def: Remove a redundant pass.
79 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
82 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
84 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
86 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
88 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
90 (armv2, armv2a, armv3, armv3m): Delete architectures.
91 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
92 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
93 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
95 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
96 (*mulsidi3adddi): Likewise.
98 (*mulsidi3_nov6): Likewise.
99 (umulsidi3): Likewise.
100 (umulsidi3_nov6): Likewise.
101 (umaddsidi4): Likewise.
102 (*umulsidi3adddi): Likewise.
103 (smulsi3_highpart): Likewise.
104 (*smulsi3_highpart_nov6): Likewise.
105 (umulsi3_highpart): Likewise.
106 (*umulsi3_highpart_nov6): Likewise.
107 * config/arm/arm.h (arm_arch3m): Delete.
108 * config/arm/arm.c (arm_arch3m): Delete.
109 (arm_option_override_internal): Update armv3-related comment.
110 (arm_configure_build_target): Delete use of isa_bit_mode32.
111 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
112 (arm_rtx_costs_internal): Delete check of arm_arch3m.
113 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
116 * config/arm/arm-protos.h (arm_arch3m): Delete.
117 * config/arm/arm-tables.opt: Regenerate.
118 * config/arm/arm-tune.md: Likewise.
119 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
120 deleted architectures.
122 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
124 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
125 (armv5t, armv5te): New features.
126 (ARMv5, ARMv5e): Delete fgroups.
127 (ARMv5t, ARMv5te): Adjust for above changes.
129 (armv5, armv5e): Delete arches.
130 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
132 (*call_reg_arm): Likewise.
133 (*call_value_reg_armv5): Likewise.
134 (*call_value_reg_arm): Likewise.
135 (*call_symbol): Likewise.
136 (*call_value_symbol): Likewise.
137 (*sibcall_insn): Likewise.
138 (*sibcall_value_insn): Likewise.
140 (prefetch): Likewise.
141 (define_split and define_peephole2 dependent on arm_arch5):
143 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
145 (TARGET_ARM_QBIT): Likewise.
146 (TARGET_DSP_MULTIPLY): Likewise.
147 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
148 (arm_arch5, arm_arch5e): Delete.
149 (arm_arch5t, arm_arch5te): Declare.
150 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
151 (arm_arch5t): Declare.
152 (arm_option_reconfigure_globals): Update for the above.
153 (arm_options_perform_arch_sanity_checks): Update comment, replace
154 use of arm_arch5 with arm_arch5t.
155 (use_return_insn): Likewise.
156 (arm_emit_call_insn): Likewise.
157 (output_return_instruction): Likewise.
158 (arm_final_prescan_insn): Likewise.
159 (arm_coproc_builtin_available): Likewise.
160 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
161 arm_arch5e with arm_arch5t and arm_arch5te.
162 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
163 (arm_arch5t, arm_arch5te): Declare.
164 * config/arm/arm-tables.opt: Regenerate.
165 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
166 * config/arm/t-multilib: Likewise.
167 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
168 instead of arm_arch5.
169 (*call_reg_thumb1): Likewise.
170 (*call_value_reg_thumb1_v5): Likewise.
171 (*call_value_reg_thumb1): Likewise.
172 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
174 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
176 2018-05-18 Martin Liska <mliska@suse.cz>
178 PR gcov-profile/84846
179 * doc/gcov.texi: Document -t option of gcov tool.
181 2018-05-18 Martin Liska <mliska@suse.cz>
183 PR gcov-profile/84846
184 * gcov.c (print_usage): Add new -t option.
185 (process_args): Handle the option.
186 (generate_results): Use stdout as output when requested by
189 2018-05-18 Martin Liska <mliska@suse.cz>
191 PR gcov-profile/84846
192 * coverage.c (coverage_init): Write PWD to .gcno file.
193 * doc/gcov.texi: Document how working directory is printed.
194 * gcov-dump.c (dump_gcov_file): Print PWD.
195 * gcov.c (output_intermediate_file): Likewise.
196 (read_graph_file): Read PWD string.
197 (output_lines): Print PWD.
199 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
202 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
203 for retval and return false if all args to phi are zero.
205 2018-05-18 Richard Biener <rguenther@suse.de>
207 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
209 (evrp_dom_walker::before_dom_children): Call it.
211 2018-05-18 Richard Biener <rguenther@suse.de>
213 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
214 results when processing array refs with variable index.
216 2018-05-18 Toon Moene <toon@moene.org>
218 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
219 directly after that of -floop-interchange. Indicate that both
220 options are enabled by default when specifying -O3.
222 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
224 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
225 iterator. Delete separate integer-mode vec_set<mode> expander.
226 (aarch64_simd_vec_setv2di): Delete.
227 (vec_setv2di): Delete.
228 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
229 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
230 the "w, r" alternative.
232 2018-05-18 Martin Liska <mliska@suse.cz>
234 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
235 * tree-pass.h (make_pass_lower_switch_O0): New function.
236 * tree-switch-conversion.c (node_has_low_bound): Remove.
237 (node_has_high_bound): Likewise.
238 (node_is_bounded): Likewise.
239 (class pass_lower_switch): Make it a template type and create
241 (pass_lower_switch::execute): Add template argument.
242 (make_pass_lower_switch): New function.
243 (make_pass_lower_switch_O0): New function.
244 (do_jump_if_equal): Remove.
245 (emit_case_nodes): Simplify to just handle all 3 cases and leave
246 all the hard work to tree optimization passes.
248 2018-05-18 Martin Liska <mliska@suse.cz>
250 * dbgcnt.c (limit_low): Renamed from limit.
251 (limit_high): New variable.
252 (dbg_cnt_is_enabled): Check for upper limit.
253 (dbg_cnt): Adjust dumping.
254 (dbg_cnt_set_limit_by_index): Add new argument for high
256 (dbg_cnt_set_limit_by_name): Likewise.
257 (dbg_cnt_process_single_pair): Parse new format.
258 (dbg_cnt_process_opt): Use strtok.
259 (dbg_cnt_list_all_counters): Remove 'value' and add
261 * doc/invoke.texi: Document changes.
263 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
265 * doc/sourcebuild.texi (scalar_all_fma): Document.
266 * tree.def (FMA_EXPR): Delete.
267 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
268 * internal-fn.c (ternary_direct): New macro.
269 (expand_ternary_optab_fn): Likewise.
270 (direct_ternary_optab_supported_p): Likewise.
271 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
272 * builtins.c (fold_builtin_fma): Delete.
273 (fold_builtin_3): Don't call it.
274 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
275 * expr.c (expand_expr_real_2): Likewise.
276 * fold-const.c (operand_equal_p): Likewise.
277 (fold_ternary_loc): Likewise.
278 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
279 * gimple.c (DEFTREECODE): Likewise.
280 * gimplify.c (gimplify_expr): Likewise.
281 * optabs-tree.c (optab_for_tree_code): Likewise.
282 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
283 * tree-eh.c (operation_could_trap_p): Likewise.
284 (stmt_could_throw_1_p): Likewise.
285 * tree-inline.c (estimate_operator_cost): Likewise.
286 * tree-pretty-print.c (dump_generic_node): Likewise.
287 (op_code_prio): Likewise.
288 * tree-ssa-loop-im.c (stmt_cost): Likewise.
289 * tree-ssa-operands.c (get_expr_operands): Likewise.
290 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
291 * fold-const-call.h (fold_fma): Delete.
292 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
293 CFN_FNMA and CFN_FNMS.
295 * genmatch.c (combined_fn): New enum.
296 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
297 (commutative_op): New function.
298 (commutate): Use it. Handle more than 2 operands.
299 (dt_operand::gen_gimple_expr): Use commutative_op.
300 (parser::parse_expr): Allow :c to be used with non-binary
301 operators if the commutative operand is known.
302 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
303 CFN_FMS, CFN_FNMA and CFN_FNMS.
304 (backprop::process_assign_use): Remove FMA_EXPR handling.
305 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
306 (gen_hsa_fma): New function.
307 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
308 IFN_FNMA and IFN_FNMS.
309 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
310 * gimple-fold.h (follow_all_ssa_edges): Declare.
311 * gimple-fold.c (follow_all_ssa_edges): New function.
312 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
313 gimple_build interface and use follow_all_ssa_edges to fold the result.
314 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
315 instead of checking for optabs directly.
316 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
317 rather than FMA_EXPRs.
318 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
319 call to IFN_FMA instead of an FMA_EXPR.
321 2018-05-17 Jim Wilson <jimw@sifive.com>
323 * expr.c (do_tablejump): When converting index to Pmode, if we have a
324 sign extended promoted subreg, and the range does not have the sign bit
325 set, then do a sign extend.
327 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
328 test, check for sign extended subreg and/or constant operands, and
329 do a sign extend in that case.
331 2018-05-17 Steve Ellcey <sellcey@cavium.com>
333 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
334 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
336 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
337 Change logics_shift_reg to logics_shift_imm.
338 (thunderx2t99_fp_loadpair_basic): Delete.
339 (thunderx2t99_fp_storepair_basic): Delete.
340 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
341 (thunderx2t99_asimd_polynomial): Delete.
342 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
343 and neon_fp_mul_d_scalar_q.
344 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
345 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
346 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
347 (thunderx2t99_asimd_lut): Add missing tbl types.
348 (thunderx2t99_asimd_ext): Delete.
349 (thunderx2t99_asimd_load1_1_mult): Delete.
350 (thunderx2t99_asimd_load1_2_mult): Delete.
351 (thunderx2t99_asimd_load1_ldp): New.
352 (thunderx2t99_asimd_load1): New.
353 (thunderx2t99_asimd_load2): Add missing *load2* types.
354 (thunderx2t99_asimd_load3): New.
355 (thunderx2t99_asimd_load4): New.
356 (thunderx2t99_asimd_store1_1_mult): Delete.
357 (thunderx2t99_asimd_store1_2_mult): Delete.
358 (thunderx2t99_asimd_store2_mult): Delete.
359 (thunderx2t99_asimd_store2_onelane): Delete.
360 (thunderx2t99_asimd_store_stp): New.
361 (thunderx2t99_asimd_store1): New.
362 (thunderx2t99_asimd_store2): New.
363 (thunderx2t99_asimd_store3): New.
364 (thunderx2t99_asimd_store4): New.
366 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
368 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
369 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
371 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
372 Segher Boessenkool <segher@kernel.crashing.org>
375 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
378 2018-05-17 Richard Biener <rguenther@suse.de>
380 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
381 for pruning loop and prune defs feeding only already visited PHIs.
383 2018-05-17 Richard Biener <rguenther@suse.de>
385 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
387 2018-05-17 Bin Cheng <bin.cheng@arm.com>
388 Richard Biener <rguenther@suse.de>
390 PR tree-optimization/85793
391 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
392 for VMAT_ELEMENTWISE.
394 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
396 * internal-fn.h (lookup_internal_fn): Declare
397 * internal-fn.c (lookup_internal_fn): New function.
398 * gimple.c (gimple_build_call_from_tree): Handle calls to
400 * gimple-pretty-print.c (dump_gimple_call): Print "." before
401 internal function names.
402 * tree-pretty-print.c (dump_generic_node): Likewise.
403 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
405 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
407 * gimple-fold.h (gimple_build): Make the function forms take
408 combined_fn rather than built_in_function.
409 (gimple_simplify): Likewise.
410 * gimple-match-head.c (gimple_simplify): Likewise.
411 * gimple-fold.c (gimple_build): Likewise.
412 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
413 rather than gimple_build_call_internal.
414 (get_initial_defs_for_reduction): Likewise.
415 (vect_create_epilog_for_reduction): Likewise.
416 (vectorizable_live_operation): Likewise.
418 2018-05-17 Martin Liska <mliska@suse.cz>
420 * gimple-ssa-sprintf.c (format_directive): Do not use
421 space in between 'G_' and '('.
423 2018-05-17 Jakub Jelinek <jakub@redhat.com>
426 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
427 even if the mask is not all ones.
430 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
432 (ix86_gimple_fold_builtin): Likewise.
435 * config/i386/i386.c: Include tree-vector-builder.h.
436 (ix86_vector_shift_count): New function.
437 (ix86_fold_builtin): Fold shift builtins by scalar count.
438 (ix86_gimple_fold_builtin): Likewise.
440 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
441 _mm512_setzero): New intrinsics.
443 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
444 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
446 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
447 code generation for cases where splatting a value is not useful.
448 * simplify-rtx.c (simplify_ternary_operation): Simplify
449 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
450 mode to a vec_concat.
452 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
454 * config.gcc: Support "goldmont-plus".
455 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
457 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
458 PROCESSOR_GOLDMONT_PLUS.
459 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
460 (processor_target_table): Add "goldmont-plus".
461 (PTA_GOLDMONT_PLUS): Define.
462 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
463 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
464 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
465 (fold_builtin_cpu): Add "goldmont-plus".
466 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
467 (ix86_option_override_internal): Add "goldmont-plus".
468 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
469 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
470 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
471 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
473 2018-05-17 Richard Biener <rguenther@suse.de>
475 PR tree-optimization/85757
476 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
477 remove defs that only feed that PHI from further processing.
479 2018-05-16 Jim Wilson <jimw@sifive.com>
481 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
483 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
485 2018-05-16 Mark Wielaard <mark@klomp.org>
487 * dwarf2out.c (count_index_strings): New function.
488 (output_indirect_strings): Call count_index_strings and generate
489 header for dwarf_version >= 5.
491 2018-05-16 Mark Wielaard <mark@klomp.org>
493 * dwarf2out.c (dwarf_FORM): New function.
494 (set_indirect_string): Use dwarf_FORM.
495 (reset_indirect_string): Likewise.
496 (size_of_die): Likewise.
497 (value_format): Likewise.
498 (output_die): Likewise.
499 (add_skeleton_AT_string): Likewise.
500 (output_macinfo_op): Likewise.
501 (index_string): Likewise.
502 (output_index_string_offset): Likewise.
503 (output_index_string): Likewise.
504 (count_index_strings): Likewise.
506 2018-05-16 Carl Love <cel@us.ibm.com>
508 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
509 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
511 2018-05-16 Martin Jambor <mjambor@suse.cz>
513 * ipa-prop.c (ipa_free_all_edge_args): Remove.
514 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
516 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
518 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
519 (fnma<mode>4): Likewise.
520 (fms<mode>4): Likewise.
521 (fnms<mode>4): Likewise.
522 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
523 (aarch64_fnma<mode>4): Likewise.
524 (aarch64_fms<mode>4): Likewise.
525 (aarch64_fnms<mode>4): Likewise.
526 (aarch64_fnmadd<mode>4): Likewise.
528 2018-05-16 Jason Merrill <jason@redhat.com>
530 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
532 2018-05-16 Richard Biener <rguenther@suse.de>
534 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
535 (dump_stmt_cost): Declare.
536 (add_stmt_cost): Dump cost we add.
537 (add_stmt_costs): New function.
538 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
540 (vect_analyze_stmt): Adjust prototype.
541 (vectorizable_condition): Likewise.
542 (vectorizable_live_operation): Likewise.
543 (vectorizable_reduction): Likewise.
544 (vectorizable_induction): Likewise.
545 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
546 cost vector to pass to vectorizable_ and record afterwards.
547 (vect_model_reduction_cost): Take cost vector argument and adjust.
548 (vect_model_induction_cost): Likewise.
549 (vectorizable_reduction): Likewise.
550 (vectorizable_induction): Likewise.
551 (vectorizable_live_operation): Likewise.
552 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
553 SLP_TREE_NUMBER_OF_VEC_STMTS.
554 (vect_analyze_slp_cost_1): Remove.
555 (vect_analyze_slp_cost): Likewise.
556 (vect_slp_analyze_node_operations): Take visited args and
557 a target cost vector. Avoid processing already visited stmt sets.
558 (vect_slp_analyze_operations): Use a local cost vector to gather
559 costs and register those of non-discarded instances.
560 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
561 (vect_schedule_slp_instance): Remove copying of
562 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
564 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
565 adding cost. Record cost entry location.
566 (vect_prologue_cost_for_slp_op): Function to compute cost of
567 a constant or invariant generated for SLP vect in the prologue,
568 split out from vect_analyze_slp_cost_1.
569 (vect_model_simple_cost): Make static. Adjust for SLP costing.
570 (vect_model_promotion_demotion_cost): Likewise.
571 (vect_model_store_cost): Likewise, make static.
572 (vect_model_load_cost): Likewise.
573 (vectorizable_bswap): Add cost vector arg and adjust.
574 (vectorizable_call): Likewise.
575 (vectorizable_simd_clone_call): Likewise.
576 (vectorizable_conversion): Likewise.
577 (vectorizable_assignment): Likewise.
578 (vectorizable_shift): Likewise.
579 (vectorizable_operation): Likewise.
580 (vectorizable_store): Likewise.
581 (vectorizable_load): Likewise.
582 (vectorizable_condition): Likewise.
583 (vectorizable_comparison): Likewise.
584 (can_vectorize_live_stmts): Likewise.
585 (vect_analyze_stmt): Likewise.
586 (vect_transform_stmt): Adjust calls to vectorizable_*.
587 * tree-vectorizer.c: Include gimple-pretty-print.h.
588 (dump_stmt_cost): New function.
590 2018-05-16 Richard Biener <rguenther@suse.de>
592 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
593 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
594 * tree-ssa-dse.c: Include tree-ssa-loop.h.
595 (check_name): New callback.
596 (dse_classify_store): Track cycles via a visited bitmap of PHI
597 defs and simplify handling of in-loop and across loop dead stores
598 and properly fail for loop-variant refs. Handle byte-tracking with
599 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
602 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
604 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
605 (vect_get_mask_type_for_stmt): Likewise.
606 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
608 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
609 to determine the statement's vector type and the vector type that
610 should be used for calculating nunits. Deal with cases in which
611 the type has to be deferred.
612 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
613 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
614 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
615 (vect_determine_vf_for_stmt): New functions, split out from...
616 (vect_determine_vectorization_factor): ...here.
617 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
618 (vect_get_mask_type_for_stmt): New functions, split out from
619 vect_determine_vectorization_factor.
621 2018-05-16 Richard Biener <rguenther@suse.de>
623 * tree-cfg.c (verify_gimple_assign_ternary): Properly
624 verify the [VEC_]COND_EXPR embedded comparison.
626 2018-05-15 Martin Sebor <msebor@redhat.com>
628 PR tree-optimization/85753
629 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
630 RECORD_TYPE in addition to ARRAY_TYPE.
632 2018-05-15 Martin Sebor <msebor@redhat.com>
635 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
637 2018-05-15 Richard Biener <rguenther@suse.de>
639 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
640 add by_clobber_p one. Change algorithm to collect all defs
641 representing uses we need to walk and try reducing them to
642 a single one before failing.
643 (dse_dom_walker::dse_optimize_stmt): Adjust.
645 2018-05-13 Mark Wielaard <mark@klomp.org>
647 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
648 (size_of_loc_descr): Likewise.
649 (output_loc_operands): Likewise.
650 (output_loc_operands_raw): Likewise.
651 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
652 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
653 (hash_loc_operands): Likewise.
654 (compare_loc_operands): Likewise.
656 2018-05-14 Mark Wielaard <mark@klomp.org>
658 * dwarf2out.c (count_index_addrs): New function.
659 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
661 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
663 PR tree-optimization/83648
664 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
665 return value as malloc candidate.
667 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
670 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
671 as true in call to suggest_attribute.
673 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
675 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
678 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
680 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
682 (aarch64_crypto_aesd_fused): Likewise.
684 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
686 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
687 (movsi_aarch64): Likewise.
688 (load_pairsi): Likewise.
689 (load_pairdi): Likewise.
690 (store_pairsi): Likewise.
691 (store_pairdi): Likewise.
692 (load_pairsf): Likewise.
693 (load_pairdf): Likewise.
694 (store_pairsf): Likewise.
695 (store_pairdf): Likewise.
696 (zero_extend): Likewise.
697 (trunc): Swap alternatives.
698 (fcvt_target): Add '?' to prefer w over r.
700 2018-05-14 Jakub Jelinek <jakub@redhat.com>
703 * config/i386/i386.md: Disallow non-commutative arithmetics in
704 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
705 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
706 in the peephole2 before it.
708 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
710 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
711 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
712 (ix86_handle_option): Handle -mcldemote.
713 * config.gcc: New header.
714 * config/i386/cldemoteintrin.h: New file.
715 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
716 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
718 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
719 OPTION_MASK_ISA_CLDEMOTE.
720 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
721 (ix86_valid_target_attribute_inner_p): Ditto.
722 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
723 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
724 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
725 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
726 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
728 * config/i386/i386.opt: Add -mcldemote.
729 * config/i386/x86intrin.h: New header.
730 * doc/invoke.texi: Add -mcldemote.
732 2018-05-14 Richard Biener <rguenther@suse.de>
734 * doc/match-and-simplify.texi: Adjust :s documentation.
736 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
738 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
739 intended memcpy size.
740 (REORDER_45): Likewise.
742 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
745 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
746 * vec.c (qsort_chk): Use gcc_qsort.
747 * Makefile.in (OBJS-libcommon): Add sort.o.
748 (build/sort.o): New target. Use it...
749 (BUILD_RTL): ... here, and...
750 (build/gencfn-macros): ... here, and...
751 (build/genmatch): ... here.
753 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
754 Chung-Ju Wu <jasonwucj@gmail.com>
756 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
757 * config/nds32/nds32-graywolf.md: New file.
758 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
759 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
761 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
762 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
763 * config/nds32/nds32.md (pipeline_model): Add graywolf.
764 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
765 * config/nds32/pipelines.md: Include n15 settings.
767 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
768 Chung-Ju Wu <jasonwucj@gmail.com>
770 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
771 * config/nds32/nds32-n13.md: New file.
772 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
773 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
775 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
776 * config/nds32/nds32.md (pipeline_model): Add n13.
777 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
778 * config/nds32/pipelines.md: Include n13 settings.
780 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
781 Chung-Ju Wu <jasonwucj@gmail.com>
783 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
784 * config/nds32/nds32-n10.md: New file.
785 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
786 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
788 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
789 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
790 * config/nds32/nds32.md (pipeline_model): Add n10.
791 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
792 * config/nds32/pipelines.md: Include n10 settings.
794 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
795 Kito Cheng <kito.cheng@gmail.com>
796 Chung-Ju Wu <jasonwucj@gmail.com>
798 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
799 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
800 Add enum values for DSP extension instructions.
801 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
803 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
804 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
806 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
807 * config/nds32/nds32-dspext.md: New file for DSP implementation.
808 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
809 * config/nds32/nds32-intrinsic.md: Likewise.
810 * config/nds32/nds32_intrinsic.h: Likewise.
811 * config/nds32/nds32-md-auxiliary.c: Likewise.
812 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
813 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
814 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
815 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
816 * config/nds32/nds32-protos.h: New declarations for DSP extension.
817 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
818 TYPE_DMAC in switch statement.
819 * config/nds32/nds32.c: New checking and implementation for DSP
820 extension instructions.
821 * config/nds32/nds32.h: Likewise.
822 * config/nds32/nds32.md: Likewise.
823 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
824 * config/nds32/predicates.md: Implement new predicates for DSP
827 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
829 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
830 Reformat alternatives and attributes so it is easier to identify
831 which constraints/attributes go with which instruction.
832 (mov<mode>_hardfloat32, FMOVE64): Likewise.
833 (mov<mode>_softfloat32, FMOVE64): Likewise.
834 (mov<mode>_hardfloat64, FMOVE64): Likewise.
835 (mov<mode>_softfloat64, FMOVE64): Likewise.
837 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
839 * doc/extend.texi (PowerPC Built-in Functions): Rename this
841 (Basic PowerPC Built-in Functions): The new name of the
842 subsection previously known as "PowerPC Built-in Functions".
843 (Basic PowerPC Built-in Functions Available on all Configurations):
845 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
846 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
847 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
848 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
850 2018-05-11 Martin Jambor <mjambor@suse.cz>
853 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
856 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
859 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
861 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
863 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
864 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
865 (ix86_handle_option): Handle -mwaitpkg.
866 * config.gcc: New header.
867 * config/i386/cpuid.h (bit_WAITPKG): New bit.
868 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
869 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
871 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
872 OPTION_MASK_ISA_WAITPKG.
873 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
874 (ix86_option_override_internal): Add PTA_WAITPKG.
875 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
876 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
878 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
879 __builtin_ia32_umwait and __builtin_ia32_tpause.
880 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
881 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
882 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
883 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
884 UNSPECV_TPAUSE): New.
885 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
886 * config/i386/i386.opt: Add -mwaitpkg.
887 * config/i386/waitpkgintrin.h: New file.
888 * config/i386/x86intrin.h: New header.
889 * doc/invoke.texi: Add -mwaitpkg.
891 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
894 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
896 (cortex-m0): Use armv6s-m isa.
897 (cortex-m0plus): Likewise.
898 (cortex-m1): Likewise.
899 (cortex-m0.small-multiply): Likewise.
900 (cortex-m0plus.small-multiply): Likewise.
901 (cortex-m1.small-multiply): Likewise.
903 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
904 Jakub Jelinek <jakub@redhat.com>
906 PR tree-optimization/85692
907 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
908 source permute as well.
910 2018-05-11 Martin Liska <mliska@suse.cz>
913 * doc/extend.texi: Document LLVM style format for no_sanitize
916 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
918 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
919 mode_supports_vsx_dform_quad to mode_supports_dq_form.
920 (mode_supports_vsx_dform_quad): Likewise.
921 (mode_supports_vmx_dform): Move these functions to be next to the
922 other mode_supports functions.
923 (mode_supports_dq_form): Likewise.
924 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
925 mode_supports_dq_form.
926 (reg_offset_addressing_ok_p): Likewise.
927 (offsettable_ok_by_alignment): Likewise.
928 (rs6000_legitimate_offset_address_p): Likewise.
929 (legitimate_lo_sum_address_p): Likewise.
930 (rs6000_legitimize_address): Likewise.
931 (rs6000_legitimize_reload_address): Likewise.
932 (rs6000_secondary_reload_inner): Likewise.
933 (rs6000_preferred_reload_class): Likewise.
934 (rs6000_output_move_128bit): Likewise.
936 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
938 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
939 Generate SImode target register for null target.
940 <case IX86_BUILTIN_XGETBV>: Ditto.
941 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
942 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
944 2018-05-10 Carl Love <cel@us.ibm.com>
946 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
947 dcbtt and dcbtstt if operands[2] is 0.
949 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
952 * config/i386/sse.md (usadv64qi): New expander.
954 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
956 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
957 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
958 -maltivec=be support.
959 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
960 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
961 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
962 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
963 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
964 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
965 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
966 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
967 altivec_vsumsws): Adjust.
968 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
969 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
970 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
972 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
973 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
974 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
975 (altivec_lve<VI_char>x): Delete expand.
976 (*altivec_lve<VI_char>x_internal): Rename to...
977 (altivec_lve<VI_char>x): ... this.
978 (altivec_lvxl_<mode>): Delete expand.
979 (*altivec_lvxl_<mode>_internal): Rename to ...
980 (altivec_lvxl_<mode>): ... this.
981 (altivec_stvxl_<mode>): Delete expand.
982 (*altivec_stvxl_<mode>_internal): Rename to ...
983 (altivec_stvxl_<mode>): ... this.
984 (altivec_stve<VI_char>x): Delete expand.
985 (*altivec_stve<VI_char>x_internal): Rename to ...
986 (altivec_stve<VI_char>x): ... this.
987 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
988 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
989 reduc_plus_scal_<mode>): Adjust.
990 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
992 (rs6000_cpu_cpp_builtins): Adjust.
993 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
994 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
995 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
996 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
997 -maltivec=be support.
998 (rs6000_split_vec_extract_var): Adjust.
999 (rs6000_split_v4si_init): Adjust.
1000 (swap_selector_for_mode): Delete.
1001 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1002 altivec_expand_stvex_be): Delete.
1003 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1004 -maltivec=be support.
1005 (rs6000_gimple_fold_builtin): Ditto.
1006 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1008 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1009 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1010 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1011 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1012 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1013 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1014 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1015 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1016 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1017 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1018 anonymous split): Adjust.
1019 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1020 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1022 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1024 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1025 when --with-gxx-include-dir is also specified.
1026 * configure: Regenerate.
1028 2018-05-09 Jim Wilson <jimw@sifive.com>
1031 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1032 * config/riscv/t-withmultilib: New.
1033 * config/riscv/withmultilib.h: New.
1034 * doc/install.texi: Document RISC-V --with-multilib-list support.
1036 2018-05-09 Richard Biener <rguenther@suse.de>
1038 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1040 (vect_bb_vectorization_profitable_p): Adjust. Compute
1041 actual scalar cost using the cost vector and the add_stmt_cost
1044 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1046 PR rtl-optimization/85645
1047 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1048 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1050 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1052 PR rtl-optimization/85645
1053 * shrink-wrap.c (spread_components): Return a boolean saying if
1054 anything was changed.
1055 (try_shrink_wrapping_separate): Iterate spread_components until
1056 nothing changes anymore.
1058 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1060 PR rtl-optimization/85645
1061 * regrename.c (build_def_use): Also kill the chains that include the
1062 destination of a REG_CFA_REGISTER note.
1064 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1066 PR rtl-optimization/85645
1067 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1068 insn that has a REG_CFA_REGISTER note.
1070 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1072 * cfgexpand.c (expand_clobber): New function.
1073 (expand_gimple_stmt_1): Use it.
1074 * tree-vect-stmts.c (vect_clobber_variable): New function,
1076 (vectorizable_simd_clone_call): ...here.
1077 (vectorizable_store): Emit a clobber either side of an
1078 IFN_STORE_LANES sequence.
1079 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1081 2018-05-09 Tom de Vries <tom@codesourcery.com>
1084 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1085 (define_insn "trap_if_false"): Add exit after trap.
1087 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1089 PR rtl-optimization/85638
1090 * bb-reorder.c: Include common/common-target.h.
1091 (create_forwarder_block): New function extracted from...
1092 (fix_up_crossing_landing_pad): ...here. Rename into...
1093 (dw2_fix_up_crossing_landing_pad): ...this.
1094 (sjlj_fix_up_crossing_landing_pad): New function.
1095 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1096 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1097 from both partitions and exit the loop after one iteration.
1099 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1102 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1104 (Basic PowerPC Built-in Functions): The new name of the
1105 subsection previously known as "PowerPC Built-in Functions".
1106 (Basic PowerPC Built-in Functions Available on all Configurations):
1108 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1110 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1112 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1114 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1117 2018-05-08 Jim Wilson <jimw@sifive.com>
1119 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1120 (LD_EMUL_SUFFIX): New.
1121 (LINK_SPEC): Use it.
1123 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1125 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1127 (Basic PowerPC Built-in Functions): The new name of the
1128 subsection previously known as "PowerPC Built-in Functions".
1129 (Basic PowerPC Built-in Functions Available on all Configurations):
1131 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1133 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1135 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1137 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1140 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1143 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1144 after cmpelim optimization.
1146 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1148 * config.gcc: Support "goldmont".
1149 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1150 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1152 * config/i386/i386.c (m_GOLDMONT): Define.
1153 (processor_target_table): Add "goldmont".
1154 (PTA_GOLDMONT): Define.
1155 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1156 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1157 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1158 (fold_builtin_cpu): Add "goldmont".
1159 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1160 (ix86_option_override_internal): Add "goldmont".
1161 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1162 (processor_type): Add PROCESSOR_GOLDMONT.
1163 * config/i386/i386.md: Add CPU "glm".
1164 * config/i386/glm.md: New file.
1165 * config/i386/x86-tune.def: Add m_GOLDMONT.
1166 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1168 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1171 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1173 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1174 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1175 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1178 * config/i386/i386.c (ix86_fold_builtin): Handle
1179 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1182 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1183 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1185 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1188 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1189 (check_arch): Likewise.
1190 (check_fpu): Return the result rather than printing it.
1191 (end arch): Fix operator precedence.
1192 (end cpu): Likewise.
1193 (END): Print the result from check_fpu.
1195 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1196 Alan Hayward <alan.hayward@arm.com>
1197 David Sherwood <david.sherwood@arm.com>
1199 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1200 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1201 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1202 (*fcmuo<mode>_and): New patterns.
1204 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1206 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1207 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1208 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1209 (cmp_op, sve_imm_con): New code attributes.
1210 (SVE_COND_INT_CMP, imm_con): Delete.
1211 (cmp_op): Remove above unspecs from int attribute.
1212 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1214 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1215 comparison-specific unspecs.
1216 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1217 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1218 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1219 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1220 (*vec_fcm<cmp_op><mode>): Rename to...
1221 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1222 (*vec_fcmuo<mode>): Rename to...
1223 (*fcmuo<mode>): ...this and adjust likewise.
1224 (*pred_fcm<cmp_op><mode>): New pattern.
1225 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1226 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1228 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1230 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1231 (aarch64_emit_sve_predicated_cond): New function.
1232 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1233 (aarch64_emit_unspec_cond_or): Replace with...
1234 (aarch64_emit_sve_or_conds): ...this new function. Use
1235 aarch64_emit_sve_ptrue_op for the individual comparisons and
1236 aarch64_emit_binop to OR them together.
1237 (aarch64_emit_inverted_unspec_cond): Replace with...
1238 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1239 aarch64_emit_sve_ptrue_op for the comparison and
1240 aarch64_emit_unop to invert the result.
1241 (aarch64_expand_sve_vec_cmp_float): Update after the above
1242 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1244 2018-05-07 Nathan Sidwell <nathan@acm.org>
1246 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1247 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1248 (Backwards Compatibility): Likewise.
1250 2018-05-07 Luis Machado <luis.machado@linaro.org>
1254 2018-05-07 Luis Machado <luis.machado@linaro.org>
1256 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1257 <prefetch_dynamic_strides>: New const bool field.
1258 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1259 prefetch_dynamic_strides.
1260 (exynosm1_prefetch_tune): Likewise.
1261 (thunderxt88_prefetch_tune): Likewise.
1262 (thunderx_prefetch_tune): Likewise.
1263 (thunderx2t99_prefetch_tune): Likewise.
1264 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1265 (aarch64_override_options_internal): Update to set
1266 PARAM_PREFETCH_DYNAMIC_STRIDES.
1267 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1268 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1269 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1270 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1271 prefetch-dynamic-strides setting.
1273 2018-05-07 Luis Machado <luis.machado@linaro.org>
1275 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1276 <minimum_stride>: New const int field.
1277 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1278 minimum_stride field.
1279 (exynosm1_prefetch_tune): Likewise.
1280 (thunderxt88_prefetch_tune): Likewise.
1281 (thunderx_prefetch_tune): Likewise.
1282 (thunderx2t99_prefetch_tune): Likewise.
1283 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1284 (aarch64_override_options_internal): Update to set
1285 PARAM_PREFETCH_MINIMUM_STRIDE.
1286 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1287 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1288 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1289 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1290 stride is constant and is below the minimum stride threshold.
1292 2018-05-07 Luis Machado <luis.machado@linaro.org>
1294 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1297 2018-05-07 Luis Machado <luis.machado@linaro.org>
1299 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1300 <prefetch_dynamic_strides>: New const bool field.
1301 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1302 prefetch_dynamic_strides.
1303 (exynosm1_prefetch_tune): Likewise.
1304 (thunderxt88_prefetch_tune): Likewise.
1305 (thunderx_prefetch_tune): Likewise.
1306 (thunderx2t99_prefetch_tune): Likewise.
1307 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1308 (aarch64_override_options_internal): Update to set
1309 PARAM_PREFETCH_DYNAMIC_STRIDES.
1310 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1311 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1312 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1313 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1314 prefetch-dynamic-strides setting.
1316 2018-05-07 Luis Machado <luis.machado@linaro.org>
1318 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1319 <minimum_stride>: New const int field.
1320 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1321 minimum_stride field.
1322 (exynosm1_prefetch_tune): Likewise.
1323 (thunderxt88_prefetch_tune): Likewise.
1324 (thunderx_prefetch_tune): Likewise.
1325 (thunderx2t99_prefetch_tune): Likewise.
1326 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1327 (aarch64_override_options_internal): Update to set
1328 PARAM_PREFETCH_MINIMUM_STRIDE.
1329 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1330 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1331 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1332 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1333 stride is constant and is below the minimum stride threshold.
1335 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1338 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1339 the type is addressable. Don't force op into register if it has
1342 2018-05-05 Roland McGrath <mcgrathr@google.com>
1345 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1346 any section for which we don't know a specific type it should have,
1347 regardless of name. Previously this was done only for the exact
1348 names ".init_array", ".fini_array", and ".preinit_array".
1349 (default_elf_asm_named_section): Add comment about
1350 relationship with default_section_type_flags and SECTION_NOTYPE.
1351 (get_section): Don't consider it a type conflict if one side has
1352 SECTION_NOTYPE and the other doesn't, as long as neither has the
1353 SECTION_BSS et al used in the default_section_type_flags logic.
1355 2018-05-05 Tom de Vries <tom@codesourcery.com>
1358 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1359 (workaround_barsyncs): New function.
1360 (nvptx_reorg): Use workaround_barsyncs.
1361 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1362 (define_expand "nvptx_membar_cta"): New define_expand.
1363 (define_insn "*nvptx_membar_cta"): New insn.
1365 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1367 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1368 To improve optimization opportunities.
1369 * builtin-types.def: The new needed builtin types for the above.
1371 2018-05-04 Richard Biener <rguenther@suse.de>
1373 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1374 * gimple-ssa-store-merging.c
1375 (imm_store_chain_info::output_merged_store): Remove redundant create,
1376 release split_store vector contents on failure.
1377 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1378 scalar stmt vector on cache hit.
1380 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1382 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1384 * config.gcc (powerpc-xilinx-eabi*): Remove.
1385 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1387 (fusion_addis_mem_combo_load): Ditto.
1388 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1390 (rs6000_cpu_cpp_builtins): Ditto.
1391 * config/rs6000/rs6000-linux.c
1392 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1393 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1394 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1396 (rs6000_setup_reg_addr_masks): Ditto.
1397 (rs6000_init_hard_regno_mode_ok): Ditto.
1398 (rs6000_option_override_internal): Ditto.
1399 (legitimate_lo_sum_address_p): Ditto.
1400 (rs6000_legitimize_address): Ditto.
1401 (rs6000_legitimize_reload_address): Ditto.
1402 (rs6000_legitimate_address_p): Ditto.
1403 (abi_v4_pass_in_fpr): Ditto.
1404 (setup_incoming_varargs): Ditto.
1405 (rs6000_gimplify_va_arg): Ditto.
1406 (rs6000_split_multireg_move): Ditto.
1407 (rs6000_savres_strategy): Ditto.
1408 (rs6000_emit_prologue_components): Ditto.
1409 (rs6000_emit_epilogue_components): Ditto.
1410 (rs6000_emit_prologue): Ditto.
1411 (rs6000_emit_epilogue): Ditto.
1412 (rs6000_elf_file_end): Ditto.
1413 (rs6000_function_value): Ditto.
1414 (rs6000_libcall_value): Ditto.
1415 * config/rs6000/rs6000.h: Ditto.
1416 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1417 (TARGET_MINMAX): ... this. New.
1418 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1419 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1420 (*movsi_internal1_single): Delete.
1421 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1422 mfpu=, mxilinx-fpu): Delete.
1423 * config/rs6000/singlefp.h: Delete.
1424 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1425 * config/rs6000/t-rs6000: Ditto.
1426 * config/rs6000/t-xilinx: Delete.
1427 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1428 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1429 (VStype_simple): Delete.
1430 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1431 * config/rs6000/xfpu.h: Delete.
1432 * config/rs6000/xfpu.md: Delete.
1433 * config/rs6000/xilinx.h: Delete.
1434 * config/rs6000/xilinx.opt: Delete.
1435 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1436 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1438 2018-05-04 Tom de Vries <tom@codesourcery.com>
1441 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1444 2018-05-04 Richard Biener <rguenther@suse.de>
1447 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1448 (expand_complex_div_wide): Likewise.
1449 (expand_complex_operations_1): Likewise.
1450 (expand_complex_libcall): Preserve EH info of the original stmt.
1451 (tree_lower_complex): Handle removed blocks.
1452 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1453 on complex multiplication and division libcall builtins.
1455 2018-05-04 Richard Biener <rguenther@suse.de>
1458 * fold-const.c (negate_expr_p): Restrict negation of operand
1459 zero of a division to when we know that can happen without
1461 (fold_negate_expr_1): Likewise.
1463 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1466 * real.h (real_nextafter): Declare.
1467 * real.c (real_nextafter): New function.
1468 * fold-const-call.c (fold_const_nextafter): New function.
1469 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1470 CASE_CFN_NEXTTOWARD.
1471 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1472 even when arg1_mode is different from arg0_mode.
1474 2018-05-03 Nathan Sidwell <nathan@acm.org>
1476 * doc/extend.texi (Deprecated Features): Remove
1478 (Backwards Compatibility): Likewise.
1479 * doc/invoke.texi (C++ Language Options): Likewise.
1480 (C++ Dialect Options): Likewise.
1482 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1485 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1486 _mm512_mask_mullox_epi64): New intrinsics.
1488 2018-05-03 Tom de Vries <tom@codesourcery.com>
1491 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1492 dump files): Add offload-tree.
1494 2018-05-03 Richard Biener <rguenther@suse.de>
1496 PR tree-optimization/85615
1497 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1498 to loops not nested in BBs loop father to avoid creating multi-entry
1501 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1503 PR tree-optimization/70291
1504 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1505 arguments. Change return type to tree. Emit libcall as a new
1506 statement rather than replacing existing one when inplace_p is true.
1507 (expand_complex_multiplication_components): New function.
1508 (expand_complex_multiplication): Expand floating-point complex
1509 multiplication using the above.
1510 (expand_complex_division): Rename inner_type parameter to type.
1511 Update expand_complex_libcall call-site.
1512 (expand_complex_operations_1): Update expand_complex_multiplication
1513 and expand_complex_division call-sites.
1515 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1518 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1519 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1520 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1521 the highest significant bit of the shift count mask is clear. In
1522 check whether and[sq]i3 is needed verify that all significant bits
1523 of the shift count other than the highest are set.
1525 2018-05-02 Tom de Vries <tom@codesourcery.com>
1528 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1529 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1530 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1531 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1532 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1533 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1534 __builtin_goacc_parlevel_size.
1536 2018-05-02 Richard Biener <rguenther@suse.de>
1538 PR tree-optimization/85597
1539 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1540 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1543 2018-05-02 Tom de Vries <tom@codesourcery.com>
1546 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1547 dump files): Add ltrans-tree.
1549 2018-05-02 Tom de Vries <tom@codesourcery.com>
1552 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1553 dump files): Add wpa-ipa.
1555 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1557 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1558 powerpc*-*-linux*paired* target.
1559 * config/rs6000/750cl.h: Delete.
1560 * config/rs6000/paired.h: Delete.
1561 * config/rs6000/paired.md: Delete.
1562 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1564 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1565 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1566 comment. Remove paired float support.
1567 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1568 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1570 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1571 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1573 * config/rs6000/rs6000.c: Remove paired float support.
1574 (paired_expand_vector_init, paired_expand_vector_move,
1575 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1576 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1577 paired_expand_builtin, paired_expand_predicate_builtin,
1578 paired_init_builtins): Delete.
1579 * config/rs6000/rs6000.h: Remove paired float support.
1580 * config/rs6000/rs6000.md: Remove paired float support.
1581 (move_from_CR_ov_bit): Delete.
1582 * config/rs6000/rs6000.opt (mpaired): Delete.
1583 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1584 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1586 2018-05-02 Richard Biener <rguenther@suse.de>
1589 * gimplify.c (gimplify_save_expr): When in SSA form allow
1590 SAVE_EXPRs to compute to SSA vars.
1592 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1595 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1596 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1597 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1598 clobber operands[2], instead use a new pseudo. Formatting fixes.
1600 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1602 PR tree-optimization/85586
1603 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1604 exit early for statements in the same group if the accesses are
1607 2018-05-02 Tom de Vries <tom@codesourcery.com>
1610 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1613 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1615 PR tree-optimization/85143
1616 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1618 2018-05-01 Tom de Vries <tom@codesourcery.com>
1621 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1622 not found" error message.
1624 2018-05-01 Tom de Vries <tom@codesourcery.com>
1627 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1628 * vec.c (test_ordered_remove_if): New function.
1629 (vec_c_tests): Call test_ordered_remove_if.
1630 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1631 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1632 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1633 VEC_ORDERED_REMOVE_IF.
1635 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1637 PR tree-optimization/82665
1638 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1639 pointer subtraction where arguments come from a memchr call.
1641 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1643 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1644 --push-state --as-needed and --pop-state instead of --as-needed and
1645 --no-as-needed if ld supports it.
1646 * configure: Regenerated.
1649 * doc/install.texi2html: Replace _002d with - and _002a with * in
1650 generated html files using sed.
1652 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1655 * gcc-rich-location.c (blank_line_before_p): New function.
1656 (use_new_line): New function.
1657 (gcc_rich_location::add_fixit_insert_formatted): New function.
1658 * gcc-rich-location.h
1659 (gcc_rich_location::add_fixit_insert_formatted): New function.
1661 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1663 * selftest.c (assert_streq): Rename "expected" and "actual" to
1664 "val1" and "val2". Extend NULL-handling to cover both inputs
1665 symmetrically, while still requiring both to be non-NULL for a pass.
1666 * selftest.h (assert_streq): Rename "expected" and "actual" to
1668 (ASSERT_EQ): Likewise.
1669 (ASSERT_EQ_AT): Likewise.
1670 (ASSERT_KNOWN_EQ): Likewise.
1671 (ASSERT_KNOWN_EQ_AT): Likewise.
1672 (ASSERT_NE): Likewise.
1673 (ASSERT_MAYBE_NE): Likewise.
1674 (ASSERT_MAYBE_NE_AT): Likewise.
1675 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1676 the assertion to pass.
1677 (ASSERT_STREQ_AT): Likewise.
1679 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1681 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1682 interaction with -pie.
1684 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1686 * selftest.h: Fix alphabetization of per-source-file selftest
1689 2018-04-30 Jason Merrill <jason@redhat.com>
1691 PR c++/61982 - dead stores to destroyed objects.
1692 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1695 2018-04-30 Jason Merrill <jason@redhat.com>
1697 * tree.c (build_clobber): New.
1698 * tree.h: Declare it.
1699 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1701 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1703 * diagnostic-show-locus.c (layout::layout): Update for
1704 location_get_source_line returning a char_span.
1705 (struct char_span): Move to input.h.
1706 (struct correction): Update for fields in char_span becoming
1708 (struct source_line): Update for location_get_source_line
1709 returning a char_span.
1710 (layout::print_line): Likewise.
1711 * edit-context.c (edited_file::print_content): Likewise.
1712 (edited_file::print_diff_hunk): Likewise.
1713 (edited_file::print_run_of_changed_lines): Likewise.
1714 (edited_file::get_num_lines): Likewise.
1715 (edited_line::edited_line): Likewise.
1716 * final.c (asm_show_source): Likewise.
1717 * input.c (location_get_source_line): Convert return type
1718 from const char * to char_span, losing the final "line_len"
1720 (dump_location_info): Update for the above.
1721 (get_substring_ranges_for_loc): Likewise. Use a char_span
1722 when handling the literal within the line.
1723 (test_reading_source_line): Update for location_get_source_line
1724 returning a char_span.
1725 * input.h (class char_span): Move here from
1726 diagnostic-show-locus.c, converting from a struct to a class.
1727 Make data members private.
1728 (char_span::operator bool): New.
1729 (char_span::length): New.
1730 (char_span::get_buffer): New.
1731 (char_span::operator[]): New.
1732 (char_span::subspan): Make const.
1733 (char_span::xstrdup): New.
1734 (location_get_source_line): Convert return type from const char *
1735 to char_span, losing the final "line_size" param.
1737 2018-04-30 Jan Hubicka <jh@suse.cz>
1739 * lto-wrapper.c (ltrans_priorities): New static var.
1740 (cmp_priority): New.
1741 (run_gcc): Read priorities and if doing parallel build order
1742 the Makefile by them.
1744 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1746 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1748 2018-04-30 Richard Biener <rguenther@suse.de>
1750 * tree-cfg.c (verify_address): Remove base argument, add
1751 flag whether to check TREE_ADDRESSABLE and do that.
1752 (verify_expr): Remove.
1753 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1754 (verify_gimple_assign_single): Likewise.
1755 (verify_gimple_switch): Likewise.
1756 (verify_expr_location_1): Dereference tp once. Add (disabled)
1757 piece from verify_expr.
1758 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1760 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1762 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1764 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1766 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1767 (small_data_pattern): Likewise.
1768 (arc_rewrite_small_data): Likewise.
1769 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1770 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1771 (get_symbol_alignment): New function.
1772 (legitimate_small_data_address_p): Likewise.
1773 (legitimate_scaled_address): Update, call
1774 legitimate_small_data_address_p.
1775 (output_sdata): New static variable.
1776 (arc_print_operand): Update how we handle small data operands.
1777 (arc_print_operand_address): Likewise.
1778 (arc_legitimate_address_p): Update, use
1779 legitimate_small_data_address_p.
1780 (arc_rewrite_small_data_p): Remove.
1781 (arc_rewrite_small_data_1): Likewise.
1782 (arc_rewrite_small_data): Likewise.
1783 (small_data_pattern): Likewise.
1784 (compact_sda_memory_operand): Update to use
1785 legitimate_small_data_address_p and get_symbol_alignment.
1786 (prepare_move_operands): Don't rewite sdata pattern.
1787 (prepare_extend_operands): Remove.
1788 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1790 (zero_extendqisi2): Likewise.
1791 (zero_extendhisi2): Likewise.
1792 (extendqihi2): Likewise.
1793 (extendqisi2): Likewise.
1794 (extendhisi2): Likewise.
1798 * config/arc/constraints.md (Usd): Change it to memory constraint.
1800 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1802 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1803 as source of std instructions.
1804 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1805 allow 6-bit constants as source for store instructions.
1806 (movdi_insn): Update instruction pattern to allow 6-bit constants
1807 as source for store instructions.
1809 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1811 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1813 2018-04-30 Nathan Sidwell <nathan@acm.org>
1814 Sandra Loosemore <sandra@codesourcery.com>
1816 * dumpfile.c (dump_open): Allow '-' for stdout.
1817 * doc/invoke.texi (Developer Options): Document dump filename
1818 determination early. Document stdin/stdout selection.
1820 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1822 Microblaze Target: PIC data text relative
1824 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1825 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1827 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1828 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1829 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1830 New addressing mode for data-text relative position indepenedent code.
1831 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1832 'ADDRESS_SYMBOLIC_TXT_REL'.
1833 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1834 (microblaze_legitimate_pic_operand): Exclude function calls from
1835 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1836 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1838 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1839 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1840 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1841 for 'address + offset'.
1842 (microblaze_expand_prologue): Add new function prologue call for
1844 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1845 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1846 table in case of TARGET_PIC_DATA_TEXT_REL.
1847 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1848 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1849 Add new macros 'UNSPEC_TEXT',
1850 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1851 + exclude function calls from 'UNSPEC_PLT' in case of data text
1853 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1854 new target hook for generating address diff vector tables in case of
1856 * doc/tm.texi : Regenerate.
1857 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1858 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1859 of addr diff vector generation.
1860 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1861 target hook definition.
1862 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1863 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1864 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1865 option for data text relative.
1867 2018-04-30 Richard Biener <rguenther@suse.de>
1869 * tree-chrec.h (evolution_function_is_constant_p): Remove
1871 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1873 2018-04-30 Richard Biener <rguenther@suse.de>
1876 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1878 2018-04-30 Richard Biener <rguenther@suse.de>
1880 PR tree-optimization/28364
1881 PR tree-optimization/85275
1882 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1883 copying first exit test.
1885 2018-04-28 Mark Wielaard <mark@klomp.org>
1887 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1889 (dwarf_AT): Handle DW_AT_addr_base.
1890 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1892 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1895 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1896 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1897 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1898 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1900 2018-04-28 Richard Biener <rguenther@suse.de>
1902 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1903 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1904 to reflect use. Only add interesting stmts.
1906 2018-04-27 Martin Jambor <mjambor@suse.cz>
1909 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1910 the jump function allows for passing through aggregate values.
1912 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1914 * input.h (in_system_header_at): Convert from macro to inline
1916 (from_macro_expansion_at): Likewise.
1917 (from_macro_definition_at): Likewise.
1919 2018-04-27 Jeff Law <law@redhat.com>
1921 * config.gcc: Mark tile* targets as deprecated/obsolete.
1923 2018-04-27 Richard Biener <rguenther@suse.de>
1925 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1928 2018-04-27 Richard Biener <rguenther@suse.de>
1930 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1932 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1934 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1935 with Yd constraint. Set "preferred_for_speed" attribute from
1936 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1938 (*movdi_internal): Ditto.
1939 (movti_interunit splitters): Remove
1940 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1941 (movdi_interunit splitters): Ditto.
1942 * config/i386/constraints.md (Ye): Remove.
1943 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1945 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1948 * config/aarch64/constraints.md (Usg): Limit to 31.
1951 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1953 PR tree-optimization/85529
1954 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1955 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1956 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1957 zero extension or masking of the MSB bit.
1958 (optimize_range_tests): Add FIRST_BB argument, pass it through
1959 to optimize_range_tests_var_bound.
1960 (maybe_optimize_range_tests, reassociate_bb): Adjust
1961 optimize_range_tests callers.
1963 2018-04-26 Richard Biener <rguenther@suse.de>
1964 Jakub Jelinek <jakub@redhat.com>
1966 * cgraph.h (symbol_table): Just declare debug method here.
1967 * symtab.c (symbol_table::debug): Define.
1969 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1971 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1973 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1975 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1976 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1977 (*movdi_internal): Substitute Yi and Yj constraint with x
1978 and Ym and Yn constraint with y constraint. Update "isa"
1979 attribute and set "preferred_for_speed" attribute from
1980 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1981 (*movsi_internal): Ditto.
1982 (*movdf_internal): Ditto.
1983 (*movsf_internal): Ditto.
1984 (*zero_extendsidi2): Ditto.
1985 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1986 (sse2_loadld): Ditto.
1987 (*vec_extract<ssevecmodelower>_0): Ditto.
1988 (*vec_extractv4si_0_zext_sse4): Ditto.
1989 (vec_concatv2di): Ditto.
1990 (*vec_dup<mode>): Ditto.
1991 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1992 * config/i386/constraints.md (Yi): Remove.
1997 2018-04-26 Nathan Sidwell <nathan@acm.org>
1999 * dumpfile.c (dump_open): New.
2000 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2001 (dump_finish): Detect stdio/stderr by value not name.
2003 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2005 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2007 2018-04-26 Tom de Vries <tom@codesourcery.com>
2010 * config/nvptx/nvptx.c (verify_neutering_jumps)
2011 (verify_neutering_labels): New function
2012 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2014 2018-04-26 Tom de Vries <tom@codesourcery.com>
2017 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2018 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2021 2018-04-26 Richard Biener <rguenther@suse.de>
2022 Tom de Vries <tom@codesourcery.com>
2025 * lto-streamer-out.c (output_function): Fixup loops if required to match
2026 discovery done in the reader.
2028 2018-04-26 Richard Biener <rguenther@suse.de>
2030 PR tree-optimization/85116
2031 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2032 have a loop exit from the single latch predecessor. Remove
2033 case of header with just condition.
2034 (ch_base::copy_headers): Exclude infinite loops from any
2036 (pass_ch::execute): Record exits.
2038 2018-04-26 Richard Biener <rguenther@suse.de>
2040 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2041 prologue cost vector and pass it to vect_get_load_cost.
2042 (vect_get_peeling_costs_all_drs): Likewise.
2043 (vect_peeling_hash_get_lowest_cost): Likewise.
2044 (vect_enhance_data_refs_alignment): Likewise.
2046 2018-04-26 Richard Biener <rguenther@suse.de>
2049 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2050 checking of integer<->pointer conversions.
2051 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2052 sign-/zero-extending pointer types.
2053 (expand_omp_for_static_chunk): Likewise.
2055 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2056 Jean Lee <xiaoyur347@gmail.com>
2058 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2059 (TARGET_ASAN_SHADOW_OFFSET): Define.
2060 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2061 true for -fsanitize=address.
2063 2018-04-25 Mark Wielaard <mark@klomp.org>
2065 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2068 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2070 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2071 than "alu", remove explicit "memory" and "imm_disp" attributes.
2072 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2075 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2076 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2077 gen_lowpart_no_emit.
2079 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2082 * config/i386/i386.c (ix86_expand_builtin): Change memory
2083 operand to XI, extend p0 to Pmode.
2084 * config/i386/i386.md: Change unspec volatile and operand
2085 1 mode to XI, change operand 0 mode to P.
2087 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2089 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2090 GET_MODE_MASK before any checking.
2091 (nds32_can_use_bset_p): Likewise.
2092 (nds32_can_use_btgl_p): Likewise.
2094 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2096 * config/nds32/nds32-doubleword.md: New define_split pattern for
2097 illegal register number.
2099 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2101 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2103 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2105 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2107 2018-04-25 Richard Biener <rguenther@suse.de>
2109 * lto-streamer.h (LTO_major_version): Bump to 8.
2111 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2113 * BASE-VER: Set to 9.0.0.
2115 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2117 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2118 in __abskf2 and __powikf2.
2120 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2123 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2124 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2125 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2126 Use the above on operand 2. Reindent.
2127 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2129 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2132 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2134 * config/i386/i386.opt (mcet): Removed.
2135 * doc/install.texi: Remove -mcet documentation.
2136 * doc/invoke.texi: Likewise.
2138 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2141 * doc/install.texi: Remove -mcet from bootstrap-cet.
2143 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2146 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2147 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2151 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2152 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2153 containing a CONST_VECTOR.
2155 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2157 * doc/install.texi: Update newlib dependency for nvptx.
2159 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2162 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2163 instead of INTVAL when shifting x left.
2165 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2167 PR tree-optimization/85478
2168 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2169 vect_grouped_store_supported for single element vectors.
2171 2018-04-24 Richard Biener <rguenther@suse.de>
2174 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2175 load cost increase to the case of non-constant step.
2177 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2180 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2181 destination if any_malformed_asm.
2183 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2186 * expr.c (store_field): In the bitfield case, if the value comes from
2187 a function call and is returned in registers by means of a PARALLEL,
2188 do not change the mode of the temporary unless BLKmode and VOIDmode.
2190 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2192 PR rtl-optimization/85423
2193 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2194 dependencies to debug insns when the previous insn is non-debug.
2196 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2198 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2199 enums into a single definition.
2200 (fls): Fix predicates and printing.
2203 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2205 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2206 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2207 and short u6 immediate.
2208 (check_if_valid_sleep_operand): Remove.
2209 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2211 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2213 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2214 flag_always_save_lp condition.
2215 * config/nds32/nds32.opt (malways-save-lp): New option.
2217 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2219 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2220 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2221 * config/nds32/nds32.h
2222 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2223 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2225 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2227 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2228 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2230 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2231 Chung-Ju Wu <jasonwucj@gmail.com>
2233 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2235 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2236 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2238 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2240 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2242 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2244 * config/nds32/nds32-protos.h (nds32_data_alignment,
2245 nds32_local_alignment): Declare.
2246 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2247 nds32_local_alignment): New functions.
2248 (TARGET_CONSTANT_ALIGNMENT): Define.
2249 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2251 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2253 * config/nds32/nds32.c
2254 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2255 (TARGET_MODES_TIEABLE_P): Likewise.
2257 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2259 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2262 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2263 Chung-Ju Wu <jasonwucj@gmail.com>
2265 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2266 for unaligned access.
2267 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2269 * config/nds32/nds32-intrinsic.md: Likewise.
2270 * config/nds32/nds32_intrinsic.h: Likewise.
2271 * config/nds32/nds32.h (nds32_builtins): Likewise.
2272 * config/nds32/nds32.opt (munaligned-access): New option.
2273 * config/nds32/nds32.c (nds32_asm_file_start): Display
2274 flag_unaligned_access status.
2276 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2278 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2279 -mno-relax is present.
2280 * config/riscv/linux.h (LINK_SPEC): Ditto.
2282 2018-04-20 Martin Sebor <msebor@redhat.com>
2285 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2287 (gimple_fold_builtin_stxcpy_chk): Same.
2288 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2290 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2293 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2294 __powikf2 when long double is IEEE 128-bit.
2296 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2298 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2299 step to make sure stack always aligned.
2301 2018-04-20 Carl Love <cel@us.ibm.com>
2304 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2305 size check for arg0.
2307 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2308 Tom de Vries <tom@codesourcery.com>
2311 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2312 Emit insns for calls too.
2313 (nvptx_find_par): Always look for worker-level predecessor insn.
2314 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2316 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2317 (nvptx_process_pars): Propagate frames for calls.
2319 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2322 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2324 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2325 (ix86_handle_option): Don't handle OPT_mibt.
2326 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2328 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2330 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2331 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2332 (ix86_target_macros): Define __CET__ with flag_cf_protection
2333 for -fcf-protection.
2334 * config/i386/i386.c (isa2_opts): Remove -mibt.
2335 * config/i386/i386.h (TARGET_IBT): Removed.
2336 (TARGET_IBT_P): Likewise.
2337 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2338 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2339 * config/i386/i386.opt (mcet): Update help message.
2342 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2343 -mcet as an alias for -mshstk.
2345 2018-04-20 Richard Biener <rguenther@suse.de>
2348 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2349 complexity by forcing a single use of the multiply operand.
2351 2018-04-20 Martin Jambor <mjambor@suse.cz>
2354 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2355 recursion dependency to only apply to non-clones.
2357 2018-04-20 Martin Jambor <mjambor@suse.cz>
2360 * ipa-cp.c (create_specialized_node): Check that clones of
2361 self-recursive edges exist during IPA-CP.
2363 2018-04-19 Toon Moene <toon@moene.org>
2365 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2368 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2370 PR tree-optimization/85467
2371 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2372 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2373 VECTOR_CST element to type.
2375 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2378 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2379 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2380 (builtin_longjmp): Likewise.
2381 (save_stack_nonlocal): New pattern.
2382 (restore_stack_nonlocal): Likewise.
2384 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2387 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2388 Replace ASM_OUTPUT_LABEL with fprintf.
2390 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2393 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2394 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2395 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2396 define __IBT__ and __SHSTK__ for -fcf-protection.
2397 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2399 (ix86_trampoline_init): Likewise.
2400 (x86_output_mi_thunk): Likewise.
2401 (ix86_notrack_prefixed_insn_p): Likewise.
2402 (ix86_option_override_internal): Don't disallow -fcf-protection.
2403 * config/i386/i386.md (rdssp<mode>): Also enable for
2405 (incssp<mode>): Likewise.
2406 (nop_endbr): Likewise.
2407 * config/i386/i386.opt (mcet): Change help message to built-in
2411 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2412 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2413 enable CET built-in functions.
2415 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2417 * common/config/i386/i386-common.c
2418 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2419 OPTION_MASK_ISA_MOVDIRI_UNSET,
2420 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2421 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2422 * config.gcc (movdirintrin.h): New header.
2423 * config/i386/cpuid.h (bit_MOVDIRI,
2424 bit_MOVDIR64B): New bits.
2425 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2427 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2428 (VOID, PVOID, PCVOID)): New function types.
2429 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2430 __builtin_ia32_directstoreu_u64,
2431 __builtin_ia32_movdir64b): New builtins.
2432 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2433 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2435 (ix86_valid_target_attribute_inner_p): Ditto.
2436 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2437 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2438 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2439 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2440 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2441 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2442 (movdiri<mode>, movdir64b_<mode>): New.
2443 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2444 * config/i386/immintrin.h: Include movdirintrin.h.
2445 * config/i386/movdirintrin.h: New file.
2446 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2448 2018-04-19 Richard Biener <rguenther@suse.de>
2451 * cfg.c (clear_bb_flags): When loop state says we have
2452 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2454 2018-04-19 Richard Biener <rguenther@suse.de>
2456 PR tree-optimization/84737
2457 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2458 copying restrict info.
2459 (vect_setup_realignment): Use it.
2460 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2461 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2462 the first DR to all generated stores.
2463 (vectorizable_load): Likewise for loads.
2465 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2467 PR tree-optimization/85446
2468 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2469 the integral and pointer types to have the same precision.
2471 * doc/install.texi: Document --disable-cet being the default and
2474 2018-04-18 Martin Liska <mliska@suse.cz>
2476 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2479 2018-04-18 Martin Liska <mliska@suse.cz>
2482 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2485 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2486 arguments if they are comparable.
2488 2018-04-18 Martin Liska <mliska@suse.cz>
2491 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2494 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2497 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2500 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2501 ENDBR after calling __morestack.
2503 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2506 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2507 by using gcc_base_ver to generate a gcc_driver_version, and use
2508 it when generating GCC_DRIVER_NAME.
2509 * configure: Regenerate.
2511 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2514 * config.gcc: Obsolete powerpc*-*-*spe*.
2516 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2519 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2520 (stabstr_D): Change type of unum from unsigned int to
2521 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2524 2018-04-17 Jim Wilson <jimw@sifive.com>
2527 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2528 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2529 Set arg_pointer_offset after using pretend_args_size.
2531 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2533 PR rtl-optimization/85431
2534 * dse.c (record_store): Ignore zero width stores.
2537 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2538 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2539 __builtin_stack_restore rather than after it.
2540 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2541 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2542 argument instead of virtual_dynamic_stack_rtx.
2544 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2546 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2548 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2549 Add note to error message to explain internal mapping of overloaded
2550 built-in function name to non-overloaded built-in function name.
2551 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2554 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2557 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2558 where the inputs overlap with the output.
2560 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2562 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2563 (=v, v) alternative and explicit "memory" attribute.
2564 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2565 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2567 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2568 "sselog1" type instead of "sselog".
2569 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2570 "sselog". Remove explicit "memory" attribute.
2571 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2572 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2574 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2575 "sselog1" type instead of "sselog". Remove explicit "memory"
2577 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2578 use "sselog1" type instead of "sselog". Remove explicit "memory"
2580 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2581 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2583 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2584 "sselog1" type instead of "sselog". Remove explicit "memory"
2586 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2587 use "sselog1" type instead of "sselog". Remove explicit "memory"
2591 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2594 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2597 2018-04-17 Martin Jambor <mjambor@suse.cz>
2600 * ipa-cp.c (create_specialized_node): Call
2601 expand_all_artificial_thunks if necessary.
2603 2018-04-17 Martin Liska <mliska@suse.cz>
2606 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2607 in message, remote space in between '_G' and '('.
2609 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2612 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2613 avx512f_vmcmp<mode>3<round_saeonly_name>,
2614 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2615 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2616 avx512f_rndscale<mode><round_saeonly_name>,
2617 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2618 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2619 Use %<iptr>2 instead of %2 for -masm=intel.
2620 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2621 avx512f_vcvttss2usi<round_saeonly_name>,
2622 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2624 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2625 avx512f_vcvttsd2usi<round_saeonly_name>,
2626 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2627 Use %q1 instead of %1 for -masm=intel.
2628 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2629 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2630 of %3 for -masm=intel.
2631 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2633 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2635 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2637 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2638 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2640 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2642 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2643 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2644 %g1 and one with %0 and %1.
2645 (avx512er_vmrcp28<mode><round_saeonly_name>,
2646 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2648 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2649 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2650 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2651 of %0 and %{%4%} for -masm=intel.
2652 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2653 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2654 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2655 order of %0 and %{%5%}%{z%} for -masm=intel.
2657 2018-04-17 Jan Hubicka <jh@suse.cz>
2660 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2662 2018-04-17 Martin Liska <mliska@suse.cz>
2665 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2666 for target_clone error message. Make default implementation
2667 clone to be a local declaration.
2668 (separate_attrs): Add new argument and check for an empty
2670 (expand_target_clones): Handle it.
2671 (ipa_target_clone): Make redirection just for target_clones
2674 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2675 Tom de Vries <tom@codesourcery.com>
2678 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2679 tiled basic blocks without omp continue statements.
2681 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2684 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2685 vec_extract expression as having side effects to make sure it gets
2688 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2691 * config/i386/i386.c (get_builtin_code_for_version): Check
2694 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2697 * gcc/config.gcc: Support "skylake".
2698 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2700 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2701 (processor_target_table): Add "skylake".
2702 (ix86_option_override_internal): Add "skylake".
2703 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2704 PROCESSOR_CANNONLAKE.
2705 (get_builtin_code_for_version): Fix priority for
2706 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2707 PROCESSOR_SKYLAKE-AVX512.
2708 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2709 (processor_type): Add PROCESSOR_SKYLAKE.
2711 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2712 Jason Merrill <jason@redhat.com>
2715 * convert.c (convert_to_integer_1): Use direct recursion for
2716 enumeral types and types with a precision less than the number
2717 of bits in their mode.
2719 2018-04-16 Julia Koval <julia.koval@intel.com>
2722 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2723 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2725 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2728 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2729 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2730 and -mno-direct-move.
2732 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2735 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2736 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2737 Ensure negative shifts result in {0}.
2739 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2741 PR rtl-optimization/79916
2742 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2743 regs (if any) to define how to gnerate SD moves when LRA is in
2746 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2748 PR rtl-optimization/85393
2749 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2750 * except.c (expand_dw2_landing_pad_for_region): Make static.
2751 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2752 a label and unconditional jump to old_bb, rather than
2753 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2756 PR rtl-optimization/85376
2757 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2758 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2759 instead of a specific value.
2761 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2762 Bin Cheng <bin.cheng@arm.com>
2764 PR tree-optimization/82965
2765 PR tree-optimization/83991
2766 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2767 by_profile_only parameter.
2768 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2769 information if the loop was predicted to iterate too many times.
2770 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2772 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2775 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2778 2018-04-13 Martin Liska <mliska@suse.cz>
2779 Jakub Jelinek <jakub@redhat.com>
2782 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2783 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2784 * builtins.c (expand_builtin_memory_copy_args): Use
2785 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2786 handle dest_addr == pc_rtx.
2788 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2791 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2792 asked to not generate direct moves.
2793 (fix_trunc<mode>si2_stfiwx): Similar.
2794 (fix_trunc<mode>si2_internal): Similar.
2796 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2799 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2800 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2801 lookup if dest in some wider mode is known to be const0_rtx and
2802 if so, record permanent equivalence for it to be ZERO_EXTEND of
2803 the narrower mode destination.
2805 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2807 * lto-streamer-out.c (output_function): Revert 259346.
2808 * omp-expand.c (expand_oacc_for): Likewise.
2810 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2812 PR rtl-optimization/85354
2813 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2814 * sel-sched.c (sel_global_init): ... here.
2816 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2819 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2820 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2821 mode for PE-COFF targets.
2822 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2823 (i386_pe_asm_lto_end): Likewise.
2824 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2825 (TARGET_ASM_LTO_END): Likewise.
2826 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2827 (i386_pe_asm_lto_start): New function.
2828 (i386_pe_asm_lto_end): Likewise.
2830 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2831 Richard Biener <rguenther@suse.de>
2834 * lto-streamer-out.c (output_function): Fix CFG loop state before
2836 * omp-expand.c (expand_oacc_for): Handle calls to internal
2837 functions like regular functions.
2839 2018-04-12 Richard Biener <rguenther@suse.de>
2842 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2843 for the early LTO debug to properly generate references to it
2844 during DIE emission. Do not re-use that for the skeleton for
2846 (dwarf2out_early_finish): Likewise.
2848 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2851 * config/i386/sse.md
2852 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2853 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2854 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2855 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2856 and output is a reg, avoid creating invalid lowpart subreg, but
2857 instead split into a 512-bit move. Don't split if not AVX512VL,
2858 input is xmm16+ reg and output is a mem.
2859 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2860 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2861 xmm16+ reg and output is a mem.
2863 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2865 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2866 also for flag_dwarf2_cfi_asm.
2868 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2870 PR rtl-optimization/85342
2871 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2872 a bool scalar var inside of the loop instead. Don't try to update
2873 recog_data.operand after failed apply_change_group.
2875 2018-04-12 Tom de Vries <tom@codesourcery.com>
2878 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2879 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2880 array with flexible array member as array without given dimension.
2881 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2882 argument for undefined param to true.
2884 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2887 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2888 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2889 from PowerPC section.
2890 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2891 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2892 help text that is too long.
2893 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2894 help text that is too long.
2895 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2896 help text that is too long.
2898 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2900 * config/alpha/alpha.md (stack_probe_internal): Rename
2901 from "probe_stack". Update all callers.
2903 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2905 PR rtl-optimization/84566
2906 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2907 sched_macro_fuse_insns.
2909 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2912 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2913 (compute_block_dependences): ... from here.
2915 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2917 PR tree-optimization/85331
2918 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2919 from int to HOST_WIDE_INT.
2921 2018-04-11 Martin Jambor <mjambor@suse.cz>
2924 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2925 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2926 not the same as the source val.
2927 (cgraph_edge_brings_value_p): New parameter.
2928 (gather_edges_for_value): Pass destination value to
2929 cgraph_edge_brings_value_p.
2930 (perhaps_add_new_callers): Likewise.
2931 (get_info_about_necessary_edges): Likewise and exclude values brought
2932 only by self-recursive edges.
2933 (create_specialized_node): Redirect only clones of self-calling edges.
2934 (+self_recursive_pass_through_p): New function.
2935 (find_more_scalar_values_for_callers_subset): Use it.
2936 (find_aggregate_values_for_callers_subset): Likewise.
2937 (known_aggs_to_agg_replacement_list): Removed.
2938 (decide_whether_version_node): Re-calculate known constants for all
2939 remaining context clones.
2941 2018-04-11 Richard Biener <rguenther@suse.de>
2944 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2945 from early DWARF output.
2946 (dwarf2out_early_finish): Output line info unconditionally into
2947 early DWARF and add reference to it.
2949 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2952 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2953 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2954 other than V2DFmode using iptr mode attribute.
2955 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2957 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2959 PR rtl-optimization/84659
2960 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2962 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2965 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2967 (output_loc_list): Pass address of a dummy size variable even in the
2968 locview handling loop.
2969 (index_location_lists): Add comment on why skip_loc_list_entry can't
2972 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2975 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2978 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2981 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2982 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2983 and -mstring-compare-inline-limit.
2985 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2988 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2989 for stack clash protection in a register whenever we need it to be in
2992 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2994 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2995 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2997 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3000 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3002 (mlong-double-): Ditto.
3003 * config/rs6000/sysv4.opt (msdata=): Ditto.
3004 (mtls-size=): Ditto.
3006 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3008 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3009 erroneous entries for
3010 "vector int vec_ldl (int, long int *)", and
3011 "vector unsigned int vec_ldl (int, unsigned long int *)".
3012 Add comments and entries for
3013 "vector bool char vec_ldl (int, bool char *)",
3014 "vector bool short vec_ldl (int, bool short *)",
3015 "vector bool int vec_ldl (int, bool int *)",
3016 "vector bool long long vec_ldl (int, bool long long *)",
3017 "vector pixel vec_ldl (int, pixel *)",
3018 "vector long long vec_ldl (int, long long *)",
3019 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3020 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3021 type tree bool_long_long_type_node and correct definition of
3022 bool_V2DI_type_node to make reference to this new type tree.
3023 (rs6000_mangle_type): Replace erroneous reference to
3024 bool_long_type_node with bool_long_long_type_node.
3025 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3026 comments to emphasize sign distinctions for char and int types and
3027 replace RS6000_BTI_bool_long constant with
3028 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3029 use of RS6000_BTI_pixel.
3030 (bool_long_type_node): Remove this macro definition.
3031 (bool_long_long_type_node): New macro definition
3033 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3035 PR rtl-optimization/85300
3036 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3037 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3038 simplify_unary_operation fails.
3040 2018-04-10 Martin Liska <mliska@suse.cz>
3042 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3043 cgraph_edge and ipa_ref.
3045 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3049 * config/i386/sse.md
3050 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3051 computation of the VEC_MERGE selector from mask.
3052 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3053 Fix decoding of the VEC_MERGE selector into mask.
3055 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3057 PR tree-optimization/85286
3058 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3060 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3062 * final.c (final_1): Set insn_last_address as well as
3063 insn_current_address.
3065 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3068 * explow.c (emit_stack_probe): Call validize_mem on memory location
3069 before passing it to gen_probe_stack. Create address operand and
3070 legitimize it for the probe_stack_address case.
3072 2018-04-09 Jan Hubicka <jh@suse.cz>
3075 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3076 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3077 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3079 (free_lang_data): Rebuild type inheritance graph.
3081 2018-04-09 Martin Sebor <msebor@redhat.com>
3083 * invoke.texi (-finline-small-functions): Mention other optimization
3085 (-findirect-inlining, -fpartial-inlining): Same.
3086 (-finline-functions-called-once): Same.
3087 (-freorder-blocks-and-partition): Same.
3089 2018-04-09 Jan Hubicka <jh@suse.cz>
3092 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3093 jumps; choose last target that matches the criteria (i.e.
3094 no partition changes for non-crossing jumps).
3095 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3096 support for redirecting crossing jumps to non-crossing.
3098 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3100 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3101 also for naked functions.
3103 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3105 * config/arc/arc.md (add_shift): New pattern.
3106 (add_shift2): Likewise.
3107 (sub_shift): Likewise.
3108 (sub_shift_cmp0_noout): Likewise.
3109 (compare_si_ashiftsi): Likewise.
3110 (xbfu_cmp0_noout): New combine pattern.
3111 (xbfu_cmp0"): Likewise.
3112 (movsi_set_cc_insn): Place the predicable variant first.
3113 (commutative_binary_cmp0_noout): Remove clobber.
3114 (commutative_binary_cmp0): New pattern.
3115 (noncommutative_binary_cmp0): Likewise.
3116 (noncommutative_binary_cmp0_noout): Likewise.
3117 (noncommutative_binary_comparison_result_used): Removed.
3118 (rsub_cmp0): New pattern.
3119 (rsub_cmp0_noout): Likewise.
3120 (extzvsi): Changed, keep only meaningful variants.
3121 (SQH, SEZ): New iterators.
3122 (SQH_postfix): New mode attribute.
3123 (SEZ_prefix): New code attribute.
3124 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3125 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3126 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3128 (noncommutative_operator): Check the availability of barrel
3131 2018-04-09 Richard Biener <rguenther@suse.de>
3133 PR tree-optimization/85284
3134 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3135 Only use the niter constraining form of simple_iv when the exit
3138 2018-04-09 Tom de Vries <tom@codesourcery.com>
3141 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3142 (define_expand "*memory_barrier"): New define_expand.
3143 (define_insn "memory_barrier"): New insn.
3145 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3147 PR rtl-optimization/80463
3148 PR rtl-optimization/83972
3149 PR rtl-optimization/83480
3151 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3152 correct producer for the insn.
3153 (tidy_control_flow): Fixup seqnos in case of debug insns.
3155 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3157 PR rtl-optimization/83913
3159 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3160 different sched-times when merging exprs.
3162 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3164 PR rtl-optimization/83962
3166 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3167 tidy_fallthru_edge and tidy_control_flow.
3169 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3171 PR rtl-optimization/83530
3173 * sel-sched.c (force_next_insn): New global variable.
3174 (remove_insn_for_debug): When force_next_insn is true, also leave only
3175 next insn in the ready list.
3176 (sel_sched_region): When the region wasn't scheduled, make another pass
3177 over it with force_next_insn set to 1.
3179 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3181 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3183 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3184 for interrupt control.
3185 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3186 functions for interrupt control.
3187 * config/nds32/nds32-intrinsic.md: Likewise.
3188 * config/nds32/nds32_intrinsic.h: Likewise.
3189 * config/nds32/nds32.h (nds32_builtins): Likewise.
3191 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3193 * config/nds32/nds32.c (nds32_init_machine_status,
3194 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3195 strict_aligned_p field.
3196 (nds32_expand_to_rtl_hook): New function.
3197 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3198 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3200 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3201 Chung-Ju Wu <jasonwucj@gmail.com>
3203 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3204 * config/nds32/nds32-n7.md: New file.
3205 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3206 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3208 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3209 * config/nds32/nds32.md (pipeline_model): Add n7.
3210 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3211 * config/nds32/pipelines.md: Include n7 settings.
3213 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3214 Chung-Ju Wu <jasonwucj@gmail.com>
3216 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3217 * config/nds32/nds32-e8.md: New file.
3218 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3219 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3221 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3222 * config/nds32/nds32.md (pipeline_model): Add e8.
3223 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3224 * config/nds32/pipelines.md: Include e8 settings.
3226 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3227 Chung-Ju Wu <jasonwucj@gmail.com>
3229 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3230 * config/nds32/nds32-n8.md: New file.
3231 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3232 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3234 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3235 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3236 * config/nds32/nds32.md (pipeline_model): Add n8.
3237 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3238 * config/nds32/pipelines.md: Include n8 settings.
3240 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3241 Chung-Ju Wu <jasonwucj@gmail.com>
3243 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3244 * config/nds32/nds32-n9-2r1w.md: New file.
3245 * config/nds32/nds32-n9-3r2w.md: New file.
3246 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3247 nds32_register_ports): New or modify for cpu n9.
3248 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3250 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3251 * config/nds32/nds32-utils.c: New file.
3252 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3253 TARGET_MUL_SLOW): Define.
3254 * config/nds32/nds32.md (pipeline_model): New attribute.
3255 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3256 New options that support cpu n9.
3257 * config/nds32/pipelines.md: Include n9 settings.
3258 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3260 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3262 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3263 information if necessary.
3264 (output_cond_branch_compare_zero): Likewise.
3265 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3266 (nds32_target_alignment): Refine for alignment.
3267 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3268 (FUNCTION_BOUNDARY): Modify.
3269 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3271 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3273 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3275 * config/nds32/constants.md (unspec_volatile_element): Add values for
3276 TLB operation and data prefetch.
3277 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3278 functions for TLB operation and data prefetch.
3279 * config/nds32/nds32-intrinsic.md: Likewise.
3280 * config/nds32/nds32_intrinsic.h: Likewise.
3281 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3282 (nds32_print_operand): Likewise.
3283 * config/nds32/nds32.h (nds32_builtins): Likewise.
3285 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3286 Andrew Pinski <pinsika@gcc.gnu.org>
3289 * match.pd: Use constant_boolean_node of correct type instead of
3290 boolean_true_node or boolean_false_node for simplifying
3291 pointer comparisons to zero.
3293 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3295 PR tree-optimization/80021
3296 * tree.c (verify_type_variant): Make error call in verify_variant_match
3297 translatable and remove final full stop.
3299 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3301 * config/nds32/constants.md (unspec_volatile_element): Add
3302 UNSPEC_VOLATILE_EH_RETURN.
3303 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3304 nds32_output_stack_pop): Support dwarf exception handling process.
3305 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3306 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3307 exception handling process.
3308 (nds32_compute_stack_frame): Likewise.
3309 (nds32_return_addr_rtx): Likewise.
3310 (nds32_initial_elimination_offset): Likewise.
3311 (nds32_expand_prologue): Likewise.
3312 (nds32_expand_epilogue): Likewise.
3313 (nds32_dynamic_chain_address): New function.
3314 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3316 (DYNAMIC_CHAIN_ADDRESS): Define.
3317 (EH_RETURN_DATA_REGNO): Define.
3318 (EH_RETURN_STACKADJ_RTX): Define.
3319 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3320 patterns for dwarf exception handling.
3322 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3324 * config/nds32/nds32.h: Clean up obsolete macros.
3326 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3328 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3329 Add enum values for particular instructions.
3330 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3331 particular intrinsic functions.
3332 * config/nds32/nds32-intrinsic.md: Likewise.
3333 * config/nds32/nds32_intrinsic.h: Likewise.
3334 * config/nds32/nds32.h (nds32_builtins): Likewise.
3335 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3336 (btst, ave): New patterns for particular instructions.
3338 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3340 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3341 Add enum values for atomic load/store and memory sync.
3342 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3344 * config/nds32/nds32-intrinsic.md: Likewise.
3345 * config/nds32/nds32_intrinsic.h: Likewise.
3346 * config/nds32/nds32.h (nds32_builtins): Likewise.
3348 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3350 PR tree-optimization/85257
3351 * fold-const.c (native_encode_vector): If not all elts could fit
3352 and off is -1, return 0 rather than offset.
3353 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3354 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3355 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3356 adjust buffer in native_interpret_expr call.
3358 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3360 * config/nds32/constants.md (unspec_volatile_element): Add cache
3361 control enum values.
3362 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3363 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3364 * config/nds32/nds32.c (nds32_cctl_names): New.
3365 (nds32_print_operand): Handle cache control register names.
3366 * config/nds32/nds32.h (nds32_builtins): New enum values.
3367 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3369 * config/nds32/nds32.md (type): Add mmu.
3370 * config/nds32/pipelines.md (simple_insn): Add mmu.
3372 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3374 * config/nds32/nds32.md (type): Remove call.
3375 * config/nds32/pipelines.md (simple_insn): Likewise.
3377 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3379 * config/nds32/constants.md (unspec_volatile_element): Add
3380 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3381 UNSPEC_VOLATILE_FMFCFG.
3382 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3383 description for fmfcfg and fmfcsr.
3384 (bdesc_1arg): Add fmtcsr.
3385 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3386 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3387 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3388 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3389 unspec_fmfcfg): New patterns.
3390 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3391 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3392 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3393 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3394 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3395 __nds32__fmfcfg): Define.
3397 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3399 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3400 intrinsic register names.
3401 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3402 intrinsic register enum values and macros.
3404 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3406 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3407 for load/store addressing form.
3408 (nds32_print_operand_address): Likewise.
3410 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3413 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3414 based on LABEL_REF. Remove useless assertion.
3415 (pic_address_needs_scratch): Fix formatting.
3416 (sparc_legitimize_pic_address): Minor tweaks.
3417 (sparc_delegitimize_address): Adjust assertion accordingly.
3418 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3419 into symbolic_operand.
3420 (movsi_high_pic_label_ref): Likewise.
3421 (movsi_lo_sum_pic_label_ref): Likewise.
3422 (movdi_pic_label_ref): Likewise.
3423 (movdi_high_pic_label_ref): Likewise.
3424 (movdi_lo_sum_pic_label_ref): Likewise.
3426 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3428 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3429 custom LIB_SPEC setup.
3431 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3432 Kito Cheng <kito.cheng@gmail.com>
3434 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3435 * config/riscv/freebsd.h: New.
3437 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3439 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3440 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3443 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3444 Kito Cheng <kito.cheng@gmail.com>
3446 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3447 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3448 * config/nds32/nds32-protos.h (nds32_output_call,
3449 nds32_output_return): Declare.
3450 * config/nds32/nds32.md: Refine all the call and return patterns.
3452 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3455 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3456 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3458 PR rtl-optimization/84872
3459 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3460 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3463 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3465 * expr.c (copy_blkmode_to_reg): Revert 254862.
3466 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3468 2018-04-06 Richard Biener <rguenther@suse.de>
3471 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3472 after seeing a component reference with an adjacent field. Treat
3473 refs to arrays at struct end of external decls similar to
3474 refs to unconstrained commons.
3476 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3479 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3480 look through SAVE_EXPRs with non-side-effects argument. Adjust
3482 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3485 2018-04-06 Richard Biener <rguenther@suse.de>
3488 * alias.c (find_base_term): New wrapper around find_base_term
3489 unwinding CSELIB_VAL_PTR changes.
3490 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3493 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3495 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3497 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3498 constant definitions.
3499 ("nop"): lr 0,0 -> nopr r0
3500 ("nop_lr0", "nop_lr1"): New insn definitions.
3502 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3504 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3505 NDS32_V3PUSH_AVAILABLE_P macro.
3507 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3508 Chung-Ju Wu <jasonwucj@gmail.com>
3510 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3511 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3512 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3513 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3514 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3515 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3516 * config/nds32/constraints.md: New constraints and checking for hard
3517 float configuration.
3518 * config/nds32/iterators.md: New mode iterator and attribute for hard
3519 float configuration.
3520 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3522 * config/nds32/nds32-fpu.md: New file.
3523 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3524 deal with hard float code generation.
3525 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3527 (abi_type, float_reg_number): New enum type.
3528 * config/nds32/nds32-predicates.c: New predicates for hard float.
3529 * config/nds32/nds32-protos.h: Declare functions for hard float.
3530 * config/nds32/nds32.c: Implementation for hard float configuration.
3531 * config/nds32/nds32.h: Definitions for hard float configuration.
3532 * config/nds32/nds32.md: Include hard float machine description and
3533 modify patterns for hard float configuration.
3534 * config/nds32/nds32.opt: New options for hard float configuration.
3535 * config/nds32/predicates.md: New predicates for hard float
3538 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3540 * common/config/nds32/nds32-common.c
3541 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3543 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3546 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3547 CONSTRUCTOR_ELT (ctor, ...)->value.
3549 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3552 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3554 2018-04-05 Tom de Vries <tom@codesourcery.com>
3557 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3560 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3561 Kito Cheng <kito.cheng@gmail.com>
3563 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3564 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3565 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3566 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3568 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3569 Kito Cheng <kito.cheng@gmail.com>
3571 * config/nds32/constraints.md (Ufe): New memory constraint.
3572 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3573 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3574 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3576 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3577 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3579 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3581 * config/nds32/nds32.md: Use optimize_size in the condition for
3582 alu-shift instructions.
3584 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3586 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3588 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3590 * config/nds32/nds32.md (negsi2): Refine pattern.
3592 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3593 Chung-Ju Wu <jasonwucj@gmail.com>
3595 * config/nds32/iterators.md (shift_rotate): New code iterator.
3596 (shift): New code attribute.
3597 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3598 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3599 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3600 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3601 bit-wise operations.
3602 (andsi3, *andsi3): Ditto.
3603 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3604 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3605 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3606 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3607 nds32_ior_operand, nds32_xor_operand): New predicates.
3609 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3611 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3612 (addsi3, subsi3): ... this.
3614 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3616 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3618 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3620 * config/nds32/nds32.md: Adjust indention.
3622 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3624 * config/nds32/nds32.md (feature): New attribute.
3626 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3628 * config/nds32/nds32.md (subtype): New attribute.
3630 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3633 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3634 expansion to perform a bitwise AND of the argument followed by a
3635 boolean negation of the result.
3637 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3639 PR rtl-optimization/84878
3640 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3641 the basic block. Assert the use reference is not artificial and that
3642 it has an associated insn.
3644 2018-04-04 Michael Matz <matz@suse.de>
3646 * builtins.c (compute_objsize): Pass correct operand
3647 to array_at_struct_end_p.
3649 2018-04-04 Richard Biener <rguenther@suse.de>
3652 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3653 from contexts for DINFO_LEVEL_TERSE and below.
3655 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3657 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3658 resiter_operand condition.
3659 * config/nds32/nds32.md (*move<mode>): Ditto.
3661 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3662 Monk Chiang <sh.chiang04@gmail.com>
3664 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3666 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3668 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3670 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3671 Kito Cheng <kito.cheng@gmail.com>
3673 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3674 nds32_cond_code_str, output_cond_branch,
3675 output_cond_branch_compare_zero, nds32_expand_cbranch,
3676 nds32_expand_cstore, nds32_expand_movcc,
3677 nds32_output_cbranchsi4_equality_zero,
3678 nds32_output_cbranchsi4_equality_reg,
3679 nds32_output_cbranchsi4_equality_reg_or_const_int,
3680 nds32_output_cbranchsi4_greater_less_zero: New functions.
3681 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3682 nds32_expand_cstore, nds32_expand_movcc,
3683 nds32_output_cbranchsi4_equality_zero,
3684 nds32_output_cbranchsi4_equality_reg,
3685 nds32_output_cbranchsi4_equality_reg_or_const_int,
3686 nds32_output_cbranchsi4_greater_less_zero): Declare.
3687 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3688 nds32_rimm11s_operand): New predicates.
3689 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3690 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3693 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3695 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3696 * config/nds32/nds32.md: Ditto.
3697 * config/nds32/pipelines.md: Ditto.
3699 2018-04-04 Richard Biener <rguenther@suse.de>
3701 PR tree-optimization/85168
3702 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3703 propagating abnormals.
3705 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3707 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3709 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3710 Kito Cheng <kito.cheng@gmail.com>
3712 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3713 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3714 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3715 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3716 * config/nds32/nds32.md (sibcall_internal): New.
3717 (sibcall_register): Remove.
3718 (sibcall_immediate): Remove.
3719 (sibcall_value_internal): New.
3720 (sibcall_value_register): Remove.
3721 (sibcall_value_immediate): Remove.
3722 * config/nds32/predicates.md (nds32_general_register_operand): New.
3723 (nds32_call_address_operand): New.
3725 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3727 PR rtl-optimization/85167
3728 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3729 bb_defs if *split_p, instead preinitialize it to NULL.
3731 PR tree-optimization/85156
3732 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3733 evaluating the argument multiple times.
3735 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3737 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3739 (_mm_cvtpd_ps): Likewise.
3740 (_mm_cvttpd_epi32): Likewise.
3741 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3742 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3743 vector, pixel, and bool following altivec.h include.
3745 2018-04-03 Martin Sebor <msebor@redhat.com>
3747 * doc/extend.texi (Common Function Attributes): Clarify.
3748 (const attribute): Likewise.
3749 (pure attribute): Likewise.
3751 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3754 * config/i386/i386.c (ix86_expand_vector_set): Use
3755 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3757 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3759 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3760 instructions when changing rounding bits to preserve precision bits
3761 in the x87 control word.
3763 2018-04-03 Martin Liska <mliska@suse.cz>
3765 PR tree-optimization/82491
3766 * rtl.h (strip_offset_and_add): Replace += suboffset with
3767 poly_uint64 () + suboffset.
3769 2018-03-29 Martin Liska <mliska@suse.cz>
3770 Martin Jambor <mjambor@suse.cz>
3773 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3774 param_type is not an integral or pointer type.
3776 2018-04-03 Richard Biener <rguenther@suse.de>
3778 * sese.h (recompute_all_dominators): Remove.
3780 2018-04-02 Martin Sebor <msebor@redhat.com>
3782 * doc/invoke.texi (-Wrestrict): Fix typos.
3784 2018-04-02 Jim Wilson <jimw@sifive.com>
3786 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3787 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3788 (<optab>di3, <optab>si3_extend): Likewise.
3789 (<optab>si3_mask, <optab>si3_mask_1): New.
3790 (<optab>di3_mask, <optab>di3_mask_1): New.
3791 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3792 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3793 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3795 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3797 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3800 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3802 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3803 (nds32_canonicalize_comparison): New function.
3805 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3806 Kito Cheng <kito.cheng@gmail.com>
3807 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3809 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3810 * config/nds32/constants.md (unspec_volatile_element): Add
3811 UNSPEC_VOLATILE_RELAX_GROUP.
3812 * config/nds32/nds32-relax-opt.c: New file.
3813 * config/nds32/nds32-predicates.c
3814 (nds32_symbol_load_store_p): New function.
3815 * config/nds32/nds32-protos.h
3816 (nds32_symbol_load_store_p): Declare function.
3817 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3818 * config/nds32/nds32.c
3819 (nds32_register_pass): New function to register pass.
3820 (nds32_register_passes): New function to register passes.
3821 * config/nds32/nds32.md (relax_group): New pattern.
3822 * config/nds32/nds32.opt (mrelax-hint): New option.
3823 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3825 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3827 * config/nds32/t-nds32: Modify files dependency.
3829 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3831 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3832 (PROFILE_HOOK): Define its implementation.
3834 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3836 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3837 type and 32-bit size.
3839 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3842 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3843 (V_128_256): New mode iterator.
3844 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3845 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3846 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3848 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3849 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3851 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3854 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3855 NaN inputs correctly.
3857 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3860 * config/rs6000/vsx.md (??r): New mode attribute.
3861 (*vsx_mov<mode>_64bit): Use it.
3862 (*vsx_mov<mode>_32bit): Likewise.
3864 2018-03-30 Martin Sebor <msebor@redhat.com>
3866 PR tree-optimization/84818
3867 * builtins.c (check_access): Use warning_n.
3869 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3872 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3874 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3877 2018-03-30 Julia Koval <julia.koval@intel.com>
3880 * x86-tune.def (movx, partial_reg_dependency): Enable for
3883 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3886 * lra-constraints.c (process_alt_operands): Move setting
3887 this_alternative_matches below.
3889 2018-03-29 Martin Liska <mliska@suse.cz>
3892 * doc/invoke.texi: Document how LTO works with debug info.
3893 Describe auto-load support of binutils. Mention 'x86-64'
3894 as valid option value of -march option.
3896 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3898 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3901 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3902 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3903 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3906 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3909 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3910 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3911 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3912 for RS6000_BTM_POWERPC64.
3913 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3914 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3915 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3920 2018-03-28 Carl Love <cel@us.ibm.com>
3923 2017-09-27 Carl Love <cel@us.ibm.com>
3925 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3926 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3927 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3930 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3932 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3933 instead of __vector bool.
3934 (_mm_max_pu8): Likewise.
3935 (_mm_min_pi16): Likewise.
3937 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3940 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3941 (DIVWEUO): Likewise.
3943 (DIVDEUO): Likewise.
3944 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3945 DIVWEUO and DIVDEUO.
3946 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3947 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3948 (div_extend): Likewise.
3949 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3951 (__builtin_divweuo): Likewise.
3952 (__builtin_divdeo): Likewise.
3953 (__builtin_divdeuo): Likewise.
3955 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3958 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3959 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3961 PR tree-optimization/82004
3962 * gimple-match-head.c (optimize_pow_to_exp): New function.
3963 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3964 Don't fold to exp if optimize_pow_to_exp is false.
3966 2018-03-28 Martin Liska <mliska@suse.cz>
3969 * calls.c (initialize_argument_information): Fix trailing space.
3970 * common.opt: Fix typo and provide better explanation for
3971 -fsanitize-coverage option.
3972 * config/i386/i386.opt: Fix typo.
3974 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3975 Martin Liska <mliska@suse.cz>
3978 * gimplify.c (asan_poison_variable): Don't do the check for
3979 gimplify_omp_ctxp here.
3980 (gimplify_decl_expr): Do it here.
3981 (gimplify_target_expr): Likewise.
3983 2018-03-28 Martin Liska <mliska@suse.cz>
3986 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3987 chkp_type_bounds_count if MPX is not enabled.
3989 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3991 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3993 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3996 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3997 function to create the function decl for complex long double
3998 multiply and divide for -mabi=ieeelongdouble.
3999 (init_float128_ieee): Call it.
4001 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4004 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4005 -fcf-protection=branch -mibt.
4006 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4008 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4011 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4013 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4016 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4017 extern array declarations.
4019 2018-03-27 Richard Biener <rguenther@suse.de>
4022 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4023 explicit single_use checks.
4025 2018-03-27 Richard Biener <rguenther@suse.de>
4027 PR tree-optimization/85082
4028 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4031 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4033 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4034 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4035 Turn on fasynchronous-unwind-tables and funwind-tables.
4037 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4040 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4041 (*bmi_blsr_<mode>_ccz): Ditto.
4043 2018-03-26 Tom de Vries <tom@codesourcery.com>
4045 PR tree-optimization/85063
4046 * omp-general.c (offloading_function_p): New function. Factor out
4048 * omp-offload.c (pass_omp_target_link::gate): ... here.
4049 * omp-general.h (offloading_function_p): Declare.
4050 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4051 with attribute omp declare target for offloading functions.
4053 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4055 PR tree-optimization/84005
4056 * tree-data-ref.h (get_base_for_alignment): Declare.
4057 * tree-data-ref.c (get_base_for_alignment_1): New function.
4058 (get_base_for_alignment): Likewise.
4059 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4060 get_base_for_alignment to find a suitable base object, instead
4061 of always using drb->base_address.
4063 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4066 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4067 known size by default.
4069 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4072 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4073 and non BLKmode operands.
4075 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4078 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4079 Clean up attributes.
4081 2018-03-23 Richard Biener <rguenther@suse.de>
4084 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4085 we are going to emit early debug for LTO.
4087 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4090 * function.c (match_asm_constraints_1): Don't optimize if input
4091 doesn't satisfy general_operand predicate for output's mode.
4094 * alias.c (write_dependence_p): Don't require for x_canonicalized
4095 non-VOIDmode if x has VOIDmode.
4098 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4099 just don't try to optimize it rather than assert it never happens.
4101 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4103 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4104 macro expansions for definition of ST_INTERNAL_<mode> and
4105 LD_INTERNAL_<mode> builtins.
4106 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4108 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4110 (altivec_expand_st_builtin): Likewise.
4111 (altivec_expand_builtin): Remove calls to deleted functions.
4112 (rs6000_address_for_altivec): Delete this function.
4113 * config/rs6000/vector.md: Remove expands for
4114 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4116 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4119 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4120 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4121 re-computing once computed.
4122 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4123 (arm_init_machine_status): Initialize
4124 machine->static_chain_stack_bytes.
4126 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4129 * doc/extend.texi: Add four new prototypes for vec_ld.
4130 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4131 definitions for more logical presentation.
4132 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4133 entries for V1TI variants of __builtin_altivec_ld builtin.
4134 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4135 handling of V1TI variant of LVX icode pattern.
4136 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4137 (rs6000_gimple_fold_builtin): Likewise.
4138 (altivec_init_builtins): Add code to define
4139 __builtin_altivec_lvx_v1ti function.
4141 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4144 * function.c (match_asm_constraints_1): Don't do the optimization
4145 if input isn't a REG, SUBREG, MEM or constant.
4147 2018-03-22 Tom de Vries <tom@codesourcery.com>
4149 PR tree-optimization/84956
4150 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4151 bb_has_abnormal_pred.
4153 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4156 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4157 DECL_INITIAL (decl) to decl at the end.
4158 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4161 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4163 * doc/extend.texi (__builtin_tgmath): Document when complex
4164 integer types are treated as _Complex _Float64.
4166 2018-03-21 Tom de Vries <tom@codesourcery.com>
4168 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4170 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4172 PR tree-optimization/84960
4173 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4174 if it is ENTRY block, move them into single succ of ENTRY in that case.
4176 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4178 PR tree-optimization/84811
4179 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4180 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4181 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4182 (known_subrange_p): Update accordingly. Cast each value involved
4183 in the size comparison, rather than casting the result of the
4186 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4188 PR tree-optimization/84982
4189 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4190 by flipping the least significant bit rather than all bits from
4191 bitpos to bitpos + bitsize - 1.
4193 2018-03-21 Nathan Sidwell <nathan@acm.org>
4195 * doc/extend.texi (Deprecated Features): Remove mention of
4196 long-deleted deprecations.
4198 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4201 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4202 * configure: Regenerate.
4204 2018-03-21 Tom de Vries <tom@codesourcery.com>
4206 PR tree-optimization/83126
4207 * tree-parloops.c (num_phis): New function.
4208 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4210 2018-03-21 Nathan Sidwell <nathan@acm.org>
4212 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4213 mention anon-struct/union members and trailing attributes.
4215 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4217 PR tree-optimization/84969
4218 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4219 builtin memset partitions if they set different rhs values.
4221 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4223 PR rtl-optimization/84989
4224 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4225 VEC_DUPLICATE with scalar result mode.
4227 2018-03-21 Martin Liska <mliska@suse.cz>
4230 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4231 not intended return statement.
4233 2018-03-21 Martin Liska <mliska@suse.cz>
4236 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4237 (chkp_find_bound_slots_1): Limit number of iterations.
4239 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4242 * Minor grammar fixes for x86 options.
4244 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4247 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4248 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4250 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4253 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4254 (altivec_lvx_<mode>_1op): Likewise.
4255 (altivec_stvx_<mode>_2op): Likewise.
4256 (altivec_stvx_<mode>_1op): Likewise.
4257 (altivec_lvx_<VM2:mode>): New define_expand.
4258 (altivec_stvx_<VM2:mode>): Likewise.
4259 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4260 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4261 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4262 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4263 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4264 (rs6000_gen_lvx): Likewise.
4265 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4266 (altivec_expand_stv_builtin): Likewise.
4267 (altivec_expand_builtin): Likewise.
4268 * config/rs6000/vector.md: Likewise.
4270 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4273 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4276 2018-03-20 Richard Biener <rguenther@suse.de>
4279 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4280 sign-conversions as zero, fall back to standard scalar_stmt
4283 2018-03-20 Martin Liska <mliska@suse.cz>
4286 * predict.c (rebuild_frequencies): Handle case when we have
4287 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4289 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4292 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4293 flag_section_anchors.
4294 * varasm.c (use_blocks_for_decl_p): Remove hack for
4295 dw2_force_const_mem.
4298 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4300 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4301 be created, use lowpart_subreg of operands[0] rather than operands[0]
4303 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4304 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4305 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4306 and n constraint instead of aarch64_shift_imm_di and Usd.
4307 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4308 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4310 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4313 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4314 to favor GPR over NEON registers.
4315 (<shift>di3_neon): Likewise.
4317 2018-03-20 Tom de Vries <tom@codesourcery.com>
4320 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4321 (nvptx_process_pars): Emit bar.sync asap and alap.
4323 2018-03-20 Tom de Vries <tom@codesourcery.com>
4326 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4327 seen_label if seen_label is already set.
4329 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4332 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4333 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4334 Use 1U instead of 1. Formatting fixes.
4337 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4338 instead of TREE_TYPE (s1) for the return value.
4340 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4342 PR tree-optimization/84946
4343 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4344 bitsize + bitsize in poly_uint64 rather than poly_int64.
4347 * dwarf2asm.c: Include fold-const.c.
4348 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4349 of decl rather than decl itself.
4351 PR rtl-optimization/84643
4352 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4354 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4357 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4358 calling assemble_variable.
4360 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4363 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4364 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4366 2018-03-19 Jim Wilson <jimw@sifive.com>
4369 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4370 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4371 (riscv_first_stack_step): Likewise.
4372 (riscv_option_override): Use STACK_BOUNDARY instead of
4374 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4376 (BIGGEST_ALIGNMENT): Set to 128.
4377 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4378 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4381 2018-03-19 Richard Biener <rguenther@suse.de>
4383 PR tree-optimization/84933
4384 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4385 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4387 2018-03-19 Richard Biener <rguenther@suse.de>
4389 PR tree-optimization/84859
4390 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4391 (cond_if_else_store_replacement): Perform sinking operation on
4392 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4393 Generalize what a BB with a single eligible store is.
4395 2018-03-19 Richard Biener <rguenther@suse.de>
4397 PR tree-optimization/84929
4398 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4399 chrec_is_positive against non-chrec arg.
4401 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4404 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4406 2018-03-18 Martin Liska <mliska@suse.cz>
4408 PR rtl-optimization/84635
4409 * regrename.c (build_def_use): Use matches_mode only when
4412 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4414 PR tree-optimization/84913
4415 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4416 vectorize chains of COND_EXPRs.
4418 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4420 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4422 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4424 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4426 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4428 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4430 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4431 Kito Cheng <kito.cheng@gmail.com>
4433 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4434 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4435 (nds32_adjust_reg_alloc_order): New function.
4436 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4438 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4440 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4441 nds32_print_operand, nds32_print_operand_address): Use
4442 HOST_WIDE_INT_PRINT_DEC instead.
4444 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4446 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4448 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4451 * config/i386/i386.c (initial_ix86_tune_features,
4452 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4454 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4455 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4456 rather than 1u << ix86_tune. Formatting fix.
4457 (ix86_option_override_internal): Change ix86_arch_mask from
4458 unsigned int to unsigned HOST_WIDE_INT, initialize to
4459 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4460 (ix86_function_specific_restore): Likewise.
4462 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4465 * postreload.c (reload_combine_recognize_pattern): Perform
4466 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4467 truncate_int_for_mode the result for the destination's mode.
4470 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4471 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4473 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4476 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4479 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4480 regno_allocno_class_array and sorted_pseudos.
4481 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4482 insns where regno is used.
4484 2018-03-16 Martin Liska <mliska@suse.cz>
4487 * multiple_target.c (create_dispatcher_calls): Redirect
4488 reference in the symbol table.
4490 2018-03-16 Martin Liska <mliska@suse.cz>
4493 * multiple_target.c (create_dispatcher_calls): Redirect also
4496 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4500 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4501 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4502 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4504 2018-03-16 Julia Koval <julia.koval@intel.com>
4506 * doc/invoke.texi (Skylake Server): Add CLWB.
4507 Cannonlake): Remove CLWB.
4509 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4511 PR tree-optimization/84841
4512 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4514 (FLOAT_ONE_CONST_TYPE): Define.
4515 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4516 (sort_by_operand_rank): Put entries with higher constant_type last
4517 rather than first to match comments.
4519 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4521 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4524 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4527 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4528 check and use address of target always.
4530 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4533 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4534 (indirect_thunk_bnd_needed): Likewise.
4535 (indirect_thunks_used): Likewise.
4536 (indirect_thunks_bnd_used): Likewise.
4537 (indirect_return_needed): New.
4538 (indirect_return_bnd_needed): Likewise.
4539 (output_indirect_thunk_function): Add a bool argument for
4541 (output_indirect_thunk_function): Don't generate alias for
4542 function return thunk.
4543 (ix86_code_end): Call output_indirect_thunk_function to generate
4544 function return thunks.
4545 (ix86_output_function_return): Set indirect_return_bnd_needed
4546 and indirect_return_needed instead of indirect_thunk_bnd_needed
4547 and indirect_thunk_needed.
4549 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4551 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4552 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4553 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4555 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4556 Paul Hua <paul.hua.gm@gmail.com>
4559 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4561 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4563 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4564 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4567 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4570 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4571 instead of GET_MODE_SIZE when comparing Units.
4573 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4576 * varasm.c (hash_section): Return an unchangeble hash value
4577 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4578 Return !aarch64_can_use_per_function_literal_pools_p ().
4580 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4583 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4584 rather than address of cmode as last argument to prepare_cmp_insn.
4586 2018-03-15 Julia Koval <julia.koval@intel.com>
4588 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4589 F_AVX512VNNI, F_AVX512BITALG): New.
4591 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4594 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4595 insn for floating-point loads and stores.
4597 2018-03-14 Carl Love <cel@us.ibm.com>
4599 * config/rs6000/rs6000-c.c: Add macro definitions for
4600 ALTIVEC_BUILTIN_VEC_PERMXOR.
4601 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4602 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4603 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4604 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4606 * config/doc/extend.texi: Add prototypes for vec_permxor.
4608 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4611 * diagnostic-show-locus.c (class layout_point): Convert m_line
4612 from int to linenum_type.
4613 (line_span::comparator): Use linenum "compare" function when
4614 comparing line numbers.
4615 (test_line_span): New function.
4616 (layout_range::contains_point): Convert param "row" from int to
4618 (layout_range::intersects_line_p): Likewise.
4619 (layout::will_show_line_p): Likewise.
4620 (layout::print_source_line): Likewise.
4621 (layout::should_print_annotation_line_p): Likewise.
4622 (layout::print_annotation_line): Likewise.
4623 (layout::print_leading_fixits): Likewise.
4624 (layout::annotation_line_showed_range_p): Likewise.
4625 (struct line_corrections): Likewise for field m_row.
4626 (line_corrections::line_corrections): Likewise for param "row".
4627 (layout::print_trailing_fixits): Likewise.
4628 (layout::get_state_at_point): Likewise.
4629 (layout::get_x_bound_for_row): Likewise.
4630 (layout::print_line): Likewise.
4631 (diagnostic_show_locus): Likewise for locals "last_line" and
4633 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4634 * input.c (selftest::test_linenum_comparisons): New function.
4635 (selftest::input_c_tests): Call it.
4636 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4637 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4638 * selftest.h (ASSERT_GT): New macro.
4639 (ASSERT_GT_AT): New macro.
4640 (ASSERT_LT): New macro.
4641 (ASSERT_LT_AT): New macro.
4643 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4645 PR rtl-optimization/84780
4646 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4648 2018-03-14 Martin Liska <mliska@suse.cz>
4650 * tree.c (record_node_allocation_statistics): Use
4651 get_stats_node_kind.
4652 (get_stats_node_kind): New function extracted from
4653 record_node_allocation_statistics.
4654 (free_node): Use get_stats_node_kind.
4656 2018-03-14 Richard Biener <rguenther@suse.de>
4658 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4659 that the value-set of ANTIC_IN doesn't grow.
4662 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4664 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4665 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4667 2018-03-14 Julia Koval <julia.koval@intel.com>
4669 * config.gcc (icelake-client, icelake-server): New.
4671 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4672 (initial_ix86_arch_features): Ditto.
4673 (PTA_SKYLAKE): Add SGX.
4674 (PTA_ICELAKE): Remove.
4675 (PTA_ICELAKE_CLIENT): New.
4676 (PTA_ICELAKE_SERVER): New.
4677 (ix86_option_override_internal): Split up icelake on icelake client and
4679 (get_builtin_code_for_version): Ditto.
4680 (fold_builtin_cpu): Ditto.
4681 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4682 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4683 * config/i386/i386.h (processor_type): Ditto.
4684 * doc/invoke.texi: Ditto.
4686 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4689 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4690 INTEGER_CST offset, add it together with bitpos / 8 and
4691 sign extend based on POINTER_SIZE.
4695 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4698 * config/i386/constraints.md (Yc): New register constraint.
4699 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4700 Use Yc constraint for alternative 2 of operand 0. Remove
4701 preferred_for_speed attribute.
4703 2018-03-14 Richard Biener <rguenther@suse.de>
4705 PR tree-optimization/84830
4706 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4707 with the old one to avoid oscillations.
4709 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4712 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4714 (assign_by_spills): Return a flag of reload assignment failure.
4715 Do not process the reload assignment failures. Do not spill other
4716 reload pseudos if they has the same reg class. Update n if
4718 (lra_assign): Add a return arg. Set up from the result of
4719 assign_by_spills call.
4720 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4721 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4722 usage_insns if it is not NULL.
4723 (spill_hard_reg_in_range): New function.
4724 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4725 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4726 function prototypes.
4727 (lra_assign): Change prototype.
4728 * lra.c (lra): Add code to deal with fails by splitting hard reg
4731 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4733 * config/riscv/riscv.opt (mrelax): New option.
4734 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4735 "norelax" when riscv_mrelax is disabled.
4736 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4738 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4741 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4742 reassociation for int modes.
4744 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4746 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4747 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4749 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4750 * config/aarch64/aarch64-sve.md
4751 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4752 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4753 (*extend<mode><Vwide>2): Rename to...
4754 (aarch64_sve_extend<mode><Vwide>2): ...this.
4755 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4756 renaming the old pattern to...
4757 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4759 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4760 define_expand, renaming the old pattern to...
4761 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4762 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4763 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4764 account when deciding which SVE instruction the optab should use.
4765 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4767 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4769 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4770 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4771 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4772 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4773 (tlsdesc_small_<mode>): Turn a define_expand and use
4774 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4775 (tlsdesc_small_advsimd_<mode>): ...this.
4776 (tlsdesc_small_sve_<mode>): New pattern.
4778 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4780 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4781 (UNSPEC_UMUL_HIGHPART): New constants.
4782 (MUL_HIGHPART): New int iteraor.
4783 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4784 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4786 (*<su>mul<mode>3_highpart): New define_insn.
4788 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4791 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4794 2018-03-13 Martin Liska <mliska@suse.cz>
4797 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4799 (sem_item_optimizer::~sem_item_optimizer): Release it.
4800 (sem_item_optimizer::merge_classes): Register variable aliases.
4801 (sem_item_optimizer::fixup_pt_set): New function.
4802 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4803 * ipa-icf.h: Declare new variables and functions.
4805 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4808 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4809 integer_pow2p@2 and test integer_pow2p in condition.
4810 (A < 0 ? C : 0): Similarly for @1.
4813 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4814 characters starting at p contain '\0' character, don't look beyond
4818 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4819 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4822 * reg-stack.c (change_stack): Change update_end var from int to
4823 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4824 also call set_block_for_insn on the newly added insns and rescan.
4827 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4828 on the last operand.
4831 * tree.c (stabilize_reference_1): Return save_expr (e) for
4832 STATEMENT_LIST even if it doesn't have side-effects.
4834 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4836 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4838 2018-03-12 Renlin Li <renlin.li@arm.com>
4840 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4841 aarch64_output_scalar_simd_mov_immediate.
4843 2018-03-12 Martin Sebor <msebor@redhat.com>
4845 PR tree-optimization/83456
4846 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4847 for perfectly overlapping calls to memcpy.
4848 (gimple_fold_builtin_memory_chk): Same.
4849 (gimple_fold_builtin_strcpy): Handle no-warning.
4850 (gimple_fold_builtin_stxcpy_chk): Same.
4851 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4853 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4855 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4856 parameter. Use it for SFmode.
4857 (rs6000_function_arg_advance_1): Adjust.
4858 (rs6000_function_arg): Adjust.
4859 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4861 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4863 PR rtl-optimization/84169
4864 PR rtl-optimization/84780
4865 * combine.c (can_combine_p): Check for a 2-insn combination whether
4866 the destination register is used between the two insns, too.
4868 2018-03-12 Richard Biener <rguenther@suse.de>
4870 PR tree-optimization/84803
4871 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4872 for refs DR analysis didn't process.
4874 2018-03-12 Richard Biener <rguenther@suse.de>
4876 PR tree-optimization/84777
4877 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4878 force-vectorize loops ignore whether we are optimizing for size.
4880 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4882 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4883 (TARGET_MD_ASM_ADJUST): Define.
4885 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4886 Kito Cheng <kito.cheng@gmail.com>
4887 Chung-Ju Wu <jasonwucj@gmail.com>
4889 * config/nds32/nds32.c (nds32_compute_stack_frame,
4890 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4891 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4892 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4893 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4894 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4895 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4896 * config/nds32/nds32.md (prologue, epilogue): Use macro
4897 NDS32_V3PUSH_AVAILABLE_P to do checking.
4899 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4902 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4903 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4904 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4905 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4906 addition of most attributes on !orig_type_die or the attribute not
4907 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4909 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4910 Chung-Ju Wu <jasonwucj@gmail.com>
4912 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4914 * config/nds32/nds32.opt (mvh): New option.
4916 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4917 Chung-Ju Wu <jasonwucj@gmail.com>
4919 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4921 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4922 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4925 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4926 Chung-Ju Wu <jasonwucj@gmail.com>
4928 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4930 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4931 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4933 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4934 Kito Cheng <kito.cheng@gmail.com>
4935 Chung-Ju Wu <jasonwucj@gmail.com>
4937 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4938 UNSPEC_FFMISM and UNSPEC_FLMISM.
4939 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4940 for ffb, ffmism and flmism.
4941 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4942 (unspec_ffmism): Ditto.
4943 (unspec_flmism): Ditto.
4944 (nds32_expand_builtin_impl): Check if string extension is available.
4945 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4946 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4948 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4951 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4954 * lra-assigns.c (assign_by_spills): Return a flag of reload
4955 assignment failure. Do not process the reload assignment
4956 failures. Do not spill other reload pseudos if they has the same
4958 (lra_assign): Add a return arg. Set up from the result of
4959 assign_by_spills call.
4960 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4961 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4962 usage_insns if it is not NULL.
4963 (spill_hard_reg_in_range): New function.
4964 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4965 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4966 function prototypes.
4967 (lra_assign): Change prototype.
4968 * lra.c (lra): Add code to deal with fails by splitting hard reg
4971 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4974 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4976 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4979 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4980 (dw_val_node): Add val_symbolic_view.
4981 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4982 (symview_upper_bound): New.
4983 (new_line_info_table): Initialize symviews_since_reset.
4984 (dwarf2out_source_line): Count symviews_since_reset and set
4985 symview_upper_bound.
4986 (dw_val_equal_p): Handle symview.
4987 (add_AT_symview): New.
4988 (print_dw_val): Handle symview.
4989 (attr_checksum, attr_checksum_ordered): Likewise.
4990 (same_dw_val_p, size_of_die): Likewise.
4991 (value_format, output_die): Likewise.
4992 (add_high_low_attributes): Use add_AT_symview for entry_view.
4993 (dwarf2out_finish): Reset symview_upper_bound, clear
4996 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4999 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5000 Add strict argument and use it.
5001 (rs6000_split_multireg_move): Update for new strict argument.
5002 (mem_operand_gpr): Disallow all non-offsettable addresses.
5003 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5005 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5008 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5009 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5010 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5013 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5014 decl, use remap_type if we want to use the type.
5016 2018-03-09 Martin Sebor <msebor@redhat.com>
5018 PR tree-optimization/84526
5019 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5021 (builtin_access::generic_overlap): Be prepared to handle non-array
5024 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5026 PR rtl-optimization/84682
5027 * lra-constraints.c (process_address_1): Check is_address flag
5028 for address constraints.
5029 (process_alt_operands): Likewise.
5030 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5031 preprocess_constraints.
5032 * recog.h (preprocess_constraints): Add oploc parameter.
5034 * recog.c (preprocess_constraints): Test address_operand for
5035 CT_ADDRESS constraints.
5037 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5040 * lra-assigns.c (assign_by_spills): Return a flag of reload
5041 assignment failure. Do not process the reload assignment
5042 failures. Do not spill other reload pseudos if they has the same
5044 (lra_assign): Add a return arg. Set up from the result of
5045 assign_by_spills call.
5046 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5047 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5048 usage_insns if it is not NULL.
5049 (spill_hard_reg_in_range): New function.
5050 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5051 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5052 function prototypes.
5053 (lra_assign): Change prototype.
5054 * lra.c (lra): Add code to deal with fails by splitting hard reg
5057 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5060 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5061 Accept complain bool parameter. Only emit errors if it is true.
5062 (arm_parse_cpu_option_name): Likewise.
5063 (arm_target_thumb_only): Adjust callers of the above.
5064 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5065 prototype to take a default true bool parameter.
5066 (arm_parse_arch_option_name): Likewise.
5068 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5069 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5073 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5074 * configure: Regenerate.
5075 * configure.ac ("linker --version-script option"): New.
5076 ("linker soname option"): New.
5078 2018-03-09 Richard Biener <rguenther@suse.de>
5080 PR tree-optimization/84775
5081 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5082 immediate uses of predicate stmts and mark them modified.
5085 PR tree-optimization/84178
5086 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5088 (version_loop_for_if_conversion): Delay update_ssa call.
5089 (tree_if_conversion): Delay update_ssa until after predicate
5092 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5095 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5096 when the function accesses prior frames.
5098 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5101 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5102 gen_llsym, otherwise call maybe_gen_llsym.
5105 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5106 has ',' character inside of it.
5108 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5111 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5112 as clobbering CC_REGNUM.
5114 2018-03-08 Richard Biener <rguenther@suse.de>
5117 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5118 (follow_copies_to_constant): Do not follow SSA names registered
5121 2018-03-08 Richard Biener <rguenther@suse.de>
5123 PR tree-optimization/84178
5124 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5126 (version_loop_for_if_conversion): Delay update_ssa call.
5127 (tree_if_conversion): Delay update_ssa until after predicate
5130 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5132 PR tree-optimization/84178
5133 * tree-if-conv.c (release_bb_predicate): Remove the
5134 the assertion that the stmts have NULL use_ops.
5135 Discard the statements, asserting that they haven't
5136 yet been added to a BB.
5138 2018-03-08 Richard Biener <rguenther@suse.de>
5140 PR tree-optimization/84746
5141 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5142 (phi_translate): Pass in destination ANTIC_OUT set.
5143 (phi_translate_1): Likewise. For a simplified result lookup
5144 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5145 (phi_translate_set): Adjust.
5146 (do_pre_regular_insertion): Likewise.
5147 (do_pre_partial_partial_insertion): Likewise.
5149 2018-03-08 Martin Liska <mliska@suse.cz>
5151 PR gcov-profile/84735
5152 * doc/gcov.texi: Document usage of profile files.
5153 * gcov-io.h: Document changes in the format.
5155 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5159 * dwarf2out.c (struct dw_line_info_table): Update comments for
5161 (FORCE_RESET_NEXT_VIEW): New.
5162 (FORCE_RESETTING_VIEW_P): New.
5163 (RESETTING_VIEW_P): Check for -1 too.
5164 (ZERO_VIEW_P): Likewise.
5165 (new_line_info_table): Force-reset next view.
5166 (dwarf2out_begin_function): Likewise.
5167 (dwarf2out_source_line): Simplify zero_view_p initialization.
5168 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5169 view directly. Omit view when omitting .loc at line 0.
5171 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5173 PR tree-optimization/84740
5174 * tree-switch-conversion.c (process_switch): Call build_constructors
5175 only if info.phi_count is non-zero.
5177 PR tree-optimization/84739
5178 * tree-tailcall.c (find_tail_calls): Check call arguments against
5179 DECL_ARGUMENTS (current_function_decl) rather than
5180 DECL_ARGUMENTS (func) when checking for tail recursion.
5182 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5184 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5185 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5186 Volker Reichelt's entry and add entries for people that perform
5187 GCC fuzzy testing and report numerous bugs.
5189 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5192 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5193 readonly data in sdata, if that is disabled.
5194 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5195 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5196 -mreadonly-in-sdata option.
5198 2018-03-07 Martin Sebor <msebor@redhat.com>
5200 PR tree-optimization/84468
5201 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5202 basic block when looking for nul assignment.
5204 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5207 * except.h (output_function_exception_table): Adjust prototype.
5208 * except.c (output_function_exception_table): Remove FNNAME parameter
5209 and add SECTION parameter. Ouput one part of the table at a time.
5210 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5211 the first part of the exception table and emit unwind directives.
5212 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5213 (i386_pe_seh_cold_init): Likewise.
5214 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5215 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5216 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5217 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5218 * config/i386/winnt.c: Include except.h.
5219 (struct seh_frame_state): Add reg_offset, after_prologue and
5220 in_cold_section fields.
5221 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5222 (i386_pe_seh_cold_init): New function.
5223 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5224 to seh->in_cold_section.
5225 (seh_emit_push): Record the offset of the push.
5226 (seh_emit_save): Record the offet of the save.
5227 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5228 Test seh->after_prologue to disregard the epilogue.
5229 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5230 (i386_pe_end_cold_function): New function.
5232 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5235 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5236 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5239 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5241 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5242 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5245 * multiple_target.c: Include tree-inline.h and intl.h.
5246 (expand_target_clones): Diagnose and fail if node->definition and
5247 !tree_versionable_function_p (node->decl).
5249 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5251 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5253 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5254 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5255 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5257 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5260 * combine.c (try_combine): Use reg_or_subregno instead of handling
5261 just paradoxical SUBREGs and REGs.
5263 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5265 * config/arc/arc.c (arc_finalize_pic): Remove function.
5266 (arc_must_save_register): We use single base PIC register, remove
5267 checks to save/restore the PIC register.
5268 (arc_expand_prologue): Likewise.
5269 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5271 (arc_verify_short): Likewise.
5272 (arc_attr_type): Likewise.
5273 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5274 (walk_stores): Likewise.
5275 (arc_address_cost): Make it static.
5276 (arc_verify_short): Likewise.
5277 (branch_dest): Likewise.
5278 (arc_attr_type): Likewise.
5279 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5280 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5281 (arc_final_prescan_insn): Remove inserting the nops due to
5282 hardware hazards. It is done in reorg step.
5283 (insn_length_variant_t): Remove.
5284 (insn_length_parameters_t): Likewise.
5285 (arc_insn_length_parameters): Likewise.
5286 (arc_get_insn_variants): Likewise.
5287 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5289 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5292 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5295 PR tree-optimization/84687
5296 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5298 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5300 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5302 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5303 Rename to ppc_speculation_barrier.
5304 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5305 __builtin_ppc_speculation_barrier.
5307 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5310 * combine.c (combine_simplify_rtx): Don't try to simplify if
5311 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5314 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5316 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5317 to 32 bytes when compiling for POWER9.
5319 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5322 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5323 regparm >= 3 with no arg reg available also for calls with
5324 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5327 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5329 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5331 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5334 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5336 2018-03-05 Richard Biener <rguenther@suse.de>
5338 PR tree-optimization/84486
5339 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5340 When inserting a __builtin_assume_aligned call set the LHS
5341 SSA name alignment info accordingly.
5343 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5345 PR tree-optimization/84114
5346 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5347 Avoid reassociation of FLOAT_MODE addition.
5349 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5351 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5352 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5353 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5354 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5355 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5356 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5357 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5359 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5360 __builtin_ia32_wbinvd): New builtins.
5361 (SPECIAL_ARGS2): New.
5362 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5363 (SPECIAL_ARGS2): New.
5364 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5365 (ix86_valid_target_attribute_inner_p): Ditto.
5366 (ix86_init_mmx_sse_builtins): Add special_args2.
5367 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5368 TARGET_WBNOINVD_P): New.
5369 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5370 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5371 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5372 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5373 * config/i386/pconfigintrin.h: New file.
5374 * config/i386/wbnoinvdintrin.h: Ditto.
5375 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5376 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5378 2018-03-05 Richard Biener <rguenther@suse.de>
5380 PR tree-optimization/84670
5381 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5383 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5384 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5385 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5386 doesn't grow if all successors (recursively) were visited at least
5389 2018-03-05 Richard Biener <rguenther@suse.de>
5391 PR tree-optimization/84650
5392 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5393 if executed in the loop pipeline.
5395 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5397 * doc/configfiles.texi (Configuration Files): Move info about
5398 conditionalizing $target-protos.h to...
5399 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5400 differs from $target-protos.h.
5402 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5403 Chung-Ju Wu <jasonwucj@gmail.com>
5405 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5406 * config/nds32/nds32-multiple.md (setmemsi): Define.
5407 * config/nds32/nds32-memory-manipulation.c
5408 (nds32_gen_dup_4_byte_to_word_value): New.
5409 (emit_setmem_word_loop): New.
5410 (emit_setmem_byte_loop): New.
5411 (nds32_expand_setmem_loop): New.
5412 (nds32_expand_setmem_loop_v3m): New.
5413 (nds32_expand_setmem_unroll): New.
5414 (nds32_expand_setmem): New.
5416 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5417 Chung-Ju Wu <jasonwucj@gmail.com>
5419 * config/nds32/nds32-memory-manipulation.c
5420 (nds32_emit_load_store): New.
5421 (nds32_emit_post_inc_load_store): New.
5422 (nds32_emit_mem_move): New.
5423 (nds32_emit_mem_move_block): New.
5424 (nds32_expand_movmemsi_loop_unknown_size): New.
5425 (nds32_expand_movmemsi_loop_known_size): New.
5426 (nds32_expand_movmemsi_loop): New.
5427 (nds32_expand_movmemsi_unroll): New.
5428 (nds32_expand_movmemqi): Rename ...
5429 (nds32_expand_movmemsi): ... to this.
5430 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5431 (movmemsi): ... to this.
5432 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5433 (nds32_expand_movmemsi): ... to this.
5435 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5436 Monk Chiang <sh.chiang04@gmail.com>
5437 Chung-Ju Wu <jasonwucj@gmail.com>
5439 * config/nds32/nds32-protos.h
5440 (nds32_expand_load_multiple): New arguments.
5441 (nds32_expand_store_multiple): Ditto.
5442 (nds32_valid_multiple_load_store): Rename ...
5443 (nds32_valid_multiple_load_store_p): ... to this.
5444 * config/nds32/nds32-memory-manipulation.c
5445 (nds32_expand_load_multiple): Refine implementation.
5446 (nds32_expand_store_multiple): Ditto.
5447 * config/nds32/nds32-multiple.md
5448 (load_multiple): Update nds32_expand_load_multiple interface.
5449 (store_multiple): Update nds32_expand_store_multiple interface.
5450 * config/nds32/nds32-predicates.c
5451 (nds32_valid_multiple_load_store): Rename ...
5452 (nds32_valid_multiple_load_store_p): ... to this and refine
5454 * config/nds32/predicates.md
5455 (nds32_load_multiple_and_update_address_operation): New predicate.
5456 (nds32_store_multiple_and_update_address_operation): New predicate.
5458 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5459 Chung-Ju Wu <jasonwucj@gmail.com>
5461 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5462 (combo): New attribute.
5463 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5465 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5467 * config/nds32/nds32.opt: Change -mcmodel= default value.
5469 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5470 Monk Chiang <sh.chiang04@gmail.com>
5471 Chung-Ju Wu <jasonwucj@gmail.com>
5473 * config/nds32/constants.md (unspec_element): New enum.
5474 * config/nds32/constraints.md (Umw): New constraint.
5475 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5476 * config/nds32/nds32-intrinsic.md: Likewise.
5477 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5478 (nds32_valid_smw_lwm_base_p): New.
5479 (nds32_output_smw_single_word): New.
5480 (nds32_output_lmw_single_word): New.
5481 (nds32_expand_unaligned_load): New.
5482 (nds32_expand_unaligned_store): New.
5483 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5484 (nds32_output_smw_single_word): Declare.
5485 (nds32_output_lmw_single_word): Declare.
5486 (nds32_expand_unaligned_load): Declare.
5487 (nds32_expand_unaligned_store): Declare.
5488 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5489 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5490 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5491 NDS32_BUILTIN_UASTORE_DW.
5492 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5495 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5496 Kito Cheng <kito.cheng@gmail.com>
5497 Chung-Ju Wu <jasonwucj@gmail.com>
5499 * config/nds32/nds32-intrinsic.c
5500 (nds32_expand_builtin_null_ftype_reg): Delete.
5501 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5502 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5503 (nds32_read_argument): New.
5504 (nds32_legitimize_target): Ditto.
5505 (nds32_legitimize_argument): Ditto.
5506 (nds32_check_constant_argument): Ditto.
5507 (nds32_expand_unop_builtin): Ditto.
5508 (nds32_expand_unopimm_builtin): Ditto.
5509 (nds32_expand_binop_builtin): Ditto.
5510 (nds32_builtin_decl_impl): Ditto.
5511 (builtin_description): Ditto.
5512 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5513 (nds32_init_builtins_impl): Ditto.
5514 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5515 (nds32_builtin_decl): New.
5516 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5517 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5519 2018-03-02 Jeff Law <law@redhat.com>
5521 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5522 (redundant_insn, fill_simple_delay_slots): Likewise.
5523 (fill_slots_from_thread): Likewise.
5524 * resource.c (mark_referenced_resources): Likewise.
5525 (mark_set_resources, find_dead_or_set_registers): Likewise.
5527 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5529 * substring-locations.h (format_warning_va): Formatting fix for
5531 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5533 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5534 * substring-locations.c: Include intl.h.
5535 (format_warning_va): Turned into small wrapper around
5536 format_warning_n_va, renamed to ...
5537 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5538 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5540 (format_warning_at_substring_n): New function.
5541 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5542 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5543 format_warning_at_substring with just a shorter name instead of
5544 const function pointer.
5545 (fmtwarn_n): New function.
5546 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5547 appropriate, get rid of all the fmtstr temporaries, move conditionals
5548 with G_() wrapped string literals directly into fmtwarn arguments,
5549 cast dir.len to (int), formatting fixes.
5551 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5553 * doc/invoke.texi: Remove "Cilk Plus" references.
5555 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5556 Richard Biener <rguenther@suse.de>
5559 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5560 for error or warning attributes if CALL_FROM_THUNK_P is set.
5563 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5566 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5567 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5570 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5571 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5573 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5574 instead of -1U in last predictors element's probability member.
5576 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5579 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5580 arguments if they are comparable.
5582 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5584 PR tree-optimization/84634
5585 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5586 masks and masked_loop_p with a single loop_masks, making sure it's
5587 null for bb vectorization.
5589 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5591 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5592 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5593 loop->force_vectorize to check whether there is no alias.
5595 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5598 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5600 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5602 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5603 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5604 instead of a loop around prev_real_insn.
5605 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5609 * config/i386/i386.c (ix86_print_operand): Use conditional
5610 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5613 2018-03-02 Richard Biener <rguenther@suse.de>
5615 PR tree-optimization/84427
5616 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5617 (bitmap_set_subtract_values): Rewrite to handle multiple
5620 (prune_clobbered_mems): Likewise.
5621 (phi_translate): Take edge instead of pred/phiblock.
5622 (phi_translate_1): Likewise.
5623 (phi_translate_set): Likewise. Insert all translated
5624 exprs for a value into the set, keeping possibly multiple
5625 expressions per value.
5626 (compute_antic_aux): Adjust for phi_translate changes.
5627 When intersecting union the expressions and prune those
5628 not in the final value set, keeping possibly multiple
5629 expressions per value. Do not use value-insertion
5630 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5631 all expressions. Add verification that the value-sets
5632 only shrink during iteration.
5633 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5634 (do_pre_regular_insertion): Likewise.
5635 (do_pre_partial_partial_insertion): Likewise.
5637 2018-03-02 Richard Biener <rguenther@suse.de>
5640 * config/darwin.c (saved_debug_info_level): New static global.
5641 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5642 (darwin_asm_lto_end): Restore debug info generation settings.
5644 2018-03-01 Martin Liska <mliska@suse.cz>
5647 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5650 2018-03-01 Richard Biener <rguenther@suse.de>
5653 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5654 type annotation with LTO when debug was disabled at compile-time.
5656 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5658 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5660 (mips_final_postscan_insn): Likewise.
5662 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5664 PR rtl-optimization/84528
5665 * alias.c (init_alias_target): Add commentary.
5666 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5667 a unique base value if the frame pointer is not eliminated
5668 to the stack pointer.
5670 2018-03-01 Tom de Vries <tom@codesourcery.com>
5672 PR rtl-optimization/83327
5673 * lra-int.h (hard_regs_spilled_into): Declare.
5674 * lra.c (hard_regs_spilled_into): Define.
5675 (init_reg_info): Init hard_regs_spilled_into.
5676 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5677 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5678 (process_bb_lives): Handle hard_regs_spilled_into.
5679 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5680 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5682 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5684 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5685 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5686 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5687 * config/rs6000/aix72.h: New file.
5689 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5691 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5692 instead of warning_at with conditional singular and plural messages
5696 * stor-layout.c (update_alignment_for_field): For
5697 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5698 && !DECL_PACKED (field), do the alignment update, just use
5699 only desired_align instead of MAX (type_align, desired_align)
5701 (place_field): Don't do known_align < desired_align handling
5702 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5703 is non-NULL, instead do it after rli->prev_field handling and
5704 only if not within a bitfield word. For DECL_PACKED (field)
5705 use type_align of BITS_PER_UNIT.
5707 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5709 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5710 superfluous parentheses and trailing spaces.
5712 2018-02-28 Richard Biener <rguenther@suse.de>
5714 PR tree-optimization/84584
5715 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5716 SCoPs with fake exit edge.
5718 2018-02-28 Martin Liska <mliska@suse.cz>
5721 * timevar.c (timer::print): Fix format to properly print 100%
5724 2018-02-28 Richard Biener <rguenther@suse.de>
5727 * genmatch.c (capture_info::walk_match): Do not mark
5728 captured expressions without operands as expr_p given
5729 they act more like predicates and should be subject to
5730 "lost tail" side-effect preserving.
5732 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5734 PR rtl-optimization/81611
5735 * auto-inc-dec.c (attempt_change): Move dead note from
5736 mem_insn if it's the next use of regno
5737 (find_address): Take address use of reg holding
5738 non-incremented value. Add parm to limit search to the named
5740 (merge_in_block): Attempt to use a mem insn that is the next
5741 use of the original regno.
5743 2018-02-27 Martin Sebor <msebor@redhat.com>
5746 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5747 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5749 2018-02-27 Martin Sebor <msebor@redhat.com>
5751 PR translation/84207
5752 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5753 n argument to unsigned HOST_WIDE_INT.
5754 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5755 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5756 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5757 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5759 2018-02-27 Richard Biener <rguenther@suse.de>
5761 PR tree-optimization/84512
5762 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5763 Do not use the estimate returned from record_stmt_cost for
5764 the scalar iteration cost but sum properly using add_stmt_cost.
5766 2018-02-27 Richard Biener <rguenther@suse.de>
5768 PR tree-optimization/84466
5769 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5770 Adjust last change to less strictly validate use operands.
5772 2018-02-27 Martin Liska <mliska@suse.cz>
5774 PR gcov-profile/84548
5775 * gcov.c (process_file): Allow partial overlap and consider it
5776 also as group functions.
5777 (output_lines): Properly calculate range of lines for a group.
5779 2018-02-27 Martin Liska <mliska@suse.cz>
5781 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5782 'ggc' suffixes. Change first column width.
5783 (timer::print): Fix formatting of the column.
5785 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5787 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5788 preserve inline entry blocks for the sake of debug inline
5789 entry point markers alone.
5790 (remove_unused_locals): Suggest in comments a better place to
5791 force the preservation of inline entry blocks that are
5792 otherwise unused, but do not preserve them.
5794 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5796 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5798 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5801 * config/i386/constraints.md (Bs): Replace
5802 ix86_indirect_branch_register with
5803 TARGET_INDIRECT_BRANCH_REGISTER.
5805 * config/i386/i386.md (indirect_jump): Likewise.
5806 (tablejump): Likewise.
5807 (*sibcall_memory): Likewise.
5808 (*sibcall_value_memory): Likewise.
5809 Peepholes of indirect call and jump via memory: Likewise.
5810 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5811 (*sibcall_value_GOT_32): Likewise.
5812 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5813 (GOT_memory_operand): Likewise.
5814 (call_insn_operand): Likewise.
5815 (sibcall_insn_operand): Likewise.
5816 (GOT32_symbol_operand): Likewise.
5817 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5819 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5821 PR rtl-optimization/83496
5822 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5823 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5824 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5825 redundant insn, if any.
5826 (relax_delay_slots): Likewise.
5827 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5829 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5831 PR tree-optimization/83965
5832 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5833 that grouped statements are part of a reduction chain. Return
5834 true if the statement is not marked as a reduction itself but
5836 (vect_recog_dot_prod_pattern): Don't check whether the statement
5837 is part of a group here.
5838 (vect_recog_sad_pattern): Likewise.
5839 (vect_recog_widen_sum_pattern): Likewise.
5841 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5844 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5846 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5849 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5851 (ix86_output_indirect_function_return): New prototype.
5852 (ix86_split_simple_return_pop_internal): Likewise.
5853 * config/i386/i386.c (indirect_return_via_cx): New.
5854 (indirect_return_via_cx_bnd): Likewise.
5855 (indirect_thunk_name): Handle return va CX_REG.
5856 (output_indirect_thunk_function): Create alias for
5857 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5858 (ix86_output_indirect_jmp): Remove the bool argument.
5859 (ix86_output_indirect_function_return): New function.
5860 (ix86_split_simple_return_pop_internal): Likewise.
5861 * config/i386/i386.md (*indirect_jump): Don't pass false
5862 to ix86_output_indirect_jmp.
5863 (*tablejump_1): Likewise.
5864 (simple_return_pop_internal): Change it to define_insn_and_split.
5865 Call ix86_split_simple_return_pop_internal to split it for
5867 (simple_return_indirect_internal): Call
5868 ix86_output_indirect_function_return instead of
5869 ix86_output_indirect_jmp.
5871 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5874 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5875 memset and value initialization afterwards.
5877 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5879 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5881 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5884 * common/config/aarch64/aarch64-common.c
5885 (aarch_option_optimization_table[]): Switch
5886 off fomit-frame-pointer
5888 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5889 Chung-Ju Wu <jasonwucj@gmail.com>
5891 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5893 (store_multiple): Ditto.
5895 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5897 * config.gcc: Add --with-cpu support for nds32 target.
5898 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5899 * config/nds32/nds32.opt: Add -mcpu= option.
5901 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5903 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5904 isel=yes): Warn for these deprecated options.
5906 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5908 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5909 ISA_2_5_MASKS_EMBEDDED.
5911 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5913 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5914 p->max as pointers rather than using iterative_hash_expr.
5916 2018-02-23 Carl Love <cel@us.ibm.com>
5918 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5919 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5921 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5922 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5923 P8V_BUILTIN_VEC_VUNSIGNED2.
5925 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5928 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5929 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5931 * lra-constraints.c (curr_insn_transform): Set up
5932 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5934 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5935 * lra-eliminations.c (spill_pseudos): Ditto.
5936 (process_insn_for_elimination): Ditto.
5937 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5938 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5939 LRA_NON_CLOBBERED_ALT.
5941 2018-02-22 Martin Sebor <msebor@redhat.com>
5943 PR tree-optimization/84480
5944 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5945 to maybe_diag_stxncpy_trunc. Call it.
5946 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5947 from gimple_fold_builtin_strcpy. Print inlining stack.
5948 (handle_builtin_stxncpy): Print inlining stack.
5949 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5951 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5954 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5955 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5956 and -fcheck-pointer-bounds are used together.
5957 (indirect_thunk_prefix): New enum.
5958 (indirect_thunk_need_prefix): New function.
5959 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5960 "_nt" instead of "_bnd" for NOTRACK prefix.
5961 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5962 (output_indirect_thunk_function): Likewise.
5964 (ix86_code_end): Update output_indirect_thunk_function calls.
5965 (ix86_output_indirect_branch_via_reg): Replace
5966 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5967 (ix86_output_indirect_branch_via_push): Likewise.
5968 (ix86_output_function_return): Likewise.
5969 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5970 incompatible with -fcf-protection=branch and
5971 -fcheck-pointer-bounds.
5973 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5976 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5977 Change gcc_assert call to output_operand_lossage.
5979 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5981 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5983 2018-02-22 DJ Delorie <dj@redhat.com>
5984 Sebastian Perta <sebastian.perta@renesas.com>
5985 Oleg Endo <olegendo@gcc.gnu.org>
5987 * config/rx/rx.c (rx_rtx_costs): New function.
5988 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5990 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5992 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5994 2018-02-22 Martin Liska <mliska@suse.cz>
5997 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5998 Add "native" as a possible value.
6000 2018-02-22 Martin Liska <mliska@suse.cz>
6003 * config/i386/i386.c (ix86_option_override_internal):
6004 Add "native" as a possible value for -march and -mtune.
6006 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6009 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6010 to all type variants.
6012 PR tree-optimization/84503
6013 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6014 width as info->bitpos + info->bitsize - start.
6015 (merged_store_group::merge_overlapping): Simplify width computation.
6016 (check_no_overlap): New function.
6017 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6018 start + width and last_order of the group, fail if check_no_overlap
6020 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6021 to group if check_no_overlap fails.
6023 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6025 * config/rs6000/altivec.md: Delete contraint arguments to
6026 define_expand, define_split, and define_peephole2, and in
6027 define_insn_and_split if always unused.
6028 * config/rs6000/darwin.md: Ditto.
6029 * config/rs6000/dfp.md: Ditto.
6030 * config/rs6000/rs6000.md: Ditto.
6031 * config/rs6000/sync.md: Ditto.
6032 * config/rs6000/vector.md: Ditto.
6033 * config/rs6000/vsx.md: Ditto.
6035 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6037 * config/rs6000/altivec.md: Write output control strings as braced
6038 blocks instead of double-quoted strings.
6039 * config/rs6000/darwin.md: Ditto.
6040 * config/rs6000/rs6000.md: Ditto.
6041 * config/rs6000/vector.md: Ditto.
6042 * config/rs6000/vsx.md: Ditto.
6044 2018-02-21 Jason Merrill <jason@redhat.com>
6046 PR c++/84314 - ICE with templates and fastcall attribute.
6047 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6049 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6051 * ipa-cp.c (determine_versionability): Fix comment typos.
6053 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6056 * ipa-cp.c (determine_versionability): Do not version functions caling
6059 2018-02-21 Martin Liska <mliska@suse.cz>
6062 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6063 Add "native" as a possible value.
6064 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6065 the macro when native cpu detection is available.
6067 2018-02-21 Martin Liska <mliska@suse.cz>
6070 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6071 Add "native" as a possible value.
6072 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6073 when native cpu detection is available.
6075 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6076 Martin Sebor <msebor@redhat.com>
6078 PR tree-optimization/84478
6079 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6081 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6082 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6083 support which is conservatively correct, for 2 only stay conservative
6084 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6085 argument to the 2 argument get_range_strlen, adjust 6 arg
6086 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6088 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6089 (gimple_fold_builtin_strlen): Pass true as last argument to
6092 2018-02-20 Martin Sebor <msebor@redhat.com>
6095 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6096 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6097 (builtin_memref::builtin_memref): Factor out parts into
6098 set_base_and_offset and call it.
6100 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6103 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6104 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6105 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6106 search at the associated MODE_INT.
6108 2018-02-20 Jeff Law <law@redhat.com>
6111 PR tree-optimization/81592
6113 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6114 for range data rather than using global data.
6115 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6116 range data rather than using global data.
6117 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6118 pass it to children as needed.
6119 (struct directive::fmtresult): Similarly.
6120 (struct directive::set_width): Similarly.
6121 (struct directive::set_precision): Similarly.
6122 (format_integer, format_directive, parse_directive): Similarly.
6123 (format_none): Accept unnamed vr_values parameter.
6124 (format_percent, format_floating, format_character): Similarly.
6125 (format_string, format_plain): Similarly.
6126 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6127 the EVRP range analyzer for range data rather than using global data.
6128 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6129 gimple-ssa-evrp-analyze.h
6130 (class sprintf_dom_walker): Add after_dom_children member function.
6131 Add evrp_range_analyzer member.
6132 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6133 range analyzer as needed.
6134 (sprintf_dom_walker::after_dom_children): New member function.
6135 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6137 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6138 (evrp_range_analyzer::pop_to_marker): Likewise.
6140 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6142 PR tree-optimization/84419
6143 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6144 with the required type if its current type is compatible but
6147 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6150 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6151 after vectorization.
6153 2018-02-20 Martin Liska <mliska@suse.cz>
6156 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6157 possible values if we don't have a hint.
6159 2018-02-20 Martin Liska <mliska@suse.cz>
6163 * final.c (shorten_branches): Build align_tab array with one
6165 * opts.c (finish_options): Add alignment option limit check.
6166 (MAX_CODE_ALIGN): Likewise.
6167 (MAX_CODE_ALIGN_VALUE): Likewise.
6168 * doc/invoke.texi: Document maximum allowed option value for
6169 all -falign-* options.
6171 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6174 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6175 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6176 * var-tracking.c (emit_note_insn_var_location): Remove all references
6177 to NOTE_INSN_CALL_ARG_LOCATION.
6178 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6179 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6180 Use copy_rtx_if_shared.
6181 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6182 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6183 (dwarf2out_var_location): Remove handling of
6184 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6186 * final.c (final_scan_insn): Remove all references to
6187 NOTE_INSN_CALL_ARG_LOCATION.
6188 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6189 before dumping final insns.
6190 * except.c (emit_note_eh_region_end): Remove all references to
6191 NOTE_INSN_CALL_ARG_LOCATION.
6192 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6193 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6194 * config/arc/arc.c (hwloop_optimize): Likewise.
6195 * config/arm/arm.c (create_fix_barrier): Likewise.
6196 * config/s390/s390.c (s390_chunkify_start): Likewise.
6197 * config/sh/sh.c (find_barrier): Likewise.
6198 * config/i386/i386.c (rest_of_insert_endbranch,
6199 ix86_seh_fixup_eh_fallthru): Likewise.
6200 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6201 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6202 * config/frv/frv.c (frv_function_prologue): Likewise.
6203 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6205 (note_outside_basic_block_p): Remove all references to
6206 NOTE_INSN_CALL_ARG_LOCATION.
6207 * gengtype.c (adjust_field_rtx_def): Likewise.
6208 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6210 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6211 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6214 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6217 PR tree-optimization/84452
6218 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6219 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6222 2018-02-19 Martin Liska <mliska@suse.cz>
6225 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6227 2018-02-19 Martin Liska <mliska@suse.cz>
6228 Richard Sandiford <richard.sandiford@linaro.org>
6230 PR tree-optimization/82491
6231 * gimple-fold.c (get_base_constructor): Make earlier bail out
6234 2018-02-19 Carl Love <cel@us.ibm.com>
6236 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6237 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6239 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6240 P8V_BUILTIN_VEC_NEG.
6242 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6244 * config/rl78/rl78.md (movdf): New define expand.
6246 2018-02-19 Martin Liska <mliska@suse.cz>
6249 * doc/invoke.texi: Fix typo.
6250 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6252 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6254 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6255 handle rs6000_single_float and rs6000_double_float specially for
6258 2018-02-16 Jeff Law <law@redhat.com>
6260 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6263 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6266 * ipa-inline.c (inline_small_functions): Fix a typo.
6268 2018-02-16 Nathan Sidwell <nathan@acm.org>
6270 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6272 2018-02-16 Carl Love <cel@us.ibm.com>
6274 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6275 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6276 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6277 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6278 expansion to P8V_BUILTIN_VEC_FLOAT2.
6280 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6282 PR rtl-optimization/70023
6283 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6284 src_regno into account.
6286 2018-02-16 Carl Love <cel@us.ibm.com>
6288 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6289 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6290 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6291 * config/rs6000/rs6000.c: Remove case statements for
6292 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6293 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6294 and P9V_BUILTIN_VEC_VINSERT4B.
6295 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6296 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6297 * config/rs6000/vsx.md:
6298 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6301 2018-02-16 Carl Love <cel@us.ibm.com>
6303 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6305 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6307 * config/rs6000/rs6000-c.c: Add the definitions for
6308 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6309 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6310 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6311 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6312 definition for insert4b and define insn *insert3b_internal.
6313 * doc/extend.texi: Add documentation for vec_extract4b.
6315 2018-02-16 Nathan Sidwell <nathan@acm.org>
6317 * doc/extend.texi (Backwards Compatibility): Mention friend
6318 injection. Note for-scope is deprecated.
6319 * doc/invoke.texi (-ffriend-injection): Deprecate.
6321 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6323 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6324 that moved to I2, also allow destinations that are a paradoxical
6325 subreg (instead of a normal reg).
6327 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6330 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6333 2018-02-16 Richard Biener <rguenther@suse.de>
6335 PR tree-optimization/84037
6336 PR tree-optimization/84016
6338 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6339 Adjust vec_construct for the fact we need additional higher latency
6340 128bit inserts for AVX256 and AVX512 vector builds.
6341 (ix86_add_stmt_cost): Scale vector construction cost for
6344 2018-02-16 Richard Biener <rguenther@suse.de>
6346 PR tree-optimization/84417
6347 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6348 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6349 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6351 2018-02-16 Martin Liska <mliska@suse.cz>
6354 * internal-fn.def (ASAN_CHECK): Set proper flags.
6355 (ASAN_MARK): Likewise.
6357 2018-02-16 Julia Koval <julia.koval@intel.com>
6359 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6360 from PTA_CANNONLAKE.
6362 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6365 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6366 Use ++iter rather than iter++ for std::list iterators.
6367 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6368 defer deleting them until all nodes in the forest are processed. Do
6369 free even leaf nodes. Change to_process into auto_vec.
6372 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6373 * vec.h (vec_default_construct): Use memset instead of placement new
6374 if BROKEN_VALUE_INITIALIZATION is defined.
6375 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6376 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6379 PR rtl-optimization/83723
6380 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6381 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6382 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6384 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6386 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6388 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6390 PR rtl-optimization/81443
6391 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6392 from inner REGs to paradoxical SUBREGs.
6394 2018-02-16 Richard Biener <rguenther@suse.de>
6396 PR tree-optimization/84399
6397 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6398 For operands we can analyze at their definition make sure we can
6399 analyze them at each use as well.
6401 2018-02-16 Richard Biener <rguenther@suse.de>
6403 PR tree-optimization/84190
6404 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6405 volatile accesses if the decl isn't volatile.
6407 2018-02-15 Jason Merrill <jason@redhat.com>
6409 PR c++/84314 - ICE with templates and fastcall attribute.
6410 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6411 TYPE_CANONICAL on an existing type.
6413 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6415 PR tree-optimization/84383
6416 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6417 dstoff nor call operand_equal_p if dstbase is NULL.
6419 PR tree-optimization/84334
6420 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6421 also a CONSTANT_CLASS_P, punt.
6423 2018-02-14 Jim Wilson <jimw@sifive.com>
6425 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6426 first SMALL_OPERAND check. New local min_second_step. Move assert
6427 to where locals are set. Add TARGET_RVC support.
6428 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6430 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6432 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6434 2018-02-14 Martin Sebor <msebor@redhat.com>
6436 PR tree-optimization/83698
6437 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6438 arrays constrain the offset range to their bounds.
6439 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6440 (builtin_access::overlap): Avoid setting the size of overlap if it's
6442 (maybe_diag_overlap): Also consider arrays when deciding what values
6443 of offsets to include in diagnostics.
6445 2018-02-14 Martin Sebor <msebor@redhat.com>
6448 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6449 that correspond to the kind of a declaration.
6451 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6454 * config/pa/pa.md: Load address of PIC label using the linkage table
6455 if the label is nonlocal.
6457 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6459 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6460 warning message if user requests -maltivec=be.
6461 * doc/invoke.texi: Document deprecation of -maltivec=be.
6463 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6466 * config/rs6000/rs6000-c.c: Update definitions for
6467 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6468 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6470 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6473 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6474 add _get_ssp intrinsics. Remove argument from
6475 __builtin_ia32_rdssp[d|q].
6476 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6477 * config/i386/i386-builtin.def: Remove argument from
6478 __builtin_ia32_rdssp[d|q].
6479 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6480 ix86_expand_special_args_builtin for _rdssp[d|q].
6481 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6482 Clear register before usage.
6483 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6484 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6486 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6488 PR tree-optimization/84357
6489 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6490 operand 1 of an ARRAY_REF too.
6492 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6495 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6496 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6498 (set_of_reg): New struct.
6499 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6500 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6501 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6503 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6504 Split into bitclr, bitset, bitinvert patterns if appropriate.
6505 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6506 use rx_fuse_in_memory_bitop.
6507 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6508 to named insn, correct maximum insn length.
6510 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6513 * machmode.def: Define a complex mode for PARTIAL_INT.
6514 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6516 * doc/rtl.texi: Document CSPImode.
6517 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6519 (msp430_hard_regno_nregs_with_padding): Likewise.
6521 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6524 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6526 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6528 PR rtl-optimization/84169
6529 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6530 we generated a parallel as new i3 and we split that to new i2 and i3
6531 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6532 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6533 those to i2, not i1. Partially rewrite this scan code.
6535 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6538 * stor-layout.c (place_field): For variable length fields, adjust
6539 offset_align afterwards not just based on the field's alignment,
6540 but also on the size.
6543 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6544 of exps and logs in the use_exp2 case.
6546 2018-02-13 Jeff Law <law@redhat.com>
6548 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6551 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6554 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6558 * common.opt (gas-loc-support, gas-locview-support): New.
6559 (ginline-points, ginternal-reset-location-views): New.
6560 * doc/invoke.texi: Document them. Use @itemx where intended.
6561 (gvariable-location-views): Adjust.
6562 * target.def (reset_location_view): New.
6563 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6564 (TARGET_RESET_LOCATION_VIEW): New.
6565 * doc/tm.texi: Rebuilt.
6566 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6567 (dwarf2out_default_as_locview_support): New.
6568 (output_asm_line_debug_info): Use option variables.
6569 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6570 (output_loc_list): Likewise.
6571 (add_high_low_attributes): Check option variables.
6572 Don't output entry view attribute in strict mode.
6573 (gen_inlined_subroutine_die): Check option variables.
6574 (dwarf2out_inline_entry): Likewise.
6575 (init_sections_and_labels): Likewise.
6576 (dwarf2out_early_finish): Likewise.
6577 (maybe_reset_location_view): New, from...
6578 (dwarf2out_var_location): ... here. Call it.
6579 * debug.h (dwarf2out_default_as_loc_support): Declare.
6580 (dwarf2out_default_as_locview_support): Declare.
6581 * hooks.c (hook_int_rtx_insn_0): New.
6582 * hooks.h (hook_int_rtx_insn_0): Declare.
6583 * toplev.c (process_options): Take -gas-loc-support and
6584 -gas-locview-support from dwarf2out. Enable
6585 -gvariable-location-views by default only with locview
6586 assembler support. Enable -ginternal-reset-location-views by
6587 default only if the target defines the corresponding hook.
6588 Enable -ginline-points by default if location views are
6589 enabled; force it disabled if statement frontiers are
6591 * tree-inline.c (expand_call_inline): Check option variables.
6592 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6594 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6596 PR tree-optimization/84321
6597 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6598 handling. Also check whether the anti-range contains any values
6599 that satisfy the mask; switch to a VR_RANGE if not.
6601 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6604 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6606 2018-02-13 Martin Jambor <mjambor@suse.cz>
6609 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6610 of call statements, also set location of a load to a temporary.
6612 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6614 * config/rl78/rl78.c (add_vector_labels): New function.
6615 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6616 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6617 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6618 which checks that no arguments are passed.
6619 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6620 * doc/extend.texi: Documentation for the new attribute.
6622 2018-02-13 Andreas Schwab <schwab@suse.de>
6624 * config/riscv/linux.h (CPP_SPEC): Define.
6626 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6629 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6630 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6631 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6632 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6633 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6634 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6635 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6636 temporarily for AES and PCLMUL builtins.
6638 PR tree-optimization/84339
6639 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6640 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6644 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6645 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6646 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6648 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6650 * omp-simd-clone.h: New file.
6651 * omp-simd-clone.c: Include omp-simd-clone.h.
6652 (expand_simd_clones): No longer static.
6653 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6654 cgraph.h and omp-simd-clone.h.
6655 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6656 (vect_recog_widen_shift_pattern): Formatting fix.
6657 (vect_pattern_recog_1): Don't check optab for calls.
6660 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6661 operands[2] into a REG before using gen_lowpart on it.
6663 2018-02-12 Jeff Law <law@redhat.com>
6666 * config/sh/sh.c (find_barrier): Consider a sibling call
6669 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6670 successfully back substituting a reg.
6672 2018-02-12 Richard Biener <rguenther@suse.de>
6674 PR tree-optimization/84037
6675 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6676 parameter, move visited init to caller.
6677 (vect_slp_analyze_operations): Separate cost from validity
6678 check, initialize visited once for all instances.
6679 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6681 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6683 (vect_model_promotion_demotion_cost): Likewise.
6684 (vectorizable_bswap): Guard cost modeling with !slp_node
6685 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6687 (vectorizable_call): Likewise.
6688 (vectorizable_conversion): Likewise.
6689 (vectorizable_assignment): Likewise.
6690 (vectorizable_shift): Likewise.
6691 (vectorizable_operation): Likewise.
6692 (vectorizable_store): Likewise.
6693 (vectorizable_load): Likewise.
6694 (vectorizable_condition): Likewise.
6695 (vectorizable_comparison): Likewise.
6697 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6700 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6701 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6702 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6704 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6707 * params.def (inline-min-speedup): Increase from 8 to 15.
6708 (max-inline-insns-auto): Decrease from 40 to 30.
6709 * ipa-split.c (consider_split): Add some buffer for function to
6710 be considered inlining candidate.
6711 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6714 2018-02-12 Richard Biener <rguenther@suse.de>
6716 PR tree-optimization/84037
6717 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6718 matched stmts if we cannot swap the non-matched ones.
6720 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6722 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6723 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6724 _mm_maskz_scalef_round_ss): New intrinsics.
6725 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6726 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6727 __builtin_ia32_scalefss_round): Remove.
6728 (__builtin_ia32_scalefsd_mask_round,
6729 __builtin_ia32_scalefss_mask_round): New intrinsics.
6730 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6731 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6732 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6733 "<round_constraint>")): Changed to ...
6734 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6735 "<round_scalar_constraint>")): ... this.
6736 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6737 %0, %1, %2<round_op3>}"): Changed to ...
6738 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6739 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6740 %2<round_scalar_mask_op3>}"): ... this.
6741 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6743 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6745 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6746 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6747 (_mm_maskz_sqrt_round_ss): New intrinsics.
6748 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6749 (__builtin_ia32_sqrtsd_mask_round)
6750 (__builtin_ia32_sqrtss_mask_round): New builtins.
6751 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6752 (__builtin_ia32_sqrtss_round): Remove.
6753 (__builtin_ia32_sqrtsd_mask_round)
6754 (__builtin_ia32_sqrtss_mask_round): New builtins.
6755 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6756 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6757 ((match_operand:VF_128 1 "vector_operand"
6758 "xBm,<round_constraint>")): Changed to ...
6759 ((match_operand:VF_128 1 "vector_operand"
6760 "xBm,<round_scalar_constraint>")): ... this.
6761 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6762 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6763 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6764 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6765 %<iptr>1<round_scalar_mask_op3>}): ... this.
6766 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6767 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6769 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6772 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6773 Cast vec_cmpeq result to correct type.
6774 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6775 Cast vec_cmpgt result to correct type.
6777 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6779 * final.c (final_scan_insn_1): Renamed from...
6780 (final_scan_insn): ... this. New wrapper, to recover
6781 seen from the outermost call in recursive ones.
6782 * config/sparc/sparc.c (output_return): Drop seen from call.
6783 (output_sibcall): Likewise.
6784 * config/visium/visium.c (output_branch): Likewise.
6786 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6788 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6791 2018-02-10 Alan Modra <amodra@gmail.com>
6794 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6795 Specify LR as an input.
6797 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6800 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6801 remove_member_access_dummy_vars): New functions.
6802 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6803 lower_omp_1, execute_lower_omp): Use them.
6805 PR rtl-optimization/84308
6806 * shrink-wrap.c (spread_components): Release todo vector.
6808 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6810 PR rtl-optimization/57193
6811 * ira-color.c (struct allocno_color_data): Add member
6812 conflict_allocno_hard_prefs.
6813 (update_conflict_allocno_hard_prefs): New.
6814 (bucket_allocno_compare_func): Add a preference based on
6815 conflict_allocno_hard_prefs.
6816 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6817 (color_allocnos): Remove a dead code. Initiate
6818 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6820 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6823 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6824 constraint from =wa to wa. Avoid a subreg on the output operand,
6825 instead use a pseudo and subreg it in a move.
6826 (p9_xxbrd_<mode>): Changed to ...
6827 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6828 (p9_xxbrd_v2df): New expander.
6829 (p9_xxbrw_<mode>): Changed to ...
6830 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6831 (p9_xxbrw_v4sf): New expander.
6833 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6835 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6837 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6840 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6841 multiply in 32-bit mode.
6842 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6843 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6846 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6848 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6849 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6850 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6851 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6853 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6856 * dwarf2out.c (is_trivial_indirect_ref): New function.
6857 (dwarf2out_late_global_decl): Do not generate a location
6858 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6859 and that are not defined in the current unit.
6861 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6863 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6864 instead of a libcall for UNORDERED.
6866 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6869 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6870 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6872 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6875 * config/s390/s390.c (s390_set_current_function): Invoke
6876 s390_indirect_branch_settings also if fndecl didn't change.
6878 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6880 * config/rs6000/rs6000.md (blockage): Set length to zero.
6882 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6884 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6886 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6889 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6890 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6894 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6895 PARALLEL incoming that failed vt_get_decl_and_offset check.
6898 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6899 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6901 (get_variable_section): For decls in named .bss* sections pass true as
6902 second argument to bss_initializer_p.
6904 2018-02-09 Marek Polacek <polacek@redhat.com>
6905 Jakub Jelinek <jakub@redhat.com>
6908 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6909 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6910 Sync some changes from cxx_fold_indirect_ref.
6912 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6914 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6916 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6917 (BLOCK_INLINE_ENTRY_LABEL): New.
6918 (dwarf2out_var_location): Disregard inline entry markers.
6919 (inline_entry_data): New struct.
6920 (inline_entry_data_hasher): New hashtable type.
6921 (inline_entry_data_hasher::hash): New.
6922 (inline_entry_data_hasher::equal): New.
6923 (inline_entry_data_table): New variable.
6924 (add_high_low_attributes): Add DW_AT_entry_pc and
6925 DW_AT_GNU_entry_view attributes if a pending entry is found
6926 in inline_entry_data_table. Add old entry_pc attribute only
6927 if debug nonbinding markers are disabled.
6928 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6929 markers are enabled.
6930 (block_within_block_p, dwarf2out_inline_entry): New.
6931 (dwarf2out_finish): Check that no entries remained in
6932 inline_entry_data_table.
6933 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6934 (final_scan_insn, notice_source_line): Likewise.
6935 (rest_of_clean_state): Skip inline entry markers.
6936 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6938 * gimple.c (gimple_build_debug_inline_entry): New.
6939 * gimple.h (enum gimple_debug_subcode): Add
6940 GIMPLE_DEBUG_INLINE_ENTRY.
6941 (gimple_build_debug_inline_entry): Declare.
6942 (gimple_debug_inline_entry_p): New.
6943 (gimple_debug_nonbind_marker_p): Adjust.
6944 * insn-notes.def (INLINE_ENTRY): New.
6945 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6946 inline entry marker notes.
6947 (print_insn): Likewise.
6948 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6949 (INSN_DEBUG_MARKER_KIND): Likewise.
6950 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6951 * tree-inline.c (expand_call_inline): Build and insert
6952 debug_inline_entry stmt.
6953 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6954 inline entry blocks early, if nonbind markers are enabled.
6955 (dump_scope_block): Dump fragment info.
6956 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6957 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6958 (gimple_build_debug_inline_entry): New.
6959 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6960 Enable/disable inline entry points too.
6961 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6962 (DEBUG_INSN): Describe inline entry markers.
6964 * common.opt (gvariable-location-views): New.
6965 (gvariable-location-views=incompat5): New.
6966 * config.in: Rebuilt.
6967 * configure: Rebuilt.
6968 * configure.ac: Test assembler for view support.
6969 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6970 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6971 * dwarf2out.c (var_loc_view): New typedef.
6972 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6973 (dwarf2out_locviews_in_attribute): New.
6974 (dwarf2out_locviews_in_loclist): New.
6975 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6976 (enum dw_line_info_opcode): Add LI_adv_address.
6977 (struct dw_line_info_table): Add view.
6978 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6979 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6980 (zero_view_p): New variable.
6981 (ZERO_VIEW_P): New macro.
6982 (output_asm_line_debug_info): New.
6983 (struct var_loc_node): Add view.
6984 (add_AT_view_list, AT_loc_list): New.
6985 (add_var_loc_to_decl): Add view param. Test it against last.
6986 (new_loc_list): Add view params. Record them.
6987 (AT_loc_list_ptr): Handle loc and view lists.
6988 (view_list_to_loc_list_val_node): New.
6989 (print_dw_val): Handle dw_val_class_view_list.
6990 (size_of_die): Likewise.
6991 (value_format): Likewise.
6992 (loc_list_has_views): New.
6993 (gen_llsym): Set vl_symbol too.
6994 (maybe_gen_llsym, skip_loc_list_entry): New.
6995 (dwarf2out_maybe_output_loclist_view_pair): New.
6996 (output_loc_list): Output view list or entries too.
6997 (output_view_list_offset): New.
6998 (output_die): Handle dw_val_class_view_list.
6999 (output_dwarf_version): New.
7000 (output_compilation_unit_header): Use it.
7001 (output_skeleton_debug_sections): Likewise.
7002 (output_rnglists, output_line_info): Likewise.
7003 (output_pubnames, output_aranges): Update version comments.
7004 (output_one_line_info_table): Output view numbers in asm comments.
7005 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7006 Call maybe_gen_llsym.
7007 (loc_list_from_tree_1): Adjust.
7008 (add_AT_location_description): Create view list attribute if
7009 needed, check it's absent otherwise.
7010 (convert_cfa_to_fb_loc_list): Adjust.
7011 (maybe_emit_file): Call output_asm_line_debug_info for test.
7012 (dwarf2out_var_location): Reset views as needed. Precompute
7013 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7014 attribute. Set view.
7015 (new_line_info_table): Reset next view.
7016 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7017 (dwarf2out_source_line): Likewise. Output view resets and labels to
7018 the assembler, or select appropriate line info opcodes.
7019 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7020 (optimize_string_length): Catch it. Adjust.
7021 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7022 dw_val_class_view_list, and remove it if no longer needed.
7023 (hash_loc_list): Hash view numbers.
7024 (loc_list_hasher::equal): Compare them.
7025 (optimize_location_lists): Check whether a view list symbol is
7026 needed, and whether the locview attribute is present, and
7027 whether they match. Remove the locview attribute if no longer
7029 (index_location_lists): Call skip_loc_list_entry for test.
7030 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7031 Use output_dwarf_version.
7032 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7033 (struct dw_val_node): Add val_view_list.
7034 * final.c (SEEN_NEXT_VIEW): New.
7035 (set_next_view_needed): New.
7036 (clear_next_view_needed): New.
7037 (maybe_output_next_view): New.
7038 (final_start_function): Rename to...
7039 (final_start_function_1): ... this. Take pointer to FIRST,
7040 add SEEN parameter. Emit param bindings in the initial view.
7041 (final_start_function): Reintroduce SEEN-less interface.
7042 (final): Rename to...
7043 (final_1): ... this. Take SEEN parameter. Output final pending
7044 next view at the end.
7045 (final): Reintroduce seen-less interface.
7046 (final_scan_insn): Output pending next view before switching
7047 sections or ending a block. Mark the next view as needed when
7048 outputting variable locations. Notify debug backend of section
7049 changes, and of location view changes.
7050 (rest_of_handle_final): Adjust.
7051 * toplev.c (process_options): Autodetect value for debug variable
7052 location views option. Warn on incompat5 without -gdwarf-5.
7053 * doc/invoke.texi (gvariable-location-views): New.
7054 (gvariable-location-views=incompat5): New.
7055 (gno-variable-location-views): New.
7057 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7059 PR tree-optimization/84136
7060 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7061 that the result of find_edge is non-NULL.
7063 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7066 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7067 storing integer register in SImode. Fix cost of 256 and 512
7068 byte aligned SSE register store.
7070 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7072 * config/i386/i386.c (ix86_multiplication_cost): Fix
7073 multiplication cost for TARGET_AVX512DQ.
7075 2018-02-08 Marek Polacek <polacek@redhat.com>
7077 PR tree-optimization/84238
7078 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7081 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7083 PR tree-optimization/84265
7084 * tree-vect-stmts.c (vectorizable_store): Don't treat
7085 VMAT_CONTIGUOUS accesses as grouped.
7086 (vectorizable_load): Likewise.
7088 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7090 PR tree-optimization/81635
7091 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7092 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7093 (test_round_for_mask): New functions.
7094 (wide_int_cc_tests): Call test_round_for_mask.
7095 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7096 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7097 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7098 range returned by get_range_info.
7100 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7103 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7104 * symtab.c: Include builtins.h
7105 (symtab_node::output_to_lto_symbol_table_p): Move here
7106 from lto-streamer-out.c:output_symbol_p.
7107 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7108 (output_symbol_p): Move all logic to symtab.c
7109 (produce_symtab): Update.
7111 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7113 * config/s390/s390-opts.h (enum indirect_branch): Define.
7114 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7115 (s390_indirect_branch_via_thunk)
7116 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7117 (enum s390_indirect_branch_type): Define.
7118 * config/s390/s390.c (struct s390_frame_layout, struct
7119 machine_function): Remove.
7120 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7121 (indirect_branch_table_label_no, indirect_branch_table_name):
7123 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7124 (enum s390_indirect_branch_option): Define.
7125 (s390_return_addr_from_memory): New function.
7126 (s390_handle_string_attribute): New function.
7127 (s390_attribute_table): Add new attribute handler.
7128 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7129 (s390_indirect_branch_via_thunk): New function.
7130 (s390_indirect_branch_via_inline_thunk): New function.
7131 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7132 sibling call optimization for non z10 compiles.
7133 (s390_emit_call): Force indirect branch target to be a single
7134 register. Add r1 clobber for non-z10 compiles.
7135 (s390_emit_epilogue): Emit return jump via return_use expander.
7136 (s390_reorg): Handle JUMP_INSNs as execute targets.
7137 (s390_option_override_internal): Perform validity checks for the
7138 new command line options.
7139 (s390_indirect_branch_attrvalue): New function.
7140 (s390_indirect_branch_settings): New function.
7141 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7142 (s390_output_indirect_thunk_function): New function.
7143 (s390_code_end): Implement target hook.
7144 (s390_case_values_threshold): Implement target hook.
7145 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7147 * config/s390/s390.h (struct s390_frame_layout)
7148 (struct machine_function): Move here from s390.c.
7149 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7150 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7151 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7152 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7153 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7154 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7155 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7156 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7157 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7158 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7159 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7160 (mnemonic attribute): Add values which aren't recognized
7162 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7163 pattern for branch conversion. Fix mnemonic attribute.
7164 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7165 indirect branch via thunk if requested.
7166 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7167 ("*indirect_jump"): Disable for branch conversion using out of
7169 ("indirect_jump_via_thunk<mode>_z10")
7170 ("indirect_jump_via_thunk<mode>")
7171 ("indirect_jump_via_inlinethunk<mode>_z10")
7172 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7173 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7174 ("casesi_jump_via_inlinethunk<mode>_z10")
7175 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7176 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7177 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7178 ("*indirect2_jump"): Disable for branch conversion.
7179 ("casesi_jump"): Turn into expander and expand patterns for branch
7181 ("return_use"): New expander.
7182 ("*return"): Emit return via thunk and rename it to ...
7183 ("*return<mode>"): ... this one.
7184 * config/s390/s390.opt: Add new options and and enum for the
7187 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7189 * lra-constraints.c (match_reload): Unconditionally use
7190 gen_lowpart_SUBREG, rather than selecting between that
7191 and equivalent gen_rtx_SUBREG code.
7193 2018-02-08 Richard Biener <rguenther@suse.de>
7195 PR tree-optimization/84233
7196 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7197 changed flag instead of boguously re-using phi_inserted.
7199 2018-02-08 Martin Jambor <mjambor@suse.cz>
7201 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7202 static local variables.
7204 2018-02-08 Richard Biener <rguenther@suse.de>
7206 PR tree-optimization/84278
7207 * tree-vect-stmts.c (vectorizable_store): When looking for
7208 smaller vector types to perform grouped strided loads/stores
7209 make sure the mode is supported by the target.
7210 (vectorizable_load): Likewise.
7212 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7214 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7215 Increase LDP/STP opportunities by adding adjacent callee-saves.
7217 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7219 PR rtl-optimization/84068
7220 PR rtl-optimization/83459
7221 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7223 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7225 PR tree-optimization/84224
7226 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7227 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7230 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7233 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7234 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7235 count, remove one USE.
7237 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7239 * doc/install.texi (Configuration): Document the
7240 --with-long-double-format={ibm,ieee} PowerPC configuration
7244 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7245 Convert from define_expand to be define_insn_and_split. Rework
7246 float/double/_Float128 conversions to QI/HI/SImode to work with
7247 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7248 conversions to QI/HImode types did a store and then a load to
7249 truncate the value. For conversions to VSX registers, don't split
7250 the insn, instead emit the code directly. Use the code iterator
7251 any_fix to combine signed and unsigned conversions.
7252 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7253 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7254 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7255 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7256 (fix_<mode>di2_hw): Likewise.
7257 (fixuns_<mode>di2_hw): Likewise.
7258 (fix_<mode>si2_hw): Likewise.
7259 (fixuns_<mode>si2_hw): Likewise.
7260 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7261 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7262 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7263 fix<uns>_trunc<SFDF:mode>si2_p8.
7264 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7266 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7267 (fix<uns>_<mode>_mem): Likewise.
7268 (fctiw<u>z_<mode>_mem): Likewise.
7269 (fix<uns>_<mode>_mem): Likewise.
7270 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7271 the register allocator from doing a direct move to the GPRs to do
7272 a store, and instead use the ISA 3.0 store byte/half-word from
7273 vector register instruction. For IEEE 128-bit floating point,
7274 also optimize stores of 32-bit ints.
7275 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7277 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7279 * genextract.c (push_pathstr_operand): New function to support
7281 (walk_rtx): Call push_pathstr_operand.
7282 (print_path): Support [a-zA-Z].
7284 2018-02-07 Richard Biener <rguenther@suse.de>
7286 PR tree-optimization/84037
7287 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7288 (cse_and_gimplify_to_preheader): Declare.
7289 (vect_get_place_in_interleaving_chain): Likewise.
7290 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7292 (_loop_vec_info::~_loop_vec_info): Delete it.
7293 (cse_and_gimplify_to_preheader): New function.
7294 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7295 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7296 (vectorizable_load): Likewise. For grouped stores always base
7297 the IV on the first element.
7298 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7299 condition before gimplifying.
7301 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7303 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7304 *DIV_EXPR and *MOD_EXPR.
7306 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7309 * config/i386/i386.c (ix86_option_override_internal): Mask out
7310 the CF_SET bit when checking -fcf-protection.
7312 2018-02-07 Tom de Vries <tom@codesourcery.com>
7315 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7318 2018-02-07 Richard Biener <rguenther@suse.de>
7320 PR tree-optimization/84204
7321 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7324 PR tree-optimization/84205
7325 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7326 special-case isl_ast_op_zdiv_r.
7328 PR tree-optimization/84223
7329 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7330 Only add conditions from within the region.
7331 (gather_bbs::after_dom_children): Adjust.
7333 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7336 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7337 * config/avr/avr.md: Only post-reload split REG-REG moves if
7338 either register is GENERAL_REG_P.
7340 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7342 PR tree-optimization/84235
7343 * tree-ssa-scopedtables.c
7344 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7345 if the subtraction is performed in floating point type where NaNs are
7346 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7347 build 1. Formatting fix.
7349 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7352 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7353 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7354 and skip it regardless of bb boundaries. Use CALL_P macro,
7355 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7356 unnecessarily, formatting fix.
7358 2018-02-06 Michael Collison <michael.collison@arm.com>
7360 * config/arm/thumb2.md:
7361 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7362 (*thumb_mov_notscc): Ditto.
7364 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7367 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7368 unsigned_fix, not "s".
7370 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7372 * configure.ac (gcc_fn_eh_frame_ro): New function.
7373 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7374 correct .eh_frame permissions.
7375 * configure: Regenerate.
7377 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7379 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7382 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7384 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7385 Display warning message for -mno-speculate-indirect-jumps.
7387 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7389 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7391 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7393 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7395 PR tree-optimization/84225
7396 * tree-eh.c (find_trapping_overflow): Only call
7397 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7399 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7402 * config/i386/i386.c: Reimplement the check of possible options
7403 -mibt/-mshstk conbination. Change error messages.
7404 * doc/invoke.texi: Fix a typo: remove extra '='.
7406 2018-02-06 Marek Polacek <polacek@redhat.com>
7408 PR tree-optimization/84228
7409 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7411 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7414 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7415 emitted arch directives.
7416 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7417 __ARM_FEATURE_COPROC before changing architectures.
7419 2018-02-06 Richard Biener <rguenther@suse.de>
7421 * config/i386/i386.c (print_reg): Fix typo.
7422 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7424 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7426 * configure: Regenerate.
7428 2018-02-05 Martin Sebor <msebor@redhat.com>
7430 PR tree-optimization/83369
7431 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7434 2018-02-05 Martin Liska <mliska@suse.cz>
7436 * doc/invoke.texi: Cherry-pick upstream r323995.
7438 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7440 * ira.c (ira_init_register_move_cost): Adjust comment.
7442 2018-02-05 Martin Liska <mliska@suse.cz>
7444 PR gcov-profile/84137
7445 * doc/gcov.texi: Fix typo in documentation.
7447 2018-02-05 Martin Liska <mliska@suse.cz>
7449 PR gcov-profile/83879
7450 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7451 using dlopen functionality.
7453 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7455 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7456 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7457 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7458 _mm_maskz_range_round_ss): New intrinsics.
7459 (__builtin_ia32_rangesd128_round)
7460 (__builtin_ia32_rangess128_round): Remove.
7461 (__builtin_ia32_rangesd128_mask_round,
7462 __builtin_ia32_rangess128_mask_round): New builtins.
7463 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7464 __builtin_ia32_rangess128_round): Remove.
7465 (__builtin_ia32_rangesd128_mask_round,
7466 __builtin_ia32_rangess128_mask_round): New builtins.
7467 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7468 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7469 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7470 "<round_saeonly_constraint>")): Changed to ...
7471 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7472 "<round_saeonly_scalar_constraint>")): ... this.
7473 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7474 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7475 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7476 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7477 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7479 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7481 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7483 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7484 Remove all values except native, 8540 and 8548.
7486 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7488 * config/i386/i386.c (ix86_output_function_return): Pass
7489 INVALID_REGNUM, instead of -1, as invalid register number to
7490 indirect_thunk_name and output_indirect_thunk.
7492 2018-02-02 Julia Koval <julia.koval@intel.com>
7494 * config.gcc: Add -march=icelake.
7495 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7496 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7497 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7498 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7499 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7500 (processor_target_table): Add icelake.
7501 (ix86_option_override_internal): Handle new PTAs.
7502 (get_builtin_code_for_version): Handle icelake.
7503 (M_INTEL_COREI7_ICELAKE): New.
7504 (fold_builtin_cpu): Handle icelake.
7505 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7506 * doc/invoke.texi: Add -march=icelake.
7508 2018-02-02 Julia Koval <julia.koval@intel.com>
7510 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7511 to wide_int_bitmask.
7512 * wide-int-bitmask.h: New.
7514 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7517 * config/i386/i386.md: Replace Pmode with word_mode in
7518 builtin_setjmp_setup and builtin_longjmp to support x32.
7520 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7524 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7526 (rs6000_supported_cpu_names): New static variable.
7527 (linux_cpu_translation_table): Likewise.
7528 (elf_platform) <cpu>: Define new static variable and use it.
7529 Translate kernel AT_PLATFORM name to canonical name if needed.
7530 Error if platform name is unknown.
7532 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7535 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7537 2018-02-01 Jeff Law <law@redhat.com>
7540 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7541 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7542 the scratch if RELEASE_VIA_POP is false.
7543 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7544 If we have to save a temporary register, decrement SIZE appropriately.
7545 Pass new arguments to release_scratch_register_on_entry.
7546 (ix86_adjust_stack_and_probe): Likewise.
7547 (ix86_emit_probe_stack_range): Pass new arguments to
7548 release_scratch_register_on_entry.
7550 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7552 PR rtl-optimization/84157
7553 * combine.c (change_zero_ext): Use REG_P predicate in
7554 front of HARD_REGISTER_P predicate.
7556 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7558 * config/avr/avr.c (avr_option_override): Move disabling of
7559 -fdelete-null-pointer-checks to...
7560 * common/config/avr/avr-common.c (avr_option_optimization_table):
7563 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7565 PR tree-optimization/81635
7566 * tree-data-ref.c (split_constant_offset_1): For types that
7567 wrap on overflow, try to use range info to prove that wrapping
7570 2018-02-01 Renlin Li <renlin.li@arm.com>
7573 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7575 (aarch64_register_move_cost): Likewise.
7576 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7578 (REG_CLASS_NAMES): Likewise.
7579 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7580 TAILCALL_ADDR_REGS. Remove IP registers.
7581 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7583 2018-02-01 Richard Biener <rguenther@suse.de>
7585 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7586 for specifying RPO order and allow NULL for that.
7587 * domwalk.c (dom_walker::dom_walker): Likewise.
7588 (dom_walker::walk): Handle NULL RPO order.
7589 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7591 (rewrite_update_dom_walker): Likewise.
7592 (mark_def_dom_walker): Likewise.
7594 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7596 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7597 (aarch64_maybe_expand_sve_subreg_move): Declare.
7598 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7599 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7601 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7602 that are semantically a reverse operation.
7603 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7604 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7605 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7607 (aarch64_can_change_mode_class): For big-endian, forbid changes
7608 between two SVE modes if they have different element sizes.
7610 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7612 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7613 the TImode handling for big-endian targets.
7615 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7617 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7618 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7620 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7621 Remove BSWAP handing for big-endian targets and use the form of
7622 LD1RQ appropariate for the mode.
7624 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7626 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7627 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7630 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7633 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7634 check for operands that need to go through aarch64_sve_reload_be.
7636 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7638 PR tree-optimization/81661
7639 PR tree-optimization/84117
7640 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7641 * tree-eh.c: Include gimplify.h.
7642 (find_trapping_overflow, replace_trapping_overflow,
7643 rewrite_to_non_trapping_overflow): New functions.
7644 * tree-vect-loop.c: Include tree-eh.h.
7645 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7646 * tree-data-ref.c: Include tree-eh.h.
7647 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7649 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7651 PR rtl-optimization/84123
7652 * combine.c (change_zero_ext): Check if hard register satisfies
7653 can_change_dest_mode before calling gen_lowpart_SUBREG.
7655 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7658 * ira.c (ira_init_register_move_cost): Remove assert.
7660 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7662 PR rtl-optimization/84071
7663 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7664 * doc/tm.texi: Regenerate.
7666 2018-01-31 Richard Biener <rguenther@suse.de>
7668 PR tree-optimization/84132
7669 * tree-data-ref.c (analyze_miv_subscript): Properly
7670 check whether evolution_function_is_affine_multivariate_p
7671 before calling gcd_of_steps_may_divide_p.
7673 2018-01-31 Julia Koval <julia.koval@intel.com>
7676 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7677 * config/i386/i386.md (rdpid_rex64) New.
7678 (rdpid): Make 32bit only.
7680 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7683 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7684 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7686 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7689 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7691 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7693 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7695 PR rtl-optimization/84071
7696 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7697 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7699 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7701 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7702 (arc_attribute_table): Add 'aux' attribute.
7703 (arc_in_small_data_p): Consider aux like variables.
7704 (arc_is_aux_reg_p): New function.
7705 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7706 (arc_get_aux_arg): New function.
7707 (prepare_move_operands): Handle aux-register access.
7708 (arc_handle_aux_attribute): New function.
7709 * doc/extend.texi (ARC Variable attributes): Add subsection.
7711 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7713 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7714 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7715 (arc_attribute_table): Add 'uncached' attribute.
7716 (arc_print_operand): Print '.di' flag for uncached memory
7718 (arc_in_small_data_p): Do not consider for small data the uncached
7720 (arc_is_uncached_mem_p): New function.
7721 * config/arc/predicates.md (compact_store_memory_operand): Check
7722 for uncached memory accesses.
7723 (nonvol_nonimm_operand): Likewise.
7724 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7726 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7729 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7730 falign-loops=): Add Optimization flag.
7732 2018-01-30 Jeff Law <law@redhat.com>
7735 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7736 INT_REGISTERS_SAVED. Check it prior to calling
7737 get_scratch_register_on_entry.
7738 (ix86_adjust_stack_and_probe): Similarly.
7739 (ix86_emit_probe_stack_range): Similarly.
7740 (ix86_expand_prologue): Corresponding changes.
7742 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7745 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7746 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7748 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7751 * lra-constraints.c (curr_insn_transform): Process AND in the
7754 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7756 PR rtl-optimization/83986
7757 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7758 dependence against last_pending_memory_flush in addition to
7761 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7763 PR tree-optimization/81611
7764 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7767 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7770 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7773 2018-01-30 Richard Biener <rguenther@suse.de>
7774 Jakub Jelinek <jakub@redhat.com>
7776 PR tree-optimization/84111
7777 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7778 inner loops added during recursion, as they don't have up-to-date
7781 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7784 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7785 (can_inline_edge_by_limits_p): ... here.
7786 (can_early_inline_edge_p, check_callers,
7787 update_caller_keys, update_callee_keys, recursive_inlining,
7788 add_new_edges_to_heap, speculation_useful_p,
7789 inline_small_functions,
7790 inline_small_functions, flatten_function,
7791 inline_to_all_callers_1): Update.
7793 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7795 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7798 2018-01-30 Richard Biener <rguenther@suse.de>
7800 PR tree-optimization/83008
7801 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7802 invariant and constant vector uses in stmts when they need
7805 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7808 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7809 * configure: Regenerate.
7811 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7813 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7815 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7816 Use gen_rtx_REG rather than gen_lowpart.
7818 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7820 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7821 rather than 0 when creating partial subregs.
7823 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7825 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7828 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7831 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7832 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7833 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7834 flags. This restores the settings used before the 2017-07-24.
7835 Turning off pre increment/decrement/modify allows IVOPTS to
7836 optimize DF/SF loops where the index is an int.
7838 2018-01-29 Richard Biener <rguenther@suse.de>
7839 Kelvin Nilsen <kelvin@gcc.gnu.org>
7842 * tree-vect-stmts.c (vectorizable_call): Don't call
7843 targetm.vectorize_builtin_md_vectorized_function if callee is
7846 2018-01-22 Carl Love <cel@us.ibm.com>
7848 * doc/extend.tex: Fix typo in second arg in
7849 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7851 2018-01-29 Richard Biener <rguenther@suse.de>
7853 PR tree-optimization/84086
7854 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7855 (flush_ssaname_freelist): When SSA names were released reset
7856 the SCEV hash table.
7858 2018-01-29 Richard Biener <rguenther@suse.de>
7860 PR tree-optimization/84057
7861 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7862 removed paths when removing edges.
7864 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7866 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7867 -mfunction-return=@var{choice}.
7869 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7872 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7874 (layout::print_source_line): Likewise.
7875 (test_get_line_width_without_trailing_whitespace): Add test cases.
7877 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7880 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7883 2018-01-26 Jim Wilson <jimw@sifive.com>
7885 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7887 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7890 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7892 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7893 and CMP + SUB-immediate -> SUBS.
7895 2018-01-26 Martin Sebor <msebor@redhat.com>
7897 PR tree-optimization/83896
7898 * tree-ssa-strlen.c (get_string_len): Rename...
7899 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7900 Avoid assuming length is constant.
7901 (handle_char_store): Use HOST_WIDE_INT for string length.
7903 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7906 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7907 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7909 2018-01-26 Richard Biener <rguenther@suse.de>
7911 PR rtl-optimization/84003
7912 * dse.c (record_store): Only record redundant stores when
7913 the earlier store aliases at least all accesses the later one does.
7915 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7917 PR rtl-optimization/83985
7918 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7919 REG_CFA_RESTORE insns.
7920 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7921 REG_CFA_RESTORE insns here.
7924 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7925 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7927 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7929 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7930 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7931 (arc_init): Likewise.
7932 (arc_override_options): Likewise.
7933 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7935 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7937 * config/arc/arc.h (TARGET_DBNZ): Define.
7938 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7939 properly set the tune attribute.
7940 (dbnz): Use TARGET_DBNZ guard.
7941 * config/arc/arc.opt (mtune): Add core3 option.
7943 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7945 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7946 recognize new pic like addresses.
7947 (arc_delegitimize_address): Clean up.
7949 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7951 * config/arc/arc-arches.def: Option mrf16 valid for all
7953 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7954 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7955 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7956 * config/arc/arc-tables.opt: Regenerate.
7957 * config/arc/arc.c (arc_conditional_register_usage): Handle
7958 reduced register file case.
7959 (arc_file_start): Set must have build attributes.
7960 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7962 * config/arc/arc.opt (mrf16): Add new option.
7963 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7964 * config/arc/genmultilib.awk: Handle new mrf16 option.
7965 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7966 * config/arc/t-multilib: Regenerate.
7967 * doc/invoke.texi (ARC Options): Document mrf16 option.
7969 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7971 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7972 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7973 (arc_attribute_table): Add 'secure_call' attribute.
7974 (arc_print_operand): Print secure call operand.
7975 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7977 (arc_is_secure_call_p): New function. * config/arc/arc.md
7978 (call_i): Add support for sjli instruction.
7979 (call_value_i): Likewise.
7980 * config/arc/constraints.md (Csc): New constraint.
7982 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7983 John Eric Martin <John.Martin@emmicro-us.com>
7985 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7986 * config/arc/arc.c (_arc_jli_section): New struct.
7987 (arc_jli_section): New type.
7988 (rc_jli_sections): New static variable.
7989 (arc_handle_jli_attribute): New function.
7990 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7991 (arc_file_end): New function.
7992 (TARGET_ASM_FILE_END): Define.
7993 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7994 (arc_add_jli_section): New function.
7995 (jli_call_scan): Likewise.
7996 (arc_reorg): Call jli_call_scan.
7997 (arc_output_addsi): Remove 'S' from printing asm operand.
7998 (arc_is_jli_call_p): New function.
7999 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8001 (movhi_insn): Likewise.
8002 (movsi_insn): Likewise.
8003 (movsi_set_cc_insn): Likewise.
8004 (loadqi_update): Likewise.
8005 (load_zeroextendqisi_update): Likewise.
8006 (load_signextendqisi_update): Likewise.
8007 (loadhi_update): Likewise.
8008 (load_zeroextendhisi_update): Likewise.
8009 (load_signextendhisi_update): Likewise.
8010 (loadsi_update): Likewise.
8011 (loadsf_update): Likewise.
8012 (movsicc_insn): Likewise.
8013 (bset_insn): Likewise.
8014 (bxor_insn): Likewise.
8015 (bclr_insn): Likewise.
8016 (bmsk_insn): Likewise.
8017 (bicsi3_insn): Likewise.
8018 (cmpsi_cc_c_insn): Likewise.
8019 (movsi_ne): Likewise.
8020 (movsi_cond_exec): Likewise.
8021 (clrsbsi2): Likewise.
8032 (call_i): Remove 'S' asm letter, add jli instruction.
8033 (call_value_i): Likewise.
8034 * config/arc/arc.op (mjli-always): New option.
8035 * config/arc/constraints.md (Cji): New constraint.
8036 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8038 (subsf3_fpx): Likewise.
8039 (mulsf3_fpx): Likewise.
8040 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8042 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8044 * doc/invoke.texi (ARC): Document mjli-always option.
8046 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8048 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8049 avoid addition with 0 and use incw and decw where possible.
8051 2018-01-26 Richard Biener <rguenther@suse.de>
8053 PR tree-optimization/81082
8054 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8055 association if it requires casting to unsigned.
8056 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8057 from fold_plusminus_mult_expr to catch important cases late when
8058 range info is available.
8060 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8062 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8063 * configure.ac (hidden_linkonce): New test.
8064 * configure: Regenerate.
8065 * config.in: Regenerate.
8067 2018-01-26 Julia Koval <julia.koval@intel.com>
8069 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8070 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8071 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8072 _mm_mask_bitshuffle_epi64_mask): Fix type.
8073 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8074 USI_FTYPE_V4DI_V4DI_USI): Remove.
8075 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8076 __builtin_ia32_vpshufbitqmb256_mask,
8077 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8078 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8079 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8081 2018-01-26 Alan Modra <amodra@gmail.com>
8084 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8085 UNSPEC_VBPERMQ. Sort other unspecs.
8087 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8089 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8091 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8094 * predict.c (drop_profile): Do not push/pop cfun; update also
8096 (handle_missing_profiles): Fix logic looking for zero profiles.
8098 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8101 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8102 on functions with #pragma omp declare simd or functions with simd
8104 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8105 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8106 Remove trailing \n from warning_at calls.
8108 2018-01-25 Tom de Vries <tom@codesourcery.com>
8111 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8112 for neutered workers.
8114 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8117 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8118 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8120 2018-01-24 Jeff Law <law@redhat.com>
8123 * i386.c (get_probe_interval): Move to earlier point.
8124 (ix86_compute_frame_layout): If -fstack-clash-protection and
8125 the frame is larger than the probe interval, then use pushes
8126 to save registers rather than reg->mem moves.
8127 (ix86_expand_prologue): Remove conditional for int_registers_saved
8130 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8133 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8134 min/max for never referenced object.
8136 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8139 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8141 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8142 attributes from DECL_ATTRIBUTES (decl) without affecting
8143 DECL_ATTRIBUTES (current_function_decl).
8144 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8145 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8147 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8149 PR tree-optimization/83979
8150 * fold-const.c (fold_comparison): Use constant_boolean_node
8151 instead of boolean_{true,false}_node.
8153 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8155 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8158 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8160 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8161 Simplify the clause that sets the length attribute.
8162 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8163 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8164 clause that sets the length attribute.
8165 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8167 2018-01-24 Tom de Vries <tom@codesourcery.com>
8170 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8171 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8172 Add strict parameter.
8173 (prevent_branch_around_nothing): Insert dummy insn between branch to
8174 label and label with no ptx insn inbetween.
8175 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8177 2018-01-24 Tom de Vries <tom@codesourcery.com>
8180 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8181 for neutered threads in warp.
8182 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8184 2018-01-24 Richard Biener <rguenther@suse.de>
8186 PR tree-optimization/83176
8187 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8190 2018-01-24 Richard Biener <rguenther@suse.de>
8192 PR tree-optimization/82819
8193 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8194 code generating pluses that are no-ops in the target precision.
8196 2018-01-24 Richard Biener <rguenther@suse.de>
8199 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8201 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8203 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8204 to merge probabilities.
8205 * predict.c (probably_never_executed): Also mark as cold functions
8206 with global 0 profile and guessed local profile.
8207 * profile-count.c (profile_probability::combine_with_count): New
8209 * profile-count.h (profile_probability::operator*,
8210 profile_probability::operator*=, profile_probability::operator/,
8211 profile_probability::operator/=): Reduce precision to adjusted
8212 and set value to guessed on contradictory divisions.
8213 (profile_probability::combine_with_freq): Remove.
8214 (profile_probability::combine_wiht_count): Declare.
8215 (profile_count::force_nonzero):: Set to adjusted.
8216 (profile_count::probability_in):: Set quality to adjusted.
8217 * tree-ssa-tail-merge.c (replace_block_by): Use
8220 2018-01-23 Andrew Waterman <andrew@sifive.com>
8221 Jim Wilson <jimw@sifive.com>
8223 * config/riscv/riscv.c (riscv_stack_boundary): New.
8224 (riscv_option_override): Set riscv_stack_boundary. Handle
8225 riscv_preferred_stack_boundary_arg.
8226 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8227 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8228 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8229 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8230 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8231 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8233 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8236 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8237 of struct ix86_frame.
8238 (ix86_expand_epilogue): Likewise. Add a local variable for
8239 the reg_save_offset field in struct ix86_frame.
8241 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8243 PR tree-optimization/82604
8244 * tree-loop-distribution.c (enum partition_kind): New enum item
8245 PKIND_PARTIAL_MEMSET.
8246 (partition_builtin_p): Support above new enum item.
8247 (generate_code_for_partition): Ditto.
8248 (compute_access_range): Differentiate cases that equality can be
8249 proven at all loops, the innermost loops or no loops.
8250 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8251 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8252 (finalize_partitions, distribute_loop): Don't fuse partition of
8253 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8254 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8257 2018-01-23 Martin Liska <mliska@suse.cz>
8259 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8260 order to ignore the predictor.
8261 (PRED_POLYMORPHIC_CALL): Likewise.
8262 (PRED_RECURSIVE_CALL): Likewise.
8264 2018-01-23 Martin Liska <mliska@suse.cz>
8266 * tree-profile.c (tree_profiling): Print function header to
8267 aware reader which function we are working on.
8268 * value-prof.c (gimple_find_values_to_profile): Do not print
8269 not interesting value histograms.
8271 2018-01-23 Martin Liska <mliska@suse.cz>
8273 * profile-count.h (enum profile_quality): Add
8274 profile_uninitialized as the first value. Do not number values
8275 as they are zero based.
8276 (profile_count::verify): Update sanity check.
8277 (profile_probability::verify): Likewise.
8279 2018-01-23 Nathan Sidwell <nathan@acm.org>
8281 * doc/invoke.texi (ffor-scope): Deprecate.
8283 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8285 PR tree-optimization/83510
8286 * domwalk.c (set_all_edges_as_executable): New function.
8287 (dom_walker::dom_walker): Convert bool param
8288 "skip_unreachable_blocks" to enum reachability. Move setup of
8289 edge flags to set_all_edges_as_executable and only do it when
8290 reachability is REACHABLE_BLOCKS.
8291 * domwalk.h (enum dom_walker::reachability): New enum.
8292 (dom_walker::dom_walker): Convert bool param
8293 "skip_unreachable_blocks" to enum reachability.
8294 (set_all_edges_as_executable): New decl.
8295 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8296 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8298 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8299 but converting true to REACHABLE_BLOCKS.
8300 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8302 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8303 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8304 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8306 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8307 if check_all_array_refs will be called.
8309 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8311 * tree.c (selftest::test_location_wrappers): Add more test
8314 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8316 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8317 (selftest::test_bit_in_range): Likewise.
8319 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8322 * doc/sourcebuild.texi (vect_float): Say that the selector
8323 only describes the situation when -funsafe-math-optimizations is on.
8324 (vect_float_strict): Document.
8326 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8328 PR tree-optimization/83965
8329 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8330 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8331 instead of checking only for a reduction.
8332 (vect_recog_widen_sum_pattern): Likewise.
8334 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8336 * predict.c (probably_never_executed): Only use precise profile info.
8337 (compute_function_frequency): Skip after inlining hack since we now
8338 have quality checking.
8340 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8342 * profile-count.h (profile_probability::very_unlikely,
8343 profile_probability::unlikely, profile_probability::even): Set
8344 precision to guessed.
8346 2018-01-23 Richard Biener <rguenther@suse.de>
8348 PR tree-optimization/83963
8349 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8350 Properly terminate dominator walk when crossing the exit edge not
8351 when visiting its source block.
8353 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8356 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8357 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8359 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8361 PR tree-optimization/83957
8362 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8363 semicolon after for body surrounded by braces.
8365 PR tree-optimization/83081
8366 * profile-count.h (profile_probability::split): New method.
8367 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8368 Use profile_probability::split.
8369 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8370 when splitting a single conditional jump into 2.
8372 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8374 PR tree-optimization/69452
8375 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8378 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8380 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8381 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8382 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8384 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8386 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8387 * config/rl78/rl78.md (movdi): New define_expand.
8388 * config/rl78/rl78.c (rl78_split_movdi): New function.
8390 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8393 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8395 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8396 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8397 128-bit to produce an UNSPEC move to get the double word with the
8398 signbit and then a shift directly to do signbit.
8399 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8400 implementation with a new version that just does either a direct
8401 move or a regular move. Move memory interface to separate insns.
8402 Move insns so they are next to the expander.
8403 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8404 with signbit move. Split big and little endian case.
8405 (signbit<mode>2_dm_mem_le): Likewise.
8406 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8407 (signbit<mode>2_dm2): Likewise.
8409 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8411 * config/rl78/rl78.md (anddi3): New define_expand.
8413 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8415 * config/rl78/rl78.md (umindi3): New define_expand.
8417 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8419 * config/rl78/rl78.md (smindi3): New define_expand.
8421 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8423 * config/rl78/rl78.md (smaxdi3): New define_expand.
8425 2018-01-22 Carl Love <cel@us.ibm.com>
8427 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8428 LVX_V1TI): Add macro expansion.
8429 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8430 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8431 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8432 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8433 Change check to determine if the instruction is a byte reversing
8434 entry. Fix typo in comment.
8435 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8436 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8437 Add def_builtin calls for new builtins.
8438 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8439 Add define_insn expansion.
8441 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8443 * config/rl78/rl78.md (umaxdi3): New define_expand.
8445 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8447 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8448 for non-QImode registers.
8450 2018-01-22 Richard Biener <rguenther@suse.de>
8452 PR tree-optimization/83963
8453 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8454 including the loop exit block.
8455 (scop_detection::merge_sese): Likewise.
8456 (scop_detection::add_scop): Do it here instead.
8458 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8460 * doc/sourcebuild.texi (arm_softfloat): Document.
8462 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8465 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8466 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8467 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8469 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8470 David Edelsohn <dje.gcc@gmail.com>
8473 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8474 Change "crset eq" to "crset 2".
8475 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8476 (*call_indirect_aix<mode>_nospec): Likewise.
8477 (*call_value_indirect_aix<mode>_nospec): Likewise.
8478 (*call_indirect_elfv2<mode>_nospec): Likewise.
8479 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8480 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8481 change assembly output from . to $.
8482 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8483 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8484 (*tablejump<mode>_internal1_nospec): Likewise.
8486 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8489 * config/sh/sh_optimize_sett_clrt.cc:
8490 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8492 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8494 PR tree-optimization/83940
8495 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8496 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8497 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8498 use it to pass back the definition type.
8499 (vect_check_store_rhs): Likewise rhs_dt_out.
8500 (vect_build_gather_load_calls): Add a mask_dt argument and use
8501 it instead of a call to vect_is_simple_use.
8502 (vectorizable_store): Update calls to vect_check_load_store_mask
8503 and vect_check_store_rhs. Use the dt returned by the latter instead
8504 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8505 instead of calls to vect_is_simple_use. Pass the scalar rather
8506 than the vector operand to vect_is_simple_use when handling
8507 second and subsequent copies of an rhs value.
8508 (vectorizable_load): Update calls to vect_check_load_store_mask
8509 and vect_build_gather_load_calls. Use the cached mask_dt and
8510 gs_info.offset_dt instead of calls to vect_is_simple_use.
8512 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8515 * tree-emutls.c: Include gimplify.h.
8516 (lower_emutls_2): New function.
8517 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8518 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8519 it before further processing.
8522 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8523 UINTVAL (trueop1) instead of INTVAL (op1).
8525 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8529 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8530 INCOMING_FRAME_SP_OFFSET if not defined.
8531 (scan_trace): Add ENTRY argument. If true and
8532 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8533 emit a note to adjust the CFA offset.
8534 (create_cfi_notes): Adjust scan_trace callers.
8535 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8536 INCOMING_FRAME_SP_OFFSET in the CIE.
8537 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8538 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8540 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8541 * doc/tm.texi: Regenerated.
8543 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8545 PR rtl-optimization/83147
8546 * lra-constraints.c (remove_inheritance_pseudos): Use
8547 lra_substitute_pseudo_within_insn.
8549 2018-01-19 Tom de Vries <tom@codesourcery.com>
8550 Cesar Philippidis <cesar@codesourcery.com>
8553 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8555 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8558 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8559 spaces for function labels.
8561 2018-01-19 Martin Liska <mliska@suse.cz>
8563 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8564 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8565 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8566 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8567 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8568 (PRED_CONST_RETURN): Change from 69 to 65.
8569 (PRED_NULL_RETURN): Change from 91 to 71.
8570 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8571 (PRED_LOOP_GUARD): Change from 66 to 73.
8573 2018-01-19 Martin Liska <mliska@suse.cz>
8575 * predict.c (predict_insn_def): Add new assert.
8576 (struct branch_predictor): Change type to signed integer.
8577 (test_prediction_value_range): Amend test to cover
8579 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8580 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8581 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8582 (PRED_LOOP_IV_COMPARE): Likewise.
8583 * predict.h (PROB_UNINITIALIZED): Define new constant.
8585 2018-01-19 Martin Liska <mliska@suse.cz>
8587 * predict.c (dump_prediction): Add new format for
8588 analyze_brprob.py script which is enabled with -details
8590 * profile-count.h (precise_p): New function.
8592 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8594 PR tree-optimization/83922
8595 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8596 there are no statements that need masking.
8597 (vect_active_double_reduction_p): New function.
8598 (vect_analyze_loop_operations): Use it when handling phis that
8599 are not in the loop header.
8601 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8603 PR tree-optimization/83914
8604 * tree-vect-loop.c (vectorizable_induction): Don't convert
8605 init_expr or apply the peeling adjustment for inductions
8606 that are nested within the vectorized loop.
8608 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8610 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8613 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8617 * function.h (gimplify_parameters): Add gimple_seq * argument.
8618 * function.c: Include gimple.h and options.h.
8619 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8620 for the added local temporaries if needed.
8621 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8622 if there are any parameter cleanups, wrap whole body into a
8623 try/finally with the cleanups.
8625 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8628 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8629 Use GET_MODE_CLASS for scalar floating point.
8631 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8635 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8636 Fix call of call_cgraph_insertion_hooks.
8638 2018-01-18 Martin Sebor <msebor@redhat.com>
8640 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8642 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8645 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8648 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8651 * common.opt: (-ffile-prefix-map): New option.
8652 * opts.c (common_handle_option): Defer it.
8653 * opts-global.c (handle_common_deferred_options): Handle it.
8654 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8655 * file-prefix-map.h: New file.
8656 (remap_debug_filename, add_debug_prefix_map): ...here.
8657 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8658 * final.c (debug_prefix_map, add_debug_prefix_map
8659 remap_debug_filename): Move to...
8660 * file-prefix-map.c: New file.
8661 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8662 generalize, get rid of alloca(), use strrchr() instead of strchr().
8663 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8664 Implement in terms of add_prefix_map().
8665 (remap_macro_filename, remap_debug_filename): Implement in term of
8667 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8668 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8669 * dbxout.c: Include file-prefix-map.h.
8670 * varasm.c: Likewise.
8671 * vmsdbgout.c: Likewise.
8672 * xcoffout.c: Likewise.
8673 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8674 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8675 * doc/invoke.texi (-ffile-prefix-map): Document.
8676 (-fdebug-prefix-map): Update description.
8678 2018-01-18 Martin Liska <mliska@suse.cz>
8680 * config/i386/i386.c (indirect_thunk_name): Document that also
8682 (output_indirect_thunk): Document why both instructions
8683 (pause and lfence) are generated.
8685 2018-01-18 Richard Biener <rguenther@suse.de>
8687 PR tree-optimization/83887
8688 * graphite-scop-detection.c
8689 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8690 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8691 (scop_detection::merge_sese): Re-implement with a flood-fill
8692 algorithm that properly finds a SESE region if it exists.
8694 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8697 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8698 pointer_diff optimizations use view_convert instead of convert.
8700 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8702 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8703 Generate different code for -mno-speculate-indirect-jumps.
8704 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8705 (*call_indirect_aix<mode>): Disable for
8706 -mno-speculate-indirect-jumps.
8707 (*call_indirect_aix<mode>_nospec): New define_insn.
8708 (*call_value_indirect_aix<mode>): Disable for
8709 -mno-speculate-indirect-jumps.
8710 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8711 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8712 -mno-speculate-indirect-jumps.
8713 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8715 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8717 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8718 long double type, set the flags for noting the default long double
8719 type, even if we don't pass or return a long double type.
8721 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8724 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8727 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8729 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8730 support for merge[hl].
8731 (fold_mergehl_helper): New helper function.
8732 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8733 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8734 (altivec_vmrglw_direct): Add xxmrglw insn.
8736 2018-01-17 Andrew Waterman <andrew@sifive.com>
8738 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8739 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8741 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8744 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8745 call the lto_location_cache before reading the
8746 DECL_SOURCE_LOCATION of the types.
8748 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8749 Richard Sandiford <richard.sandiford@linaro.org>
8751 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8752 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8753 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8754 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8755 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8757 * config/aarch64/constraints.md (aarch64_movti_operand):
8759 * config/aarch64/predicates.md (Uti): Add new constraint.
8761 2018-01-17 Carl Love <cel@us.ibm.com>
8763 * config/rs6000/vsx.md (define_expand xl_len_r,
8764 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8765 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8767 (define_expand, define_insn): Move the shift left from the
8768 define_insn to the define_expand for lxvl and stxvl instructions.
8769 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8770 and XL_LEN_R definitions to PURE.
8772 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8774 * config/i386/i386.c (indirect_thunk_name): Declare regno
8775 as unsigned int. Compare regno with INVALID_REGNUM.
8776 (output_indirect_thunk): Ditto.
8777 (output_indirect_thunk_function): Ditto.
8778 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8779 in the call to output_indirect_thunk_function.
8781 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8784 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8785 rather than the size of inner_type to determine the stack slot size
8786 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8788 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8791 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8794 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8796 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8797 endian Linux systems to optionally enable multilibs for selecting
8798 the long double type if the user configured an explicit type.
8799 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8800 have no long double multilibs if not defined.
8801 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8802 warn if the user used -mabi={ieee,ibm}longdouble and we built
8803 multilibs for long double.
8804 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8805 appropriate multilib option.
8806 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8808 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8809 for building long double multilibs.
8810 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8812 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8814 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8817 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8819 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8822 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8825 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8828 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8830 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8831 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8833 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8835 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8836 different rtl trees depending on TARGET_64BIT.
8837 (rs6000_gen_lvx): Likewise.
8839 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8841 * config/visium/visium.md (nop): Tweak comment.
8842 (hazard_nop): Likewise.
8844 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8846 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8847 -mspeculate-indirect-jumps.
8848 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8849 for -mno-speculate-indirect-jumps.
8850 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8851 (*call_value_indirect_elfv2<mode>): Disable for
8852 -mno-speculate-indirect-jumps.
8853 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8854 (indirect_jump): Emit different RTL for
8855 -mno-speculate-indirect-jumps.
8856 (*indirect_jump<mode>): Disable for
8857 -mno-speculate-indirect-jumps.
8858 (*indirect_jump<mode>_nospec): New define_insn.
8859 (tablejump): Emit different RTL for
8860 -mno-speculate-indirect-jumps.
8861 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8862 (tablejumpsi_nospec): New define_expand.
8863 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8864 (tablejumpdi_nospec): New define_expand.
8865 (*tablejump<mode>_internal1): Disable for
8866 -mno-speculate-indirect-jumps.
8867 (*tablejump<mode>_internal1_nospec): New define_insn.
8868 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8871 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8873 * caller-save.c (insert_save): Drop unnecessary parameter. All
8876 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8877 Richard Biener <rguenth@suse.de>
8880 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8881 return early, inline manually is_gimple_sizepos. Make sure if we
8882 call gimplify_expr we don't end up with a gimple constant.
8883 * tree.c (variably_modified_type_p): Don't return true for
8884 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8885 * gimplify.h (is_gimple_sizepos): Remove.
8887 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8889 PR tree-optimization/83857
8890 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8891 vectorizable_live_operation for pure SLP statements.
8892 (vectorizable_live_operation): Handle PHIs.
8894 2018-01-16 Richard Biener <rguenther@suse.de>
8896 PR tree-optimization/83867
8897 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8898 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8900 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8903 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8904 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8905 If off is not INTEGER_CST, issue a may not be aligned warning
8906 rather than isn't aligned. Use isn%'t rather than isn't.
8907 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8909 <case MULT_EXPR>: Improve the case when bottom and one of the
8910 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8911 operand, in that case check if the other operand is multiple of
8912 bottom divided by the INTEGER_CST operand.
8914 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8917 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8918 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8919 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8920 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8921 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8922 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8923 (pa_function_arg_size): New function.
8925 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8927 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8928 in a separate statement.
8930 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8932 PR tree-optimization/83847
8933 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8934 group gathers and scatters.
8936 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8938 PR rtl-optimization/86620
8939 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8941 PR rtl-optimization/83213
8942 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8943 to last if both are JUMP_INSNs.
8945 PR tree-optimization/83843
8946 * gimple-ssa-store-merging.c
8947 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8948 store_immediate_info for bswap/nop orig_stores.
8950 2018-01-15 Andrew Waterman <andrew@sifive.com>
8952 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8954 <UDIV>: Increase cost if !TARGET_DIV.
8956 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8958 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8959 (define_attr "cr_logical_3op"): New.
8960 (cceq_ior_compare): Adjust.
8961 (cceq_ior_compare_complement): Adjust.
8962 (*cceq_rev_compare): Adjust.
8963 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8964 (is_cracked_insn): Adjust.
8965 (insn_must_be_first_in_group): Adjust.
8966 * config/rs6000/40x.md: Adjust.
8967 * config/rs6000/440.md: Adjust.
8968 * config/rs6000/476.md: Adjust.
8969 * config/rs6000/601.md: Adjust.
8970 * config/rs6000/603.md: Adjust.
8971 * config/rs6000/6xx.md: Adjust.
8972 * config/rs6000/7450.md: Adjust.
8973 * config/rs6000/7xx.md: Adjust.
8974 * config/rs6000/8540.md: Adjust.
8975 * config/rs6000/cell.md: Adjust.
8976 * config/rs6000/e300c2c3.md: Adjust.
8977 * config/rs6000/e500mc.md: Adjust.
8978 * config/rs6000/e500mc64.md: Adjust.
8979 * config/rs6000/e5500.md: Adjust.
8980 * config/rs6000/e6500.md: Adjust.
8981 * config/rs6000/mpc.md: Adjust.
8982 * config/rs6000/power4.md: Adjust.
8983 * config/rs6000/power5.md: Adjust.
8984 * config/rs6000/power6.md: Adjust.
8985 * config/rs6000/power7.md: Adjust.
8986 * config/rs6000/power8.md: Adjust.
8987 * config/rs6000/power9.md: Adjust.
8988 * config/rs6000/rs64.md: Adjust.
8989 * config/rs6000/titan.md: Adjust.
8991 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8993 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8994 ix86_indirect_branch_register logic.
8996 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8998 * config/i386/constraints.md (Bs): Update
8999 ix86_indirect_branch_register check. Don't check
9000 ix86_indirect_branch_register with GOT_memory_operand.
9002 * config/i386/predicates.md (GOT_memory_operand): Don't check
9003 ix86_indirect_branch_register here.
9004 (GOT32_symbol_operand): Likewise.
9006 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9008 * config/i386/predicates.md (constant_call_address_operand):
9009 Rewrite ix86_indirect_branch_register logic.
9010 (sibcall_insn_operand): Likewise.
9012 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9014 * config/i386/constraints.md (Bs): Replace
9015 ix86_indirect_branch_thunk_register with
9016 ix86_indirect_branch_register.
9018 * config/i386/i386.md (indirect_jump): Likewise.
9019 (tablejump): Likewise.
9020 (*sibcall_memory): Likewise.
9021 (*sibcall_value_memory): Likewise.
9022 Peepholes of indirect call and jump via memory: Likewise.
9023 * config/i386/i386.opt: Likewise.
9024 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9025 (GOT_memory_operand): Likewise.
9026 (call_insn_operand): Likewise.
9027 (sibcall_insn_operand): Likewise.
9028 (GOT32_symbol_operand): Likewise.
9030 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9033 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9034 type rather than type addr's type points to.
9035 (expand_omp_atomic_mutex): Likewise.
9036 (expand_omp_atomic): Likewise.
9038 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9041 * config/i386/i386.c (output_indirect_thunk_function): Use
9042 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9043 for __x86_return_thunk.
9045 2018-01-15 Richard Biener <rguenther@suse.de>
9048 * expmed.c (extract_bit_field_1): Fix typo.
9050 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9053 * config/arm/iterators.md (VF): New mode iterator.
9054 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9055 Remove integer-related logic from pattern.
9056 (neon_vabd<mode>_3): Likewise.
9058 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9061 * common.opt (fstrict-overflow): No longer an alias.
9062 (fwrapv-pointer): New option.
9063 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9064 also for pointer types based on flag_wrapv_pointer.
9065 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9066 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9067 opts->x_flag_wrapv got set.
9068 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9069 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9070 POINTER_TYPE_OVERFLOW_UNDEFINED.
9071 * match.pd: Likewise in address comparison pattern.
9072 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9074 2018-01-15 Richard Biener <rguenther@suse.de>
9077 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9078 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9079 Reset type names to their identifier if their TYPE_DECL doesn't
9080 have linkage (and thus is used for ODR and devirt).
9081 (save_debug_info_for_decl): Remove.
9082 (save_debug_info_for_type): Likewise.
9083 (add_tree_to_fld_list): Adjust.
9084 * tree-pretty-print.c (dump_generic_node): Make dumping of
9085 type names more robust.
9087 2018-01-15 Richard Biener <rguenther@suse.de>
9089 * BASE-VER: Bump to 8.0.1.
9091 2018-01-14 Martin Sebor <msebor@redhat.com>
9094 * builtins.c (check_access): Avoid warning when the no-warning bit
9097 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9099 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9100 * ira-color (allocno_hard_regs_compare): Likewise.
9102 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9105 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9106 Use .pushsection/.popsection.
9108 2018-01-14 Martin Sebor <msebor@redhat.com>
9111 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9113 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9115 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9116 entry from extra_headers.
9117 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9118 extra_headers, make the list bitwise identical to the i?86-*-* one.
9120 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9122 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9123 -mcmodel=large with -mindirect-branch=thunk,
9124 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9125 -mfunction-return=thunk-extern.
9126 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9127 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9128 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9130 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9132 * config/i386/i386.c (print_reg): Print the name of the full
9133 integer register without '%'.
9134 (ix86_print_operand): Handle 'V'.
9135 * doc/extend.texi: Document 'V' modifier.
9137 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9139 * config/i386/constraints.md (Bs): Disallow memory operand for
9140 -mindirect-branch-register.
9142 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9143 (GOT_memory_operand): Likewise.
9144 (call_insn_operand): Likewise.
9145 (sibcall_insn_operand): Likewise.
9146 (GOT32_symbol_operand): Likewise.
9147 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9148 for -mindirect-branch-register.
9149 (tablejump): Likewise.
9150 (*sibcall_memory): Likewise.
9151 (*sibcall_value_memory): Likewise.
9152 Disallow peepholes of indirect call and jump via memory for
9153 -mindirect-branch-register.
9154 (*call_pop): Replace m with Bw.
9155 (*call_value_pop): Likewise.
9156 (*sibcall_pop_memory): Replace m with Bs.
9157 * config/i386/i386.opt (mindirect-branch-register): New option.
9158 * doc/invoke.texi: Document -mindirect-branch-register option.
9160 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9162 * config/i386/i386-protos.h (ix86_output_function_return): New.
9163 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9164 set function_return_type.
9165 (indirect_thunk_name): Add ret_p to indicate thunk for function
9167 (output_indirect_thunk_function): Pass false to
9168 indirect_thunk_name.
9169 (ix86_output_indirect_branch_via_reg): Likewise.
9170 (ix86_output_indirect_branch_via_push): Likewise.
9171 (output_indirect_thunk_function): Create alias for function
9172 return thunk if regno < 0.
9173 (ix86_output_function_return): New function.
9174 (ix86_handle_fndecl_attribute): Handle function_return.
9175 (ix86_attribute_table): Add function_return.
9176 * config/i386/i386.h (machine_function): Add
9177 function_return_type.
9178 * config/i386/i386.md (simple_return_internal): Use
9179 ix86_output_function_return.
9180 (simple_return_internal_long): Likewise.
9181 * config/i386/i386.opt (mfunction-return=): New option.
9182 (indirect_branch): Mention -mfunction-return=.
9183 * doc/extend.texi: Document function_return function attribute.
9184 * doc/invoke.texi: Document -mfunction-return= option.
9186 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9188 * config/i386/i386-opts.h (indirect_branch): New.
9189 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9190 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9191 with local indirect jump when converting indirect call and jump.
9192 (ix86_set_indirect_branch_type): New.
9193 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9194 (indirectlabelno): New.
9195 (indirect_thunk_needed): Likewise.
9196 (indirect_thunk_bnd_needed): Likewise.
9197 (indirect_thunks_used): Likewise.
9198 (indirect_thunks_bnd_used): Likewise.
9199 (INDIRECT_LABEL): Likewise.
9200 (indirect_thunk_name): Likewise.
9201 (output_indirect_thunk): Likewise.
9202 (output_indirect_thunk_function): Likewise.
9203 (ix86_output_indirect_branch_via_reg): Likewise.
9204 (ix86_output_indirect_branch_via_push): Likewise.
9205 (ix86_output_indirect_branch): Likewise.
9206 (ix86_output_indirect_jmp): Likewise.
9207 (ix86_code_end): Call output_indirect_thunk_function if needed.
9208 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9210 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9211 (ix86_attribute_table): Add indirect_branch.
9212 * config/i386/i386.h (machine_function): Add indirect_branch_type
9213 and has_local_indirect_jump.
9214 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9216 (tablejump): Likewise.
9217 (*indirect_jump): Use ix86_output_indirect_jmp.
9218 (*tablejump_1): Likewise.
9219 (simple_return_indirect_internal): Likewise.
9220 * config/i386/i386.opt (mindirect-branch=): New option.
9221 (indirect_branch): New.
9224 (thunk-inline): Likewise.
9225 (thunk-extern): Likewise.
9226 * doc/extend.texi: Document indirect_branch function attribute.
9227 * doc/invoke.texi: Document -mindirect-branch= option.
9229 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9232 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9234 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9236 * ipa-inline.c (want_inline_small_function_p): Return false if
9237 inlining has already failed with CIF_FINAL_ERROR.
9238 (update_caller_keys): Call want_inline_small_function_p before
9240 (update_callee_keys): Likewise.
9242 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9244 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9246 (rs6000_quadword_masked_address_p): Likewise.
9247 (quad_aligned_load_p): Likewise.
9248 (quad_aligned_store_p): Likewise.
9249 (const_load_sequence_p): Add comment to describe the outer-most loop.
9250 (mimic_memory_attributes_and_flags): New function.
9251 (rs6000_gen_stvx): Likewise.
9252 (replace_swapped_aligned_store): Likewise.
9253 (rs6000_gen_lvx): Likewise.
9254 (replace_swapped_aligned_load): Likewise.
9255 (replace_swapped_load_constant): Capitalize argument name in
9256 comment describing this function.
9257 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9258 and stores that access quad-word aligned addresses and replace
9259 with stvx or lvx instructions when appropriate.
9260 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9261 New function prototype.
9262 (rs6000_quadword_masked_address_p): Likewise.
9263 (rs6000_gen_lvx): Likewise.
9264 (rs6000_gen_stvx): Likewise.
9265 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9266 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9267 when memory address is aligned.
9268 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9269 this split to select lvx instruction when memory address is aligned.
9270 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9271 instruction when memory address is aligned.
9272 (*vsx_le_perm_load_v16qi): Likewise.
9273 (four unnamed splitters): Modify to select the stvx instruction
9274 when memory is aligned.
9276 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9278 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9279 which appears in the queue multiple times.
9281 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9282 Alan Hayward <alan.hayward@arm.com>
9283 David Sherwood <david.sherwood@arm.com>
9285 * tree-vectorizer.h (vec_lower_bound): New structure.
9286 (_loop_vec_info): Add check_nonzero and lower_bounds.
9287 (LOOP_VINFO_CHECK_NONZERO): New macro.
9288 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9289 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9290 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9291 fields. Make seg_len the distance travelled, not including the
9293 (dr_direction_indicator): Declare.
9294 (dr_zero_step_indicator): Likewise.
9295 (dr_known_forward_stride_p): Likewise.
9296 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9298 (runtime_alias_check_p): Allow runtime alias checks with
9300 (operator ==): Compare access_size and align.
9301 (prune_runtime_alias_test_list): Rework for new distinction between
9302 the access_size and seg_len.
9303 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9305 (get_segment_min_max): New function.
9306 (create_intersect_range_checks): Use it.
9307 (dr_step_indicator): New function.
9308 (dr_direction_indicator): Likewise.
9309 (dr_zero_step_indicator): Likewise.
9310 (dr_known_forward_stride_p): Likewise.
9311 * tree-loop-distribution.c (data_ref_segment_size): Return
9312 DR_STEP * (niters - 1).
9313 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9315 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9316 (vect_preserves_scalar_order_p): New function, split out from...
9317 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9318 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9319 (vect_vfa_access_size): New function.
9320 (vect_vfa_align): Likewise.
9321 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9322 (dump_lower_bound): New function.
9323 (vect_check_lower_bound): Likewise.
9324 (vect_small_gap_p): Likewise.
9325 (vectorizable_with_step_bound_p): Likewise.
9326 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9327 depencies if the vectorization factor is 1. Convert the checks
9328 for nonzero steps into checks on the bounds of DR_STEP. Try using
9329 a bunds check for variable steps if the minimum required step is
9330 relatively small. Update calls to the dr_with_seg_len
9331 constructor and to vect_compile_time_alias.
9332 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9334 (vect_loop_versioning): Call it.
9335 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9337 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9339 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9340 Alan Hayward <alan.hayward@arm.com>
9341 David Sherwood <david.sherwood@arm.com>
9343 * doc/sourcebuild.texi (vect_scatter_store): Document.
9344 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9346 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9348 * genopinit.c (main): Add supports_vec_scatter_store and
9349 supports_vec_scatter_store_cached to target_optabs.
9350 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9351 IFN_MASK_SCATTER_STORE.
9352 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9354 * internal-fn.h (internal_store_fn_p): Declare.
9355 (internal_fn_stored_value_index): Likewise.
9356 * internal-fn.c (scatter_store_direct): New macro.
9357 (expand_scatter_store_optab_fn): New function.
9358 (direct_scatter_store_optab_supported_p): New macro.
9359 (internal_store_fn_p): New function.
9360 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9361 IFN_MASK_SCATTER_STORE.
9362 (internal_fn_mask_index): Likewise.
9363 (internal_fn_stored_value_index): New function.
9364 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9366 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9367 * optabs-query.c (supports_vec_scatter_store_p): New function.
9368 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9369 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9370 true for scatter stores.
9371 (vect_gather_scatter_fn_p): Handle scatter stores too.
9372 (vect_check_gather_scatter): Consider using scatter stores if
9373 supports_vec_scatter_store_p.
9374 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9376 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9377 internal_fn_stored_value_index.
9378 (check_load_store_masking): Handle scatter stores too.
9379 (vect_get_store_rhs): Make public.
9380 (vectorizable_call): Use internal_store_fn_p.
9381 (vectorizable_store): Handle scatter store internal functions.
9382 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9383 when deciding whether the end of the group has been reached.
9384 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9385 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9386 (mask_scatter_store<mode>): New insns.
9388 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9389 Alan Hayward <alan.hayward@arm.com>
9390 David Sherwood <david.sherwood@arm.com>
9392 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9393 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9394 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9396 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9397 Use vect_truncate_gather_scatter_offset if we can't treat the
9398 operation as a normal gather load or scatter store.
9399 (get_group_load_store_type): Take the gather_scatter_info
9400 as argument. Try using a gather load or scatter store for
9401 single-element groups.
9402 (get_load_store_type): Update calls to get_group_load_store_type
9403 and vect_use_strided_gather_scatters_p.
9405 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9406 Alan Hayward <alan.hayward@arm.com>
9407 David Sherwood <david.sherwood@arm.com>
9409 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9410 optional tree argument.
9411 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9413 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9414 but continue to use the current value as a fallback.
9415 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9416 to compare the updates.
9417 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9418 (get_load_store_type): Use it when handling a strided access.
9419 (vect_get_strided_load_store_ops): New function.
9420 (vect_get_data_ptr_increment): Likewise.
9421 (vectorizable_load): Handle strided gather loads. Always pass
9422 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9424 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9425 Alan Hayward <alan.hayward@arm.com>
9426 David Sherwood <david.sherwood@arm.com>
9428 * doc/md.texi (gather_load@var{m}): Document.
9429 (mask_gather_load@var{m}): Likewise.
9430 * genopinit.c (main): Add supports_vec_gather_load and
9431 supports_vec_gather_load_cached to target_optabs.
9432 * optabs-tree.c (init_tree_optimization_optabs): Use
9433 ggc_cleared_alloc to allocate target_optabs.
9434 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9435 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9437 * internal-fn.h (internal_load_fn_p): Declare.
9438 (internal_gather_scatter_fn_p): Likewise.
9439 (internal_fn_mask_index): Likewise.
9440 (internal_gather_scatter_fn_supported_p): Likewise.
9441 * internal-fn.c (gather_load_direct): New macro.
9442 (expand_gather_load_optab_fn): New function.
9443 (direct_gather_load_optab_supported_p): New macro.
9444 (direct_internal_fn_optab): New function.
9445 (internal_load_fn_p): Likewise.
9446 (internal_gather_scatter_fn_p): Likewise.
9447 (internal_fn_mask_index): Likewise.
9448 (internal_gather_scatter_fn_supported_p): Likewise.
9449 * optabs-query.c (supports_at_least_one_mode_p): New function.
9450 (supports_vec_gather_load_p): Likewise.
9451 * optabs-query.h (supports_vec_gather_load_p): Declare.
9452 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9453 and memory_type field.
9454 (NUM_PATTERNS): Bump to 15.
9455 * tree-vect-data-refs.c: Include internal-fn.h.
9456 (vect_gather_scatter_fn_p): New function.
9457 (vect_describe_gather_scatter_call): Likewise.
9458 (vect_check_gather_scatter): Try using internal functions for
9459 gather loads. Recognize existing calls to a gather load function.
9460 (vect_analyze_data_refs): Consider using gather loads if
9461 supports_vec_gather_load_p.
9462 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9463 (vect_get_gather_scatter_offset_type): Likewise.
9464 (vect_convert_mask_for_vectype): Likewise.
9465 (vect_add_conversion_to_patterm): Likewise.
9466 (vect_try_gather_scatter_pattern): Likewise.
9467 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9468 (vect_vect_recog_func_ptrs): Add it.
9469 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9470 internal_fn_mask_index and internal_gather_scatter_fn_p.
9471 (check_load_store_masking): Take the gather_scatter_info as an
9472 argument and handle gather loads.
9473 (vect_get_gather_scatter_ops): New function.
9474 (vectorizable_call): Check internal_load_fn_p.
9475 (vectorizable_load): Likewise. Handle gather load internal
9477 (vectorizable_store): Update call to check_load_store_masking.
9478 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9479 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9480 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9481 (aarch64_gather_scale_operand_d): New predicates.
9482 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9483 (mask_gather_load<mode>): New insns.
9485 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9486 Alan Hayward <alan.hayward@arm.com>
9487 David Sherwood <david.sherwood@arm.com>
9489 * optabs.def (fold_left_plus_optab): New optab.
9490 * doc/md.texi (fold_left_plus_@var{m}): Document.
9491 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9492 * internal-fn.c (fold_left_direct): Define.
9493 (expand_fold_left_optab_fn): Likewise.
9494 (direct_fold_left_optab_supported_p): Likewise.
9495 * fold-const-call.c (fold_const_fold_left): New function.
9496 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9497 * tree-parloops.c (valid_reduction_p): New function.
9498 (gather_scalar_reductions): Use it.
9499 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9500 (vect_finish_replace_stmt): Declare.
9501 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9502 (needs_fold_left_reduction_p): New function, split out from...
9503 (vect_is_simple_reduction): ...here. Accept reductions that
9504 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9505 (vect_force_simple_reduction): Also store the reduction type in
9506 the assignment's STMT_VINFO_REDUC_TYPE.
9507 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9508 (merge_with_identity): New function.
9509 (vect_expand_fold_left): Likewise.
9510 (vectorize_fold_left_reduction): Likewise.
9511 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9512 scalar phi in place for it. Check for target support and reject
9513 cases that would reassociate the operation. Defer the transform
9514 phase to vectorize_fold_left_reduction.
9515 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9516 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9517 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9519 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9521 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9522 call to ifc_temp_var.
9524 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9525 Alan Hayward <alan.hayward@arm.com>
9526 David Sherwood <david.sherwood@arm.com>
9528 * target.def (legitimize_address_displacement): Take the original
9529 offset as a poly_int.
9530 * targhooks.h (default_legitimize_address_displacement): Update
9532 * targhooks.c (default_legitimize_address_displacement): Likewise.
9533 * doc/tm.texi: Regenerate.
9534 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9535 as an argument, moving assert of ad->disp == ad->disp_term to...
9536 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9537 Try calling targetm.legitimize_address_displacement before expanding
9538 the address rather than afterwards, and adjust for the new interface.
9539 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9540 Match the new hook interface. Handle SVE addresses.
9541 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9544 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9546 * Makefile.in (OBJS): Add early-remat.o.
9547 * target.def (select_early_remat_modes): New hook.
9548 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9549 * doc/tm.texi: Regenerate.
9550 * targhooks.h (default_select_early_remat_modes): Declare.
9551 * targhooks.c (default_select_early_remat_modes): New function.
9552 * timevar.def (TV_EARLY_REMAT): New timevar.
9553 * passes.def (pass_early_remat): New pass.
9554 * tree-pass.h (make_pass_early_remat): Declare.
9555 * early-remat.c: New file.
9556 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9558 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9560 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9561 Alan Hayward <alan.hayward@arm.com>
9562 David Sherwood <david.sherwood@arm.com>
9564 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9565 vfm1 with a bound_epilog parameter.
9566 (vect_do_peeling): Update calls accordingly, and move the prologue
9567 call earlier in the function. Treat the base bound_epilog as 0 for
9568 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9569 this base when peeling for gaps.
9570 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9571 with fully-masked loops.
9572 (vect_estimate_min_profitable_iters): Handle the single peeled
9573 iteration in that case.
9575 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9576 Alan Hayward <alan.hayward@arm.com>
9577 David Sherwood <david.sherwood@arm.com>
9579 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9580 single-element interleaving even if the size is not a power of 2.
9581 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9582 accesses for single-element interleaving if the group size is
9585 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9586 Alan Hayward <alan.hayward@arm.com>
9587 David Sherwood <david.sherwood@arm.com>
9589 * doc/md.texi (fold_extract_last_@var{m}): Document.
9590 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9591 * optabs.def (fold_extract_last_optab): New optab.
9592 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9593 * internal-fn.c (fold_extract_direct): New macro.
9594 (expand_fold_extract_optab_fn): Likewise.
9595 (direct_fold_extract_optab_supported_p): Likewise.
9596 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9597 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9598 EXTRACT_LAST_REDUCTION.
9599 (get_initial_def_for_reduction): Do not create an initial vector
9600 for EXTRACT_LAST_REDUCTION reductions.
9601 (vectorizable_reduction): Leave the scalar phi in place for
9602 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9603 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9604 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9605 transform phase to vectorizable_condition.
9606 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9608 (vect_finish_stmt_generation): ...here.
9609 (vect_finish_replace_stmt): New function.
9610 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9611 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9613 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9615 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9616 Alan Hayward <alan.hayward@arm.com>
9617 David Sherwood <david.sherwood@arm.com>
9619 * doc/md.texi (extract_last_@var{m}): Document.
9620 * optabs.def (extract_last_optab): New optab.
9621 * internal-fn.def (EXTRACT_LAST): New internal function.
9622 * internal-fn.c (cond_unary_direct): New macro.
9623 (expand_cond_unary_optab_fn): Likewise.
9624 (direct_cond_unary_optab_supported_p): Likewise.
9625 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9626 loops using EXTRACT_LAST.
9627 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9628 (extract_last_<mode>): ...this optab.
9629 (vec_extract<mode><Vel>): Update accordingly.
9631 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9632 Alan Hayward <alan.hayward@arm.com>
9633 David Sherwood <david.sherwood@arm.com>
9635 * target.def (empty_mask_is_expensive): New hook.
9636 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9637 * doc/tm.texi: Regenerate.
9638 * targhooks.h (default_empty_mask_is_expensive): Declare.
9639 * targhooks.c (default_empty_mask_is_expensive): New function.
9640 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9641 if the target says that empty masks are expensive.
9642 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9644 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9646 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9647 Alan Hayward <alan.hayward@arm.com>
9648 David Sherwood <david.sherwood@arm.com>
9650 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9651 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9652 (vect_use_loop_mask_for_alignment_p): New function.
9653 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9654 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9655 niters_skip argument. Make sure that the first niters_skip elements
9656 of the first iteration are inactive.
9657 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9658 Update call to vect_set_loop_masks_directly.
9659 (get_misalign_in_elems): New function, split out from...
9660 (vect_gen_prolog_loop_niters): ...here.
9661 (vect_update_init_of_dr): Take a code argument that specifies whether
9662 the adjustment should be added or subtracted.
9663 (vect_update_init_of_drs): Likewise.
9664 (vect_prepare_for_masked_peels): New function.
9665 (vect_do_peeling): Skip prologue peeling if we're using a mask
9666 instead. Update call to vect_update_inits_of_drs.
9667 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9669 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9670 alignment. Do not include the number of peeled iterations in
9671 the minimum threshold in that case.
9672 (vectorizable_induction): Adjust the start value down by
9673 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9674 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9675 Take the number of skipped iterations into account when calculating
9677 * tree-vect-stmts.c (vect_gen_while_not): New function.
9679 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9680 Alan Hayward <alan.hayward@arm.com>
9681 David Sherwood <david.sherwood@arm.com>
9683 * doc/sourcebuild.texi (vect_fully_masked): Document.
9684 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9686 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9688 (vect_analyze_loop_2): ...here. Don't check the vectorization
9689 factor against the number of loop iterations if the loop is
9692 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9693 Alan Hayward <alan.hayward@arm.com>
9694 David Sherwood <david.sherwood@arm.com>
9696 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9697 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9698 (dump_groups): Update accordingly.
9699 (iv_use::mem_type): New member variable.
9700 (address_p): New function.
9701 (record_use): Add a mem_type argument and initialize the new
9703 (record_group_use): Add a mem_type argument. Use address_p.
9704 Remove obsolete null checks of base_object. Update call to record_use.
9705 (find_interesting_uses_op): Update call to record_group_use.
9706 (find_interesting_uses_cond): Likewise.
9707 (find_interesting_uses_address): Likewise.
9708 (get_mem_type_for_internal_fn): New function.
9709 (find_address_like_use): Likewise.
9710 (find_interesting_uses_stmt): Try find_address_like_use before
9711 calling find_interesting_uses_op.
9712 (addr_offset_valid_p): Use the iv mem_type field as the type
9713 of the addressed memory.
9714 (add_autoinc_candidates): Likewise.
9715 (get_address_cost): Likewise.
9716 (split_small_address_groups_p): Use address_p.
9717 (split_address_groups): Likewise.
9718 (add_iv_candidate_for_use): Likewise.
9719 (autoinc_possible_for_pair): Likewise.
9720 (rewrite_groups): Likewise.
9721 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9722 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9723 (get_alias_ptr_type_for_ptr_address): New function.
9724 (rewrite_use_address): Rewrite address uses in calls that were
9725 identified by find_address_like_use.
9727 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9728 Alan Hayward <alan.hayward@arm.com>
9729 David Sherwood <david.sherwood@arm.com>
9731 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9733 * gimple-expr.h (is_gimple_addressable: Likewise.
9734 * gimple-expr.c (is_gimple_address): Likewise.
9735 * internal-fn.c (expand_call_mem_ref): New function.
9736 (expand_mask_load_optab_fn): Use it.
9737 (expand_mask_store_optab_fn): Likewise.
9739 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9740 Alan Hayward <alan.hayward@arm.com>
9741 David Sherwood <david.sherwood@arm.com>
9743 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9744 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9745 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9746 (cond_umax@var{mode}): Document.
9747 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9748 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9749 (cond_umin_optab, cond_umax_optab): New optabs.
9750 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9751 (COND_IOR, COND_XOR): New internal functions.
9752 * internal-fn.h (get_conditional_internal_fn): Declare.
9753 * internal-fn.c (cond_binary_direct): New macro.
9754 (expand_cond_binary_optab_fn): Likewise.
9755 (direct_cond_binary_optab_supported_p): Likewise.
9756 (get_conditional_internal_fn): New function.
9757 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9758 Cope with reduction statements that are vectorized as calls rather
9760 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9761 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9762 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9763 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9764 (UNSPEC_COND_EOR): New unspecs.
9765 (optab): Add mappings for them.
9766 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9767 (sve_int_op, sve_fp_op): New int attributes.
9769 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9770 Alan Hayward <alan.hayward@arm.com>
9771 David Sherwood <david.sherwood@arm.com>
9773 * optabs.def (while_ult_optab): New optab.
9774 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9775 * internal-fn.def (WHILE_ULT): New internal function.
9776 * internal-fn.h (direct_internal_fn_supported_p): New override
9777 that takes two types as argument.
9778 * internal-fn.c (while_direct): New macro.
9779 (expand_while_optab_fn): New function.
9780 (convert_optab_supported_p): Likewise.
9781 (direct_while_optab_supported_p): New macro.
9782 * wide-int.h (wi::udiv_ceil): New function.
9783 * tree-vectorizer.h (rgroup_masks): New structure.
9784 (vec_loop_masks): New typedef.
9785 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9787 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9788 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9789 (vect_max_vf): New function.
9790 (slpeel_make_loop_iterate_ntimes): Delete.
9791 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9792 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9793 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9794 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9795 internal-fn.h, stor-layout.h and optabs-query.h.
9796 (vect_set_loop_mask): New function.
9797 (add_preheader_seq): Likewise.
9798 (add_header_seq): Likewise.
9799 (interleave_supported_p): Likewise.
9800 (vect_maybe_permute_loop_masks): Likewise.
9801 (vect_set_loop_masks_directly): Likewise.
9802 (vect_set_loop_condition_masked): Likewise.
9803 (vect_set_loop_condition_unmasked): New function, split out from
9804 slpeel_make_loop_iterate_ntimes.
9805 (slpeel_make_loop_iterate_ntimes): Rename to..
9806 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9807 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9808 (vect_do_peeling): Update call accordingly.
9809 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9811 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9812 mask_compare_type, can_fully_mask_p and fully_masked_p.
9813 (release_vec_loop_masks): New function.
9814 (_loop_vec_info): Use it to free the loop masks.
9815 (can_produce_all_loop_masks_p): New function.
9816 (vect_get_max_nscalars_per_iter): Likewise.
9817 (vect_verify_full_masking): Likewise.
9818 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9819 retries, and free the mask rgroups before retrying. Check loop-wide
9820 reasons for disallowing fully-masked loops. Make the final decision
9821 about whether use a fully-masked loop or not.
9822 (vect_estimate_min_profitable_iters): Do not assume that peeling
9823 for the number of iterations will be needed for fully-masked loops.
9824 (vectorizable_reduction): Disable fully-masked loops.
9825 (vectorizable_live_operation): Likewise.
9826 (vect_halve_mask_nunits): New function.
9827 (vect_double_mask_nunits): Likewise.
9828 (vect_record_loop_mask): Likewise.
9829 (vect_get_loop_mask): Likewise.
9830 (vect_transform_loop): Handle the case in which the final loop
9831 iteration might handle a partial vector. Call vect_set_loop_condition
9832 instead of slpeel_make_loop_iterate_ntimes.
9833 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9834 (check_load_store_masking): New function.
9835 (prepare_load_store_mask): Likewise.
9836 (vectorizable_store): Handle fully-masked loops.
9837 (vectorizable_load): Likewise.
9838 (supportable_widening_operation): Use vect_halve_mask_nunits for
9840 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9841 (vect_gen_while): New function.
9842 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9843 (aarch64_uqdec<mode>): New insn.
9845 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9846 Alan Hayward <alan.hayward@arm.com>
9847 David Sherwood <david.sherwood@arm.com>
9849 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9850 (reduc_xor_scal_optab): New optabs.
9851 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9852 (reduc_xor_scal_@var{m}): Document.
9853 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9854 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9856 * fold-const-call.c (fold_const_call): Handle them.
9857 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9858 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9859 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9860 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9861 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9862 (UNSPEC_XORV): New unspecs.
9863 (optab): Add entries for them.
9864 (BITWISEV): New int iterator.
9865 (bit_reduc_op): New int attributes.
9867 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9868 Alan Hayward <alan.hayward@arm.com>
9869 David Sherwood <david.sherwood@arm.com>
9871 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9872 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9873 * optabs.def (vec_shl_insert_optab): New optab.
9874 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9875 (duplicate_and_interleave): Likewise.
9876 * tree-vect-loop.c: Include internal-fn.h.
9877 (neutral_op_for_slp_reduction): New function, split out from
9878 get_initial_defs_for_reduction.
9879 (get_initial_def_for_reduction): Handle option 2 for variable-length
9880 vectors by loading the neutral value into a vector and then shifting
9881 the initial value into element 0.
9882 (get_initial_defs_for_reduction): Replace the code argument with
9883 the neutral value calculated by neutral_op_for_slp_reduction.
9884 Use gimple_build_vector for constant-length vectors.
9885 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9886 but the first group_size elements have a neutral value.
9887 Use duplicate_and_interleave otherwise.
9888 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9889 Update call to get_initial_defs_for_reduction. Handle SLP
9890 reductions for variable-length vectors by creating one vector
9891 result for each scalar result, with the elements associated
9892 with other scalar results stubbed out with the neutral value.
9893 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9894 Require IFN_VEC_SHL_INSERT for double reductions on
9895 variable-length vectors, or SLP reductions that have
9896 a neutral value. Require can_duplicate_and_interleave_p
9897 support for variable-length unchained SLP reductions if there
9898 is no neutral value, such as for MIN/MAX reductions. Also require
9899 the number of vector elements to be a multiple of the number of
9900 SLP statements when doing variable-length unchained SLP reductions.
9901 Update call to vect_create_epilog_for_reduction.
9902 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9903 and remove initial values.
9904 (duplicate_and_interleave): Make public.
9905 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9906 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9908 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9909 Alan Hayward <alan.hayward@arm.com>
9910 David Sherwood <david.sherwood@arm.com>
9912 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9913 (can_duplicate_and_interleave_p): New function.
9914 (vect_get_and_check_slp_defs): Take the vector of statements
9915 rather than just the current one. Remove excess parentheses.
9916 Restriction rejectinon of vect_constant_def and vect_external_def
9917 for variable-length vectors to boolean types, or types for which
9918 can_duplicate_and_interleave_p is false.
9919 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9920 (duplicate_and_interleave): New function.
9921 (vect_get_constant_vectors): Use gimple_build_vector for
9922 constant-length vectors and suitable variable-length constant
9923 vectors. Use duplicate_and_interleave for other variable-length
9924 vectors. Don't defer the update when inserting new statements.
9926 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9927 Alan Hayward <alan.hayward@arm.com>
9928 David Sherwood <david.sherwood@arm.com>
9930 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9931 min_profitable_iters doesn't go negative.
9933 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9934 Alan Hayward <alan.hayward@arm.com>
9935 David Sherwood <david.sherwood@arm.com>
9937 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9938 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9939 * optabs.def (vec_mask_load_lanes_optab): New optab.
9940 (vec_mask_store_lanes_optab): Likewise.
9941 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9942 (MASK_STORE_LANES): Likewise.
9943 * internal-fn.c (mask_load_lanes_direct): New macro.
9944 (mask_store_lanes_direct): Likewise.
9945 (expand_mask_load_optab_fn): Handle masked operations.
9946 (expand_mask_load_lanes_optab_fn): New macro.
9947 (expand_mask_store_optab_fn): Handle masked operations.
9948 (expand_mask_store_lanes_optab_fn): New macro.
9949 (direct_mask_load_lanes_optab_supported_p): Likewise.
9950 (direct_mask_store_lanes_optab_supported_p): Likewise.
9951 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9953 (vect_load_lanes_supported): Likewise.
9954 * tree-vect-data-refs.c (strip_conversion): New function.
9955 (can_group_stmts_p): Likewise.
9956 (vect_analyze_data_ref_accesses): Use it instead of checking
9957 for a pair of assignments.
9958 (vect_store_lanes_supported): Take a masked_p parameter.
9959 (vect_load_lanes_supported): Likewise.
9960 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9961 vect_store_lanes_supported and vect_load_lanes_supported.
9962 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9963 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9964 parameter. Don't allow gaps for masked accesses.
9965 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9966 and vect_load_lanes_supported.
9967 (get_load_store_type): Take a masked_p parameter and update
9968 call to get_group_load_store_type.
9969 (vectorizable_store): Update call to get_load_store_type.
9970 Handle IFN_MASK_STORE_LANES.
9971 (vectorizable_load): Update call to get_load_store_type.
9972 Handle IFN_MASK_LOAD_LANES.
9974 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9975 Alan Hayward <alan.hayward@arm.com>
9976 David Sherwood <david.sherwood@arm.com>
9978 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9980 * config/aarch64/aarch64-protos.h
9981 (aarch64_sve_struct_memory_operand_p): Declare.
9982 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9983 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9984 (VPRED, vpred): Handle SVE structure modes.
9985 * config/aarch64/constraints.md (Utx): New constraint.
9986 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9987 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9988 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9989 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9990 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9991 structure modes. Split into pieces after RA.
9992 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9993 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9995 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9996 SVE structure modes.
9997 (aarch64_classify_address): Likewise.
9998 (sizetochar): Move earlier in file.
9999 (aarch64_print_operand): Handle SVE register lists.
10000 (aarch64_array_mode): New function.
10001 (aarch64_sve_struct_memory_operand_p): Likewise.
10002 (TARGET_ARRAY_MODE): Redefine.
10004 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10005 Alan Hayward <alan.hayward@arm.com>
10006 David Sherwood <david.sherwood@arm.com>
10008 * target.def (array_mode): New target hook.
10009 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10010 * doc/tm.texi: Regenerate.
10011 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10012 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10013 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10014 targetm.array_mode.
10015 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10018 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10019 Alan Hayward <alan.hayward@arm.com>
10020 David Sherwood <david.sherwood@arm.com>
10022 * fold-const.c (fold_binary_loc): Check the argument types
10023 rather than the result type when testing for a vector operation.
10025 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10027 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10028 * doc/tm.texi: Regenerate.
10030 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10031 Alan Hayward <alan.hayward@arm.com>
10032 David Sherwood <david.sherwood@arm.com>
10034 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10035 (sve): Document new AArch64 extension.
10036 * doc/md.texi (w): Extend the description of the AArch64
10037 constraint to include SVE vectors.
10038 (Upl, Upa): Document new AArch64 predicate constraints.
10039 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10041 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10042 (msve-vector-bits=): New option.
10043 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10044 SVE when these are disabled.
10045 (sve): New extension.
10046 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10047 modes. Adjust their number of units based on aarch64_sve_vg.
10048 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10049 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10050 aarch64_addr_query_type.
10051 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10052 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10053 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10054 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10055 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10056 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10057 (aarch64_simd_imm_zero_p): Delete.
10058 (aarch64_check_zero_based_sve_index_immediate): Declare.
10059 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10060 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10061 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10062 (aarch64_sve_float_mul_immediate_p): Likewise.
10063 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10064 rather than an rtx.
10065 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10066 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10067 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10068 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10069 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10070 (aarch64_regmode_natural_size): Likewise.
10071 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10072 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10074 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10075 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10076 for VG and the SVE predicate registers.
10077 (V_ALIASES): Add a "z"-prefixed alias.
10078 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10079 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10080 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10081 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10082 (REG_CLASS_NAMES): Add entries for them.
10083 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10084 and the predicate registers.
10085 (aarch64_sve_vg): Declare.
10086 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10087 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10088 (REGMODE_NATURAL_SIZE): Define.
10089 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10091 * config/aarch64/aarch64.c: Include cfgrtl.h.
10092 (simd_immediate_info): Add a constructor for series vectors,
10093 and an associated step field.
10094 (aarch64_sve_vg): New variable.
10095 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10096 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10097 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10098 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10099 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10100 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10101 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10102 (aarch64_get_mask_mode): New functions.
10103 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10104 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10105 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10106 predicate modes and predicate registers. Explicitly restrict
10107 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10108 to store a vector mode if it is recognized by
10109 aarch64_classify_vector_mode.
10110 (aarch64_regmode_natural_size): New function.
10111 (aarch64_hard_regno_caller_save_mode): Return the original mode
10113 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10114 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10115 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10116 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10118 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10119 does not overlap dest if the function is frame-related. Handle
10121 (aarch64_split_add_offset): New function.
10122 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10123 them aarch64_add_offset.
10124 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10125 and update call to aarch64_sub_sp.
10126 (aarch64_add_cfa_expression): New function.
10127 (aarch64_expand_prologue): Pass extra temporary registers to the
10128 functions above. Handle the case in which we need to emit new
10129 DW_CFA_expressions for registers that were originally saved
10130 relative to the stack pointer, but now have to be expressed
10131 relative to the frame pointer.
10132 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10134 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10135 IP0 and IP1 values for SVE frames.
10136 (aarch64_expand_vec_series): New function.
10137 (aarch64_expand_sve_widened_duplicate): Likewise.
10138 (aarch64_expand_sve_const_vector): Likewise.
10139 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10140 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10141 into the register, rather than emitting a SET directly.
10142 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10143 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10144 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10145 (offset_9bit_signed_scaled_p): New functions.
10146 (aarch64_replicate_bitmask_imm): New function.
10147 (aarch64_bitmask_imm): Use it.
10148 (aarch64_cannot_force_const_mem): Reject expressions involving
10149 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10150 (aarch64_classify_index): Handle SVE indices, by requiring
10151 a plain register index with a scale that matches the element size.
10152 (aarch64_classify_address): Handle SVE addresses. Assert that
10153 the mode of the address is VOIDmode or an integer mode.
10154 Update call to aarch64_classify_symbol.
10155 (aarch64_classify_symbolic_expression): Update call to
10156 aarch64_classify_symbol.
10157 (aarch64_const_vec_all_in_range_p): New function.
10158 (aarch64_print_vector_float_operand): Likewise.
10159 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10160 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10161 and the FP immediates 1.0 and 0.5.
10162 (aarch64_print_address_internal): Handle SVE addresses.
10163 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10164 (aarch64_regno_regclass): Handle predicate registers.
10165 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10167 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10168 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10169 (aarch64_convert_sve_vector_bits): New function.
10170 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10171 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10172 rather than an rtx.
10173 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10174 Handle SVE vector and predicate modes. Accept VL-based constants
10175 that need only one temporary register, and VL offsets that require
10176 no temporary registers.
10177 (aarch64_conditional_register_usage): Mark the predicate registers
10178 as fixed if SVE isn't available.
10179 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10180 Return true for SVE vector and predicate modes.
10181 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10182 rather than an unsigned int. Handle SVE modes.
10183 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10185 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10187 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10188 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10189 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10190 (aarch64_sve_float_mul_immediate_p): New functions.
10191 (aarch64_sve_valid_immediate): New function.
10192 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10193 Explicitly reject structure modes. Check for INDEX constants.
10194 Handle PTRUE and PFALSE constants.
10195 (aarch64_check_zero_based_sve_index_immediate): New function.
10196 (aarch64_simd_imm_zero_p): Delete.
10197 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10198 vector modes. Accept constants in the range of CNT[BHWD].
10199 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10200 ask for an Advanced SIMD mode.
10201 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10202 (aarch64_simd_vector_alignment): Handle SVE predicates.
10203 (aarch64_vectorize_preferred_vector_alignment): New function.
10204 (aarch64_simd_vector_alignment_reachable): Use it instead of
10206 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10207 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10209 (MAX_VECT_LEN): Delete.
10210 (expand_vec_perm_d): Add a vec_flags field.
10211 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10212 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10213 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10215 (aarch64_evpc_rev): Rename to...
10216 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10217 (aarch64_evpc_rev_global): New function.
10218 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10219 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10221 (aarch64_evpc_sve_tbl): New function.
10222 (aarch64_expand_vec_perm_const_1): Update after rename of
10223 aarch64_evpc_rev. Handle SVE permutes too, trying
10224 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10225 than aarch64_evpc_tbl.
10226 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10227 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10228 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10229 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10230 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10231 (aarch64_expand_sve_vcond): New functions.
10232 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10233 of aarch64_vector_mode_p.
10234 (aarch64_dwarf_poly_indeterminate_value): New function.
10235 (aarch64_compute_pressure_classes): Likewise.
10236 (aarch64_can_change_mode_class): Likewise.
10237 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10238 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10239 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10240 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10241 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10242 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10243 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10244 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10246 (Dn, Dl, Dr): Accept const as well as const_vector.
10247 (Dz): Likewise. Compare against CONST0_RTX.
10248 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10249 of "vector" where appropriate.
10250 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10251 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10252 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10253 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10254 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10255 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10256 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10257 (v_int_equiv): Extend to SVE modes.
10258 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10260 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10261 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10262 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10263 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10264 (SVE_COND_FP_CMP): New int iterators.
10265 (perm_hilo): Handle the new unpack unspecs.
10266 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10268 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10269 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10270 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10271 (aarch64_equality_operator, aarch64_constant_vector_operand)
10272 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10273 (aarch64_sve_nonimmediate_operand): Likewise.
10274 (aarch64_sve_general_operand): Likewise.
10275 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10276 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10277 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10278 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10279 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10280 (aarch64_sve_float_arith_immediate): Likewise.
10281 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10282 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10283 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10284 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10285 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10286 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10287 (aarch64_sve_float_arith_operand): Likewise.
10288 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10289 (aarch64_sve_float_mul_operand): Likewise.
10290 (aarch64_sve_vec_perm_operand): Likewise.
10291 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10292 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10293 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10294 as well as const_vector.
10295 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10296 in file. Use CONST0_RTX and CONSTM1_RTX.
10297 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10298 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10299 Use aarch64_simd_imm_zero.
10300 * config/aarch64/aarch64-sve.md: New file.
10301 * config/aarch64/aarch64.md: Include it.
10302 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10303 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10304 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10305 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10306 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10307 (sve): New attribute.
10308 (enabled): Disable instructions with the sve attribute unless
10310 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10311 aarch64_expand_mov_immediate.
10312 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10313 CNT[BHSD] immediates.
10314 (movti): Split CONST_POLY_INT moves into two halves.
10315 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10316 Split additions that need a temporary here if the destination
10317 is the stack pointer.
10318 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10319 (*add<mode>3_poly_1): New instruction.
10320 (set_clobber_cc): New expander.
10322 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10324 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10325 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10326 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10327 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10328 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10329 Change innermode from fixed_mode_size to machine_mode.
10330 (simplify_subreg): Update call accordingly. Handle a constant-sized
10331 subreg of a variable-length CONST_VECTOR.
10333 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10334 Alan Hayward <alan.hayward@arm.com>
10335 David Sherwood <david.sherwood@arm.com>
10337 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10338 (add_offset_to_base): New function, split out from...
10339 (create_mem_ref): ...here. When handling a scale other than 1,
10340 check first whether the address is valid without the offset.
10341 Add it into the base if so, leaving the index and scale as-is.
10343 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10346 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10347 fold_for_warn before checking if arg2 is INTEGER_CST.
10349 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10351 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10352 (store_multiple_operation): Delete.
10353 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10354 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10355 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10356 guarded by TARGET_STRING.
10357 (rs6000_output_load_multiple): Delete.
10358 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10359 OPTION_MASK_STRING / TARGET_STRING handling.
10360 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10361 (const rs6000_opt_masks) <"string">: Change mask to 0.
10362 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10363 (MASK_STRING): Delete.
10364 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10366 (load_multiple): Delete.
10373 (store_multiple): Delete.
10380 (movmemsi_8reg): Delete.
10381 (corresponding unnamed define_insn): Delete.
10382 (movmemsi_6reg): Delete.
10383 (corresponding unnamed define_insn): Delete.
10384 (movmemsi_4reg): Delete.
10385 (corresponding unnamed define_insn): Delete.
10386 (movmemsi_2reg): Delete.
10387 (corresponding unnamed define_insn): Delete.
10388 (movmemsi_1reg): Delete.
10389 (corresponding unnamed define_insn): Delete.
10390 * config/rs6000/rs6000.opt (mno-string): New.
10391 (mstring): Replace by deprecation warning stub.
10392 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10394 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10396 * regrename.c (regrename_do_replace): If replacing the same
10397 reg multiple times, try to reuse last created gen_raw_REG.
10400 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10401 main to workaround a bug in GDB.
10403 2018-01-12 Tom de Vries <tom@codesourcery.com>
10406 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10408 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10410 PR rtl-optimization/80481
10411 * ira-color.c (get_cap_member): New function.
10412 (allocnos_conflict_by_live_ranges_p): Use it.
10413 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10414 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10416 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10419 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10420 (*saddl_se_1): Ditto.
10421 (*ssubsi_1): Ditto.
10422 (*ssubl_se_1): Ditto.
10424 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10426 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10427 rather than wi::to_widest for DR_INITs.
10428 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10429 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10430 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10432 (vect_analyze_group_access_1): Note that here.
10434 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10436 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10437 polynomial type sizes.
10439 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10441 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10442 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10443 (gimple_add_tmp_var): Likewise.
10445 2018-01-12 Martin Liska <mliska@suse.cz>
10447 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10448 (gimple_alloc_sizes): Likewise.
10449 (dump_gimple_statistics): Use PRIu64 in printf format.
10450 * gimple.h: Change uint64_t to int.
10452 2018-01-12 Martin Liska <mliska@suse.cz>
10454 * tree-core.h: Use uint64_t instead of int.
10455 * tree.c (tree_node_counts): Likewise.
10456 (tree_node_sizes): Likewise.
10457 (dump_tree_statistics): Use PRIu64 in printf format.
10459 2018-01-12 Martin Liska <mliska@suse.cz>
10461 * Makefile.in: As qsort_chk is implemented in vec.c, add
10462 vec.o to linkage of gencfn-macros.
10463 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10464 passing the info to record_node_allocation_statistics.
10465 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10467 * ggc-common.c (struct ggc_usage): Add operator== and use
10468 it in operator< and compare function.
10469 * mem-stats.h (struct mem_usage): Likewise.
10470 * vec.c (struct vec_usage): Remove operator< and compare
10471 function. Can be simply inherited.
10473 2018-01-12 Martin Jambor <mjambor@suse.cz>
10476 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10477 * tree-ssa-math-opts.c: Include domwalk.h.
10478 (convert_mult_to_fma_1): New function.
10479 (fma_transformation_info): New type.
10480 (fma_deferring_state): Likewise.
10481 (cancel_fma_deferring): New function.
10482 (result_of_phi): Likewise.
10483 (last_fma_candidate_feeds_initial_phi): Likewise.
10484 (convert_mult_to_fma): Added deferring logic, split actual
10485 transformation to convert_mult_to_fma_1.
10486 (math_opts_dom_walker): New type.
10487 (math_opts_dom_walker::after_dom_children): New method, body moved
10488 here from pass_optimize_widening_mul::execute, added deferring logic
10490 (pass_optimize_widening_mul::execute): Moved most of code to
10491 math_opts_dom_walker::after_dom_children.
10492 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10493 * config/i386/i386.c (ix86_option_override_internal): Added
10494 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10496 2018-01-12 Richard Biener <rguenther@suse.de>
10499 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10500 inline instance vars.
10502 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10505 * config/rx/rx.c (rx_is_restricted_memory_address):
10506 Handle SUBREG case.
10508 2018-01-12 Richard Biener <rguenther@suse.de>
10510 PR tree-optimization/80846
10511 * target.def (split_reduction): New target hook.
10512 * targhooks.c (default_split_reduction): New function.
10513 * targhooks.h (default_split_reduction): Declare.
10514 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10515 target requests first reduce vectors by combining low and high
10517 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10518 (get_vectype_for_scalar_type_and_size): Export.
10519 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10520 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10521 * doc/tm.texi: Regenerate.
10522 * config/i386/i386.c (ix86_split_reduction): Implement
10523 TARGET_VECTORIZE_SPLIT_REDUCTION.
10525 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10528 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10529 in PIC mode except for TARGET_VXWORKS_RTP.
10530 * config/sparc/sparc.c: Include cfgrtl.h.
10531 (TARGET_INIT_PIC_REG): Define.
10532 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10533 (sparc_pic_register_p): New predicate.
10534 (sparc_legitimate_address_p): Use it.
10535 (sparc_legitimize_pic_address): Likewise.
10536 (sparc_delegitimize_address): Likewise.
10537 (sparc_mode_dependent_address_p): Likewise.
10538 (gen_load_pcrel_sym): Remove 4th parameter.
10539 (load_got_register): Adjust call to above. Remove obsolete stuff.
10540 (sparc_expand_prologue): Do not call load_got_register here.
10541 (sparc_flat_expand_prologue): Likewise.
10542 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10543 (sparc_use_pseudo_pic_reg): New function.
10544 (sparc_init_pic_reg): Likewise.
10545 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10546 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10548 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10550 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10551 Add item for branch_cost.
10553 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10555 PR rtl-optimization/83565
10556 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10557 not extend the result to a larger mode for rotate operations.
10558 (num_sign_bit_copies1): Likewise.
10560 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10563 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10565 Use values-Xc.o for -pedantic.
10566 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10568 2018-01-12 Martin Liska <mliska@suse.cz>
10571 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10573 (possible_polymorphic_call_targets): Use it.
10574 (ipa_devirt): Likewise.
10576 2018-01-12 Martin Liska <mliska@suse.cz>
10578 * profile-count.h (enum profile_quality): Use 0 as invalid
10579 enum value of profile_quality.
10581 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10583 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10584 -mext-string options.
10586 2018-01-12 Richard Biener <rguenther@suse.de>
10588 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10589 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10590 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10592 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10594 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10596 * configure.ac (--with-long-double-format): Add support for the
10597 configuration option to change the default long double format on
10599 * config.gcc (powerpc*-linux*-*): Likewise.
10600 * configure: Regenerate.
10601 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10602 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10603 used without modification.
10605 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10607 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10608 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10609 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10610 MISC_BUILTIN_SPEC_BARRIER.
10611 (rs6000_init_builtins): Likewise.
10612 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10614 (speculation_barrier): New define_insn.
10615 * doc/extend.texi: Document __builtin_speculation_barrier.
10617 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10620 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10621 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10622 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10624 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10625 integral modes instead of "ss" and "sd".
10626 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10627 vectors with 32-bit and 64-bit elements.
10628 (vecdupssescalarmodesuffix): New mode attribute.
10629 (vec_dup<mode>): Use it.
10631 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10634 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10635 frame if argument is passed on stack.
10637 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10640 * ree.c (combine_reaching_defs): Optimize also
10641 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10642 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10644 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10646 PR middle-end/83189
10647 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10649 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10651 PR middle-end/83718
10652 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10653 after they are computed.
10655 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10657 PR tree-optimization/83695
10658 * gimple-loop-linterchange.cc
10659 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10660 reset cached scev information after interchange.
10661 (pass_linterchange::execute): Remove call to scev_reset_htab.
10663 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10665 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10666 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10667 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10668 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10669 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10670 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10671 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10672 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10673 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10674 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10675 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10676 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10677 (V_lane_reg): Likewise.
10678 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10680 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10681 (vfmal_lane_low<mode>_intrinsic,
10682 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10683 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10684 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10685 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10686 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10687 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10689 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10691 * config/arm/arm-cpus.in (fp16fml): New feature.
10692 (ALL_SIMD): Add fp16fml.
10693 (armv8.2-a): Add fp16fml as an option.
10694 (armv8.3-a): Likewise.
10695 (armv8.4-a): Add fp16fml as part of fp16.
10696 * config/arm/arm.h (TARGET_FP16FML): Define.
10697 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10699 * config/arm/arm-modes.def (V2HF): Define.
10700 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10701 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10702 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10703 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10704 vfmsl_low, vfmsl_high): New set of builtins.
10705 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10706 (vfml_op): New code attribute.
10707 (VFMLHALVES): New int iterator.
10708 (VFML, VFMLSEL): New mode attributes.
10709 (V_reg): Define mapping for V2HF.
10710 (V_hi, V_lo): New mode attributes.
10711 (VF_constraint): Likewise.
10712 (vfml_half, vfml_half_selector): New int attributes.
10713 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10715 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10716 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10718 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10719 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10720 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10721 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10723 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10724 Document new effective target and option set.
10726 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10728 * config/arm/arm-cpus.in (armv8_4): New feature.
10729 (ARMv8_4a): New fgroup.
10730 (armv8.4-a): New arch.
10731 * config/arm/arm-tables.opt: Regenerate.
10732 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10733 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10734 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10735 Add matching rules for -march=armv8.4-a and extensions.
10736 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10738 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10741 * config/rx/rx.md (BW): New mode attribute.
10742 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10744 2018-01-11 Richard Biener <rguenther@suse.de>
10746 PR tree-optimization/83435
10747 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10748 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10749 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10751 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10752 Alan Hayward <alan.hayward@arm.com>
10753 David Sherwood <david.sherwood@arm.com>
10755 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10757 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10758 (aarch64_print_address_internal): Use it to check for a zero offset.
10760 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10761 Alan Hayward <alan.hayward@arm.com>
10762 David Sherwood <david.sherwood@arm.com>
10764 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10765 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10766 Return a poly_int64 rather than a HOST_WIDE_INT.
10767 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10768 rather than a HOST_WIDE_INT.
10769 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10770 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10771 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10772 final_offset from HOST_WIDE_INT to poly_int64.
10773 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10774 to_constant when getting the number of units in an Advanced SIMD
10776 (aarch64_builtin_vectorized_function): Check for a constant number
10778 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10780 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10781 attribute instead of GET_MODE_NUNITS.
10782 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10783 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10784 GET_MODE_SIZE for fixed-size registers.
10785 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10786 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10787 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10788 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10789 (aarch64_print_operand, aarch64_print_address_internal)
10790 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10791 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10792 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10793 Handle polynomial GET_MODE_SIZE.
10794 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10795 wider than SImode without modification.
10796 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10797 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10798 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10799 passing and returning SVE modes.
10800 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10801 rather than GEN_INT.
10802 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10803 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10804 (aarch64_allocate_and_probe_stack_space): Likewise.
10805 (aarch64_layout_frame): Cope with polynomial offsets.
10806 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10807 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10808 polynomial offsets.
10809 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10810 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10811 poly_int64 rather than a HOST_WIDE_INT.
10812 (aarch64_get_separate_components, aarch64_process_components)
10813 (aarch64_expand_prologue, aarch64_expand_epilogue)
10814 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10815 (aarch64_anchor_offset): New function, split out from...
10816 (aarch64_legitimize_address): ...here.
10817 (aarch64_builtin_vectorization_cost): Handle polynomial
10818 TYPE_VECTOR_SUBPARTS.
10819 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10821 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10822 number of elements from the PARALLEL rather than the mode.
10823 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10824 rather than GET_MODE_BITSIZE.
10825 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10826 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10827 (aarch64_expand_vec_perm_const_1): Handle polynomial
10828 d->perm.length () and d->perm elements.
10829 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10830 Apply to_constant to d->perm elements.
10831 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10832 polynomial CONST_VECTOR_NUNITS.
10833 (aarch64_move_pointer): Take amount as a poly_int64 rather
10835 (aarch64_progress_pointer): Avoid temporary variable.
10836 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10837 the mode attribute instead of GET_MODE.
10839 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10840 Alan Hayward <alan.hayward@arm.com>
10841 David Sherwood <david.sherwood@arm.com>
10843 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10844 x exists before using it.
10845 (aarch64_add_constant_internal): Rename to...
10846 (aarch64_add_offset_1): ...this. Replace regnum with separate
10847 src and dest rtxes. Handle the case in which they're different,
10848 including when the offset is zero. Replace scratchreg with an rtx.
10849 Use 2 additions if there is no spare register into which we can
10850 move a 16-bit constant.
10851 (aarch64_add_constant): Delete.
10852 (aarch64_add_offset): Replace reg with separate src and dest
10853 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10854 Use aarch64_add_offset_1.
10855 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10856 an rtx rather than an int. Take the delta as a poly_int64
10857 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10858 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10859 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10860 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10861 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10862 and aarch64_add_sp.
10863 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10864 aarch64_add_constant.
10866 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10868 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10869 Use scalar_float_mode.
10871 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10873 * config/aarch64/aarch64-simd.md
10874 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10875 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10876 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10877 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10878 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10879 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10880 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10881 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10882 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10883 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10885 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10888 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10889 targ_options->x_arm_arch_string is non NULL.
10891 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10893 * config/aarch64/aarch64.h
10894 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10896 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10899 * expmed.c (emit_store_flag_force): Swap if const op0
10900 and change VOIDmode to mode of op0.
10902 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10904 PR rtl-optimization/83761
10905 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10906 than bytes to mode_for_size.
10908 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10910 PR middle-end/83189
10911 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10912 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10915 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10917 PR middle-end/83575
10918 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10919 when in layout mode.
10920 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10921 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10924 2018-01-10 Michael Collison <michael.collison@arm.com>
10926 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10927 * config/aarch64/aarch64-option-extension.def: Add
10928 AARCH64_OPT_EXTENSION of 'fp16fml'.
10929 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10930 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10931 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10932 * config/aarch64/constraints.md (Ui7): New constraint.
10933 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10934 (VFMLA_SEL_W): Ditto.
10937 (VFMLA16_LOW): New int iterator.
10938 (VFMLA16_HIGH): Ditto.
10939 (UNSPEC_FMLAL): New unspec.
10940 (UNSPEC_FMLSL): Ditto.
10941 (UNSPEC_FMLAL2): Ditto.
10942 (UNSPEC_FMLSL2): Ditto.
10943 (f16mac): New code attribute.
10944 * config/aarch64/aarch64-simd-builtins.def
10945 (aarch64_fmlal_lowv2sf): Ditto.
10946 (aarch64_fmlsl_lowv2sf): Ditto.
10947 (aarch64_fmlalq_lowv4sf): Ditto.
10948 (aarch64_fmlslq_lowv4sf): Ditto.
10949 (aarch64_fmlal_highv2sf): Ditto.
10950 (aarch64_fmlsl_highv2sf): Ditto.
10951 (aarch64_fmlalq_highv4sf): Ditto.
10952 (aarch64_fmlslq_highv4sf): Ditto.
10953 (aarch64_fmlal_lane_lowv2sf): Ditto.
10954 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10955 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10956 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10957 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10958 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10959 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10960 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10961 (aarch64_fmlal_lane_highv2sf): Ditto.
10962 (aarch64_fmlsl_lane_highv2sf): Ditto.
10963 (aarch64_fmlal_laneq_highv2sf): Ditto.
10964 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10965 (aarch64_fmlalq_lane_highv4sf): Ditto.
10966 (aarch64_fmlsl_lane_highv4sf): Ditto.
10967 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10968 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10969 * config/aarch64/aarch64-simd.md:
10970 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10971 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10972 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10973 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10974 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10975 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10976 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10977 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10978 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10979 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10980 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10981 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10982 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10983 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10984 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10985 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10986 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10987 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10988 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10989 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10990 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10991 (vfmlsl_low_u32): Ditto.
10992 (vfmlalq_low_u32): Ditto.
10993 (vfmlslq_low_u32): Ditto.
10994 (vfmlal_high_u32): Ditto.
10995 (vfmlsl_high_u32): Ditto.
10996 (vfmlalq_high_u32): Ditto.
10997 (vfmlslq_high_u32): Ditto.
10998 (vfmlal_lane_low_u32): Ditto.
10999 (vfmlsl_lane_low_u32): Ditto.
11000 (vfmlal_laneq_low_u32): Ditto.
11001 (vfmlsl_laneq_low_u32): Ditto.
11002 (vfmlalq_lane_low_u32): Ditto.
11003 (vfmlslq_lane_low_u32): Ditto.
11004 (vfmlalq_laneq_low_u32): Ditto.
11005 (vfmlslq_laneq_low_u32): Ditto.
11006 (vfmlal_lane_high_u32): Ditto.
11007 (vfmlsl_lane_high_u32): Ditto.
11008 (vfmlal_laneq_high_u32): Ditto.
11009 (vfmlsl_laneq_high_u32): Ditto.
11010 (vfmlalq_lane_high_u32): Ditto.
11011 (vfmlslq_lane_high_u32): Ditto.
11012 (vfmlalq_laneq_high_u32): Ditto.
11013 (vfmlslq_laneq_high_u32): Ditto.
11014 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11015 (AARCH64_FL_FOR_ARCH8_4): New.
11016 (AARCH64_ISA_F16FML): New ISA flag.
11017 (TARGET_F16FML): New feature flag for fp16fml.
11018 (doc/invoke.texi): Document new fp16fml option.
11020 2018-01-10 Michael Collison <michael.collison@arm.com>
11022 * config/aarch64/aarch64-builtins.c:
11023 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11024 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11025 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11026 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11027 (AARCH64_ISA_SHA3): New ISA flag.
11028 (TARGET_SHA3): New feature flag for sha3.
11029 * config/aarch64/iterators.md (sha512_op): New int attribute.
11030 (CRYPTO_SHA512): New int iterator.
11031 (UNSPEC_SHA512H): New unspec.
11032 (UNSPEC_SHA512H2): Ditto.
11033 (UNSPEC_SHA512SU0): Ditto.
11034 (UNSPEC_SHA512SU1): Ditto.
11035 * config/aarch64/aarch64-simd-builtins.def
11036 (aarch64_crypto_sha512hqv2di): New builtin.
11037 (aarch64_crypto_sha512h2qv2di): Ditto.
11038 (aarch64_crypto_sha512su0qv2di): Ditto.
11039 (aarch64_crypto_sha512su1qv2di): Ditto.
11040 (aarch64_eor3qv8hi): Ditto.
11041 (aarch64_rax1qv2di): Ditto.
11042 (aarch64_xarqv2di): Ditto.
11043 (aarch64_bcaxqv8hi): Ditto.
11044 * config/aarch64/aarch64-simd.md:
11045 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11046 (aarch64_crypto_sha512su0qv2di): Ditto.
11047 (aarch64_crypto_sha512su1qv2di): Ditto.
11048 (aarch64_eor3qv8hi): Ditto.
11049 (aarch64_rax1qv2di): Ditto.
11050 (aarch64_xarqv2di): Ditto.
11051 (aarch64_bcaxqv8hi): Ditto.
11052 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11053 (vsha512h2q_u64): Ditto.
11054 (vsha512su0q_u64): Ditto.
11055 (vsha512su1q_u64): Ditto.
11056 (veor3q_u16): Ditto.
11057 (vrax1q_u64): Ditto.
11058 (vxarq_u64): Ditto.
11059 (vbcaxq_u16): Ditto.
11060 * config/arm/types.md (crypto_sha512): New type attribute.
11061 (crypto_sha3): Ditto.
11062 (doc/invoke.texi): Document new sha3 option.
11064 2018-01-10 Michael Collison <michael.collison@arm.com>
11066 * config/aarch64/aarch64-builtins.c:
11067 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11068 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11069 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11070 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11071 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11072 (AARCH64_ISA_SM4): New ISA flag.
11073 (TARGET_SM4): New feature flag for sm4.
11074 * config/aarch64/aarch64-simd-builtins.def
11075 (aarch64_sm3ss1qv4si): Ditto.
11076 (aarch64_sm3tt1aq4si): Ditto.
11077 (aarch64_sm3tt1bq4si): Ditto.
11078 (aarch64_sm3tt2aq4si): Ditto.
11079 (aarch64_sm3tt2bq4si): Ditto.
11080 (aarch64_sm3partw1qv4si): Ditto.
11081 (aarch64_sm3partw2qv4si): Ditto.
11082 (aarch64_sm4eqv4si): Ditto.
11083 (aarch64_sm4ekeyqv4si): Ditto.
11084 * config/aarch64/aarch64-simd.md:
11085 (aarch64_sm3ss1qv4si): Ditto.
11086 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11087 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11088 (aarch64_sm4eqv4si): Ditto.
11089 (aarch64_sm4ekeyqv4si): Ditto.
11090 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11091 (sm3part_op): Ditto.
11092 (CRYPTO_SM3TT): Ditto.
11093 (CRYPTO_SM3PART): Ditto.
11094 (UNSPEC_SM3SS1): New unspec.
11095 (UNSPEC_SM3TT1A): Ditto.
11096 (UNSPEC_SM3TT1B): Ditto.
11097 (UNSPEC_SM3TT2A): Ditto.
11098 (UNSPEC_SM3TT2B): Ditto.
11099 (UNSPEC_SM3PARTW1): Ditto.
11100 (UNSPEC_SM3PARTW2): Ditto.
11101 (UNSPEC_SM4E): Ditto.
11102 (UNSPEC_SM4EKEY): Ditto.
11103 * config/aarch64/constraints.md (Ui2): New constraint.
11104 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11105 * config/arm/types.md (crypto_sm3): New type attribute.
11106 (crypto_sm4): Ditto.
11107 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11108 (vsm3tt1aq_u32): Ditto.
11109 (vsm3tt1bq_u32): Ditto.
11110 (vsm3tt2aq_u32): Ditto.
11111 (vsm3tt2bq_u32): Ditto.
11112 (vsm3partw1q_u32): Ditto.
11113 (vsm3partw2q_u32): Ditto.
11114 (vsm4eq_u32): Ditto.
11115 (vsm4ekeyq_u32): Ditto.
11116 (doc/invoke.texi): Document new sm4 option.
11118 2018-01-10 Michael Collison <michael.collison@arm.com>
11120 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11121 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11122 (AARCH64_FL_FOR_ARCH8_4): New.
11123 (AARCH64_FL_V8_4): New flag.
11124 (doc/invoke.texi): Document new armv8.4-a option.
11126 2018-01-10 Michael Collison <michael.collison@arm.com>
11128 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11129 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11130 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11131 * config/aarch64/aarch64-option-extension.def: Add
11132 AARCH64_OPT_EXTENSION of 'sha2'.
11133 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11134 (crypto): Disable sha2 and aes if crypto disabled.
11135 (crypto): Enable aes and sha2 if enabled.
11136 (simd): Disable sha2 and aes if simd disabled.
11137 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11139 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11140 (TARGET_SHA2): New feature flag for sha2.
11141 (TARGET_AES): New feature flag for aes.
11142 * config/aarch64/aarch64-simd.md:
11143 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11144 conditional on TARGET_AES.
11145 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11146 (aarch64_crypto_sha1hsi): Make pattern conditional
11148 (aarch64_crypto_sha1hv4si): Ditto.
11149 (aarch64_be_crypto_sha1hv4si): Ditto.
11150 (aarch64_crypto_sha1su1v4si): Ditto.
11151 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11152 (aarch64_crypto_sha1su0v4si): Ditto.
11153 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11154 (aarch64_crypto_sha256su0v4si): Ditto.
11155 (aarch64_crypto_sha256su1v4si): Ditto.
11156 (doc/invoke.texi): Document new aes and sha2 options.
11158 2018-01-10 Martin Sebor <msebor@redhat.com>
11160 PR tree-optimization/83781
11161 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11164 2018-01-11 Martin Sebor <msebor@gmail.com>
11165 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11167 PR tree-optimization/83501
11168 PR tree-optimization/81703
11170 * tree-ssa-strlen.c (get_string_cst): Rename...
11171 (get_string_len): ...to this. Handle global constants.
11172 (handle_char_store): Adjust.
11174 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11175 Jim Wilson <jimw@sifive.com>
11177 * config/riscv/riscv-protos.h (riscv_output_return): New.
11178 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11179 (riscv_attribute_table, riscv_output_return),
11180 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11181 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11182 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11184 (riscv_expand_prologue): Add early return for naked function.
11185 (riscv_expand_epilogue): Likewise.
11186 (riscv_function_ok_for_sibcall): Return false for naked function.
11187 (riscv_set_current_function): New.
11188 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11189 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11190 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11191 * doc/extend.texi (RISC-V Function Attributes): New.
11193 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11195 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11196 check for 128-bit long double before checking TCmode.
11197 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11198 128-bit long doubles before checking TFmode or TCmode.
11199 (FLOAT128_IBM_P): Likewise.
11201 2018-01-10 Martin Sebor <msebor@redhat.com>
11203 PR tree-optimization/83671
11204 * builtins.c (c_strlen): Unconditionally return zero for the empty
11206 Use -Warray-bounds for warnings.
11207 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11208 for non-constant array indices with COMPONENT_REF, arrays of
11209 arrays, and pointers to arrays.
11210 (gimple_fold_builtin_strlen): Determine and set length range for
11211 non-constant character arrays.
11213 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11215 PR middle-end/81897
11216 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11219 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11221 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11223 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11226 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11227 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11228 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11229 indexed_or_indirect_operand predicate.
11230 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11231 (*vsx_le_perm_load_v8hi): Likewise.
11232 (*vsx_le_perm_load_v16qi): Likewise.
11233 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11234 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11235 (*vsx_le_perm_store_v8hi): Likewise.
11236 (*vsx_le_perm_store_v16qi): Likewise.
11237 (eight unnamed splitters): Likewise.
11239 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11241 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11242 * config/rs6000/emmintrin.h: Likewise.
11243 * config/rs6000/mmintrin.h: Likewise.
11244 * config/rs6000/xmmintrin.h: Likewise.
11246 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11249 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11251 * tree.c (tree_nop_conversion): Return true for location wrapper
11253 (maybe_wrap_with_location): New function.
11254 (selftest::check_strip_nops): New function.
11255 (selftest::test_location_wrappers): New function.
11256 (selftest::tree_c_tests): Call it.
11257 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11258 (maybe_wrap_with_location): New decl.
11259 (EXPR_LOCATION_WRAPPER_P): New macro.
11260 (location_wrapper_p): New inline function.
11261 (tree_strip_any_location_wrapper): New inline function.
11263 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11266 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11267 stack_realign_offset for the largest alignment of stack slot
11269 (ix86_find_max_used_stack_alignment): New function.
11270 (ix86_finalize_stack_frame_flags): Use it. Set
11271 max_used_stack_alignment if we don't realign stack.
11272 * config/i386/i386.h (machine_function): Add
11273 max_used_stack_alignment.
11275 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11277 * config/arm/arm.opt (-mbranch-cost): New option.
11278 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11281 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11284 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11285 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11287 2018-01-10 Richard Biener <rguenther@suse.de>
11290 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11291 early out so it also covers the case where we have a non-NULL
11294 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11296 PR tree-optimization/83753
11297 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11298 for non-strided grouped accesses if the number of elements is 1.
11300 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11303 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11304 * i386.h (TARGET_USE_GATHER): Define.
11305 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11307 2018-01-10 Martin Liska <mliska@suse.cz>
11310 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11311 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11313 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11314 CLEANUP_NO_PARTITIONING is not set.
11316 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11318 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11319 for vectors, as a partial revert of r254296.
11320 * rtl.h (const_vec_p): Delete.
11321 (const_vec_duplicate_p): Don't test for vector CONSTs.
11322 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11323 * expmed.c (make_tree): Likewise.
11326 * common.md (E, F): Use CONSTANT_P instead of checking for
11328 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11329 checking for CONST_VECTOR.
11331 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11333 PR middle-end/83575
11334 * predict.c (force_edge_cold): Handle in more sane way edges
11335 with no prediction.
11337 2018-01-09 Carl Love <cel@us.ibm.com>
11339 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11341 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11342 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11343 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11344 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11345 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11346 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11347 * config/rs6000/rs6000-protos.h: Add extern defition for
11348 rs6000_generate_float2_double_code.
11349 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11351 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11352 (float2_v2df): Add define_expand.
11354 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11357 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11358 op_mode in the force_to_mode call.
11360 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11362 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11363 instead of checking each element individually.
11364 (aarch64_evpc_uzp): Likewise.
11365 (aarch64_evpc_zip): Likewise.
11366 (aarch64_evpc_ext): Likewise.
11367 (aarch64_evpc_rev): Likewise.
11368 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11369 instead of checking each element individually. Return true without
11371 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11372 whether all selected elements come from the same input, instead of
11373 checking each element individually. Remove calls to gen_rtx_REG,
11374 start_sequence and end_sequence and instead assert that no rtl is
11377 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11379 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11380 order of HIGH and CONST checks.
11382 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11384 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11385 if the destination isn't an SSA_NAME.
11387 2018-01-09 Richard Biener <rguenther@suse.de>
11389 PR tree-optimization/83668
11390 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11392 (canonicalize_loop_form): ... here, renamed from ...
11393 (canonicalize_loop_closed_ssa_form): ... this and amended to
11394 swap successor edges for loop exit blocks to make us use
11395 the RPO order we need for initial schedule generation.
11397 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11399 PR tree-optimization/64811
11400 * match.pd: When optimizing comparisons with Inf, avoid
11401 introducing or losing exceptions from comparisons with NaN.
11403 2018-01-09 Martin Liska <mliska@suse.cz>
11406 * asan.c (shadow_mem_size): Add gcc_assert.
11408 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11410 Don't save registers in main().
11413 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11414 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11415 * config/avr/avr.c (avr_set_current_function): Don't error if
11416 naked, OS_task or OS_main are specified at the same time.
11417 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11419 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11421 * common/config/avr/avr-common.c (avr_option_optimization_table):
11422 Switch on -mmain-is-OS_task for optimizing compilations.
11424 2018-01-09 Richard Biener <rguenther@suse.de>
11426 PR tree-optimization/83572
11427 * graphite.c: Include cfganal.h.
11428 (graphite_transform_loops): Connect infinite loops to exit
11429 and remove fake edges at the end.
11431 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11433 * ipa-inline.c (edge_badness): Revert accidental checkin.
11435 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11438 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11439 symbols; not inline clones.
11441 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11444 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11445 hard registers. Formatting fixes.
11447 PR preprocessor/83722
11448 * gcc.c (try_generate_repro): Pass
11449 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11450 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11453 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11454 Kito Cheng <kito.cheng@gmail.com>
11456 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11457 (riscv_leaf_function_p): Delete.
11458 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11460 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11462 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11464 (do_ifelse): New function.
11465 (do_isel): New function.
11466 (do_sub3): New function.
11467 (do_add3): New function.
11468 (do_load_mask_compare): New function.
11469 (do_overlap_load_compare): New function.
11470 (expand_compare_loop): New function.
11471 (expand_block_compare): Call expand_compare_loop() when appropriate.
11472 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11473 option description.
11474 (-mblock-compare-inline-loop-limit): New option.
11476 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11479 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11480 Reverse order of second and third operands in first alternative.
11481 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11482 of first and second elements in UNSPEC_VPERMR vector.
11483 (altivec_expand_vec_perm_le): Likewise.
11485 2018-01-08 Jeff Law <law@redhat.com>
11487 PR rtl-optimizatin/81308
11488 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11489 (process_switch): If group_case_labels makes a change, then set
11491 (pass_convert_switch::execute): If a switch is converted, then
11492 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11494 PR rtl-optimization/81308
11495 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11498 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11500 PR target/83663 - Revert r255946
11501 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11502 generation for cases where splatting a value is not useful.
11503 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11504 across a vec_duplicate and a paradoxical subreg forming a vector
11505 mode to a vec_concat.
11507 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11509 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11510 -march=armv8.3-a variants.
11511 * config/arm/t-multilib: Likewise.
11512 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11514 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11516 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11518 (cceq_ior_compare_complement): Give it a name so I can use it, and
11519 change boolean_or_operator predicate to boolean_operator so it can
11520 be used to generate a crand.
11521 (eqne): New code iterator.
11522 (bd/bd_neg): New code_attrs.
11523 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11524 a single define_insn.
11525 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11526 decrement (bdnzt/bdnzf/bdzt/bdzf).
11527 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11528 with the new names of the branch decrement patterns, and added the
11529 names of the branch decrement conditional patterns.
11531 2018-01-08 Richard Biener <rguenther@suse.de>
11533 PR tree-optimization/83563
11534 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11537 2018-01-08 Richard Biener <rguenther@suse.de>
11539 PR middle-end/83713
11540 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11542 2018-01-08 Richard Biener <rguenther@suse.de>
11544 PR tree-optimization/83685
11545 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11546 references to abnormals.
11548 2018-01-08 Richard Biener <rguenther@suse.de>
11551 * dwarf2out.c (output_indirect_strings): Handle empty
11552 skeleton_debug_str_hash.
11553 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11555 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11557 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11558 (emit_store_direct): Likewise.
11559 (arc_trampoline_adjust_address): Likewise.
11560 (arc_asm_trampoline_template): New function.
11561 (arc_initialize_trampoline): Use asm_trampoline_template.
11562 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11563 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11564 * config/arc/arc.md (flush_icache): Delete pattern.
11566 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11568 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11569 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11572 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11575 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11576 by not USED_FOR_TARGET.
11577 (make_pass_resolve_sw_modes): Likewise.
11579 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11581 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11584 2018-01-08 Richard Biener <rguenther@suse.de>
11586 PR middle-end/83580
11587 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11589 2018-01-08 Richard Biener <rguenther@suse.de>
11591 PR middle-end/83517
11592 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11594 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11596 PR middle-end/81897
11597 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11598 basic blocks with a small number of successors.
11599 (convert_control_dep_chain_into_preds): Improve handling of
11601 (dump_predicates): Split apart into...
11602 (dump_pred_chain): ...here...
11603 (dump_pred_info): ...and here.
11604 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11605 (can_chain_union_be_invalidated_p): Improve check for invalidation
11607 (uninit_uses_cannot_happen): Avoid unnecessary if
11608 convert_control_dep_chain_into_preds yielded nothing.
11610 2018-01-06 Martin Sebor <msebor@redhat.com>
11612 PR tree-optimization/83640
11613 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11614 subtracting negative offset from size.
11615 (builtin_access::overlap): Adjust offset bounds of the access to fall
11616 within the size of the object if possible.
11618 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11620 PR rtl-optimization/83699
11621 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11622 extract_bit_field_as_subreg to cases in which the extracted
11623 value is also a vector.
11625 * lra-constraints.c (process_alt_operands): Test for the equivalence
11626 substitutions when detecting a possible reload cycle.
11628 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11631 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11632 by default if flag_selective_schedling{,2}. Formatting fixes.
11634 PR rtl-optimization/83682
11635 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11636 if it has non-VECTOR_MODE element mode.
11637 (vec_duplicate_p): Likewise.
11639 PR middle-end/83694
11640 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11641 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11643 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11646 * config/i386/i386-builtin.def
11647 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11648 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11649 Require also OPTION_MASK_ISA_AVX512F in addition to
11650 OPTION_MASK_ISA_GFNI.
11651 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11652 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11653 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11654 to OPTION_MASK_ISA_GFNI.
11655 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11656 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11657 OPTION_MASK_ISA_AVX512BW.
11658 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11659 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11660 addition to OPTION_MASK_ISA_GFNI.
11661 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11662 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11663 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11664 to OPTION_MASK_ISA_GFNI.
11665 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11666 a requirement for all ISAs rather than any of them with a few
11668 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11670 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11671 bitmasks to be enabled with 3 exceptions, instead of requiring any
11672 enabled ISA with lots of exceptions.
11673 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11674 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11675 Change avx512bw in isa attribute to avx512f.
11676 * config/i386/sgxintrin.h: Add license boilerplate.
11677 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11678 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11679 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11680 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11682 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11683 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11684 temporarily sse2 rather than sse if not enabled already.
11687 * config/i386/sse.md (VI248_VLBW): Rename to ...
11688 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11689 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11690 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11691 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11692 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11693 mode iterator instead of VI248_VLBW.
11695 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11697 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11698 (record_modified): Skip clobbers; add debug output.
11699 (param_change_prob): Use sreal frequencies.
11701 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11703 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11704 punt for user-aligned variables.
11706 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11708 * tree-chrec.c (chrec_contains_symbols): Return true for
11711 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11714 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11715 of (x|y) == x for BICS pattern.
11717 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11719 PR tree-optimization/83605
11720 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11721 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11724 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11726 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11727 * config/epiphany/rtems.h: New file.
11729 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11730 Uros Bizjak <ubizjak@gmail.com>
11733 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11734 QIreg_operand instead of register_operand predicate.
11735 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11736 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11737 comments instead of -fmitigate[-_]rop.
11739 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11742 * cgraphunit.c (symbol_table::compile): Switch to text_section
11743 before calling assembly_start debug hook.
11744 * run-rtl-passes.c (run_rtl_passes): Likewise.
11747 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11749 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11750 range_int_cst_p rather than !symbolic_range_p before calling
11751 extract_range_from_multiplicative_op_1.
11753 2018-01-04 Jeff Law <law@redhat.com>
11755 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11756 redundant test in assertion.
11758 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11760 * doc/rtl.texi: Document machine_mode wrapper classes.
11762 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11764 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11765 using tree_to_uhwi.
11767 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11769 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11770 the VEC_PERM_EXPR fold to fail.
11772 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11775 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11776 to switched_sections.
11778 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11781 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11782 test for d.testing.
11784 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11787 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11788 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11790 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11793 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11794 is BLKmode and bitpos not zero or mode change is needed.
11796 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11799 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11802 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11805 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11806 instead of MULT rtx. Update all corresponding splitters.
11807 (*saddl_se): Ditto.
11808 (*ssub<modesuffix>): Ditto.
11809 (*ssubl_se): Ditto.
11810 (*cmp_sadd_di): Update split patterns.
11811 (*cmp_sadd_si): Ditto.
11812 (*cmp_sadd_sidi): Ditto.
11813 (*cmp_ssub_di): Ditto.
11814 (*cmp_ssub_si): Ditto.
11815 (*cmp_ssub_sidi): Ditto.
11816 * config/alpha/predicates.md (const23_operand): New predicate.
11817 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11818 Look for ASHIFT, not MULT inner operand.
11819 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11821 2018-01-04 Martin Liska <mliska@suse.cz>
11823 PR gcov-profile/83669
11824 * gcov.c (output_intermediate_file): Add version to intermediate
11826 * doc/gcov.texi: Document new field 'version' in intermediate
11827 file format. Fix location of '-k' option of gcov command.
11829 2018-01-04 Martin Liska <mliska@suse.cz>
11832 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11834 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11836 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11838 2018-01-03 Martin Sebor <msebor@redhat.com>
11840 PR tree-optimization/83655
11841 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11842 checking calls with invalid arguments.
11844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11846 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11847 (vectorizable_mask_load_store): Delete.
11848 (vectorizable_call): Return false for masked loads and stores.
11849 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11850 instead of gimple_assign_rhs1.
11851 (vectorizable_load): Handle IFN_MASK_LOAD.
11852 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11856 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11858 (vectorizable_mask_load_store): ...here.
11859 (vectorizable_load): ...and here.
11861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11863 * tree-vect-stmts.c (vect_build_all_ones_mask)
11864 (vect_build_zero_merge_argument): New functions, split out from...
11865 (vectorizable_load): ...here.
11867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11869 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11871 (vectorizable_mask_load_store): ...here.
11872 (vectorizable_store): ...and here.
11874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11876 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11878 (vectorizable_mask_load_store): ...here.
11880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11882 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11883 (vect_model_store_cost): Take a vec_load_store_type instead of a
11885 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11886 (vect_model_store_cost): Take a vec_load_store_type instead of a
11888 (vectorizable_mask_load_store): Update accordingly.
11889 (vectorizable_store): Likewise.
11890 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11894 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11895 IFN_MASK_LOAD calls here rather than...
11896 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11899 Alan Hayward <alan.hayward@arm.com>
11900 David Sherwood <david.sherwood@arm.com>
11902 * expmed.c (extract_bit_field_1): For vector extracts,
11903 fall back to extract_bit_field_as_subreg if vec_extract
11906 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11907 Alan Hayward <alan.hayward@arm.com>
11908 David Sherwood <david.sherwood@arm.com>
11910 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11911 they are variable or constant sized.
11912 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11913 slots for constant-sized data.
11915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11916 Alan Hayward <alan.hayward@arm.com>
11917 David Sherwood <david.sherwood@arm.com>
11919 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11920 handling COND_EXPRs with boolean comparisons, try to find a better
11921 basis for the mask type than the boolean itself.
11923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11925 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11926 is calculated and how it can be overridden.
11927 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11928 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11930 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11934 Alan Hayward <alan.hayward@arm.com>
11935 David Sherwood <david.sherwood@arm.com>
11937 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11938 Remove the mode argument.
11939 (aarch64_simd_valid_immediate): Remove the mode and inverse
11941 * config/aarch64/iterators.md (bitsize): New iterator.
11942 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11943 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11944 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11945 aarch64_simd_valid_immediate.
11946 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11947 (aarch64_reg_or_bic_imm): Likewise.
11948 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11949 with an insn_type enum and msl with a modifier_type enum.
11950 Replace element_width with a scalar_mode. Change the shift
11951 to unsigned int. Add constructors for scalar_float_mode and
11952 scalar_int_mode elements.
11953 (aarch64_vect_float_const_representable_p): Delete.
11954 (aarch64_can_const_movi_rtx_p)
11955 (aarch64_simd_scalar_immediate_valid_for_move)
11956 (aarch64_simd_make_constant): Update call to
11957 aarch64_simd_valid_immediate.
11958 (aarch64_advsimd_valid_immediate_hs): New function.
11959 (aarch64_advsimd_valid_immediate): Likewise.
11960 (aarch64_simd_valid_immediate): Remove mode and inverse
11961 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11962 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11963 and aarch64_float_const_representable_p on the result.
11964 (aarch64_output_simd_mov_immediate): Remove mode argument.
11965 Update call to aarch64_simd_valid_immediate and use of
11966 simd_immediate_info.
11967 (aarch64_output_scalar_simd_mov_immediate): Update call
11970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11971 Alan Hayward <alan.hayward@arm.com>
11972 David Sherwood <david.sherwood@arm.com>
11974 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11975 (mode_nunits): Likewise CONST_MODE_NUNITS.
11976 * machmode.def (ADJUST_NUNITS): Document.
11977 * genmodes.c (mode_data::need_nunits_adj): New field.
11978 (blank_mode): Update accordingly.
11979 (adj_nunits): New variable.
11980 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11982 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11983 listed in adj_nunits.
11984 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11985 listed in adj_nunits. Don't emit case statements for such modes.
11986 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11987 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11988 nothing if adj_nunits is nonnull.
11989 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11990 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11991 (emit_mode_fbit): Update use of print_maybe_const_decl.
11992 (emit_move_size): Likewise. Treat the array as non-const
11994 (emit_mode_adjustments): Handle adj_nunits.
11996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11998 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11999 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12000 (VECTOR_MODES): Use it.
12001 (make_vector_modes): Take the prefix as an argument.
12003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12004 Alan Hayward <alan.hayward@arm.com>
12005 David Sherwood <david.sherwood@arm.com>
12007 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12008 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12009 for MODE_VECTOR_BOOL.
12010 * machmode.def (VECTOR_BOOL_MODE): Document.
12011 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12012 (make_vector_bool_mode): New function.
12013 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12015 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12016 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12018 * stor-layout.c (int_mode_for_mode): Likewise.
12019 * tree.c (build_vector_type_for_mode): Likewise.
12020 * varasm.c (output_constant_pool_2): Likewise.
12021 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12022 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12023 for MODE_VECTOR_BOOL.
12024 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12025 of mode class checks.
12026 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12027 instead of a list of mode class checks.
12028 (expand_vector_scalar_condition): Likewise.
12029 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12031 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12032 Alan Hayward <alan.hayward@arm.com>
12033 David Sherwood <david.sherwood@arm.com>
12035 * machmode.h (mode_size): Change from unsigned short to
12037 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12038 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12039 or if measurement_type is not polynomial.
12040 (fixed_size_mode::includes_p): Check for constant-sized modes.
12041 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12042 return a poly_uint16 rather than an unsigned short.
12043 (emit_mode_size): Change the type of mode_size from unsigned short
12044 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12045 (emit_mode_adjustments): Cope with polynomial vector sizes.
12046 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12048 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12050 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12051 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12052 * caller-save.c (setup_save_areas): Likewise.
12053 (replace_reg_with_saved_mem): Likewise.
12054 * calls.c (emit_library_call_value_1): Likewise.
12055 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12056 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12057 (gen_lowpart_for_combine): Likewise.
12058 * convert.c (convert_to_integer_1): Likewise.
12059 * cse.c (equiv_constant, cse_insn): Likewise.
12060 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12061 (cselib_subst_to_values): Likewise.
12062 * dce.c (word_dce_process_block): Likewise.
12063 * df-problems.c (df_word_lr_mark_ref): Likewise.
12064 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12065 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12066 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12067 (rtl_for_decl_location): Likewise.
12068 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12069 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12070 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12071 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12072 (expand_expr_real_1): Likewise.
12073 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12074 (pad_below): Likewise.
12075 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12076 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12077 * ira.c (get_subreg_tracking_sizes): Likewise.
12078 * ira-build.c (ira_create_allocno_objects): Likewise.
12079 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12080 (ira_sort_regnos_for_alter_reg): Likewise.
12081 * ira-costs.c (record_operand_costs): Likewise.
12082 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12083 (resolve_simple_move): Likewise.
12084 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12085 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12086 (lra_constraints): Likewise.
12087 (CONST_POOL_OK_P): Reject variable-sized modes.
12088 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12089 (add_pseudo_to_slot, lra_spill): Likewise.
12090 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12091 * optabs-query.c (get_best_extraction_insn): Likewise.
12092 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12093 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12094 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12095 * recog.c (offsettable_address_addr_space_p): Likewise.
12096 * regcprop.c (maybe_mode_change): Likewise.
12097 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12098 * regrename.c (build_def_use): Likewise.
12099 * regstat.c (dump_reg_info): Likewise.
12100 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12101 (find_reloads, find_reloads_subreg_address): Likewise.
12102 * reload1.c (eliminate_regs_1): Likewise.
12103 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12104 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12105 (simplify_binary_operation_1, simplify_subreg): Likewise.
12106 * targhooks.c (default_function_arg_padding): Likewise.
12107 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12108 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12109 (verify_gimple_assign_ternary): Likewise.
12110 * tree-inline.c (estimate_move_cost): Likewise.
12111 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12112 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12113 (get_address_cost_ainc): Likewise.
12114 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12115 (vect_supportable_dr_alignment): Likewise.
12116 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12117 (vectorizable_reduction): Likewise.
12118 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12119 (vectorizable_operation, vectorizable_load): Likewise.
12120 * tree.c (build_same_sized_truth_vector_type): Likewise.
12121 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12122 * var-tracking.c (emit_note_insn_var_location): Likewise.
12123 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12124 (ADDR_VEC_ALIGN): Likewise.
12126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12127 Alan Hayward <alan.hayward@arm.com>
12128 David Sherwood <david.sherwood@arm.com>
12130 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12132 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12133 or if measurement_type is polynomial.
12134 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12135 * combine.c (make_extraction): Likewise.
12136 * dse.c (find_shift_sequence): Likewise.
12137 * dwarf2out.c (mem_loc_descriptor): Likewise.
12138 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12139 (extract_bit_field, extract_low_bits): Likewise.
12140 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12141 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12142 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12143 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12144 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12145 * reload.c (find_reloads): Likewise.
12146 * reload1.c (alter_reg): Likewise.
12147 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12148 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12149 * tree-if-conv.c (predicate_mem_writes): Likewise.
12150 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12151 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12152 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12153 * valtrack.c (dead_debug_insert_temp): Likewise.
12154 * varasm.c (mergeable_constant_section): Likewise.
12155 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12158 Alan Hayward <alan.hayward@arm.com>
12159 David Sherwood <david.sherwood@arm.com>
12161 * expr.c (expand_assignment): Cope with polynomial mode sizes
12162 when assigning to a CONCAT.
12164 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12165 Alan Hayward <alan.hayward@arm.com>
12166 David Sherwood <david.sherwood@arm.com>
12168 * machmode.h (mode_precision): Change from unsigned short to
12170 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12172 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12173 or if measurement_type is not polynomial.
12174 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12175 in which the mode is already known to be a scalar_int_mode.
12176 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12177 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12179 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12180 for GET_MODE_PRECISION.
12181 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12182 for GET_MODE_PRECISION.
12183 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12185 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12186 (expand_field_assignment, make_extraction): Likewise.
12187 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12188 (get_last_value): Likewise.
12189 * convert.c (convert_to_integer_1): Likewise.
12190 * cse.c (cse_insn): Likewise.
12191 * expr.c (expand_expr_real_1): Likewise.
12192 * lra-constraints.c (simplify_operand_subreg): Likewise.
12193 * optabs-query.c (can_atomic_load_p): Likewise.
12194 * optabs.c (expand_atomic_load): Likewise.
12195 (expand_atomic_store): Likewise.
12196 * ree.c (combine_reaching_defs): Likewise.
12197 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12198 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12199 * tree.h (type_has_mode_precision_p): Likewise.
12200 * ubsan.c (instrument_si_overflow): Likewise.
12202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12203 Alan Hayward <alan.hayward@arm.com>
12204 David Sherwood <david.sherwood@arm.com>
12206 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12207 polynomial numbers of units.
12208 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12209 (valid_vector_subparts_p): New function.
12210 (build_vector_type): Remove temporary shim and take the number
12211 of units as a poly_uint64 rather than an int.
12212 (build_opaque_vector_type): Take the number of units as a
12213 poly_uint64 rather than an int.
12214 * tree.c (build_vector_from_ctor): Handle polynomial
12215 TYPE_VECTOR_SUBPARTS.
12216 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12217 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12218 (build_vector_from_val): If the number of units is variable,
12219 use build_vec_duplicate_cst for constant operands and
12220 VEC_DUPLICATE_EXPR otherwise.
12221 (make_vector_type): Remove temporary is_constant ().
12222 (build_vector_type, build_opaque_vector_type): Take the number of
12223 units as a poly_uint64 rather than an int.
12224 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12226 * cfgexpand.c (expand_debug_expr): Likewise.
12227 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12228 (store_constructor, expand_expr_real_1): Likewise.
12229 (const_scalar_mask_from_tree): Likewise.
12230 * fold-const-call.c (fold_const_reduction): Likewise.
12231 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12232 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12233 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12234 (fold_relational_const): Likewise.
12235 (native_interpret_vector): Likewise. Change the size from an
12236 int to an unsigned int.
12237 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12238 TYPE_VECTOR_SUBPARTS.
12239 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12240 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12241 duplicating a non-constant operand into a variable-length vector.
12242 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12243 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12244 * ipa-icf.c (sem_variable::equals): Likewise.
12245 * match.pd: Likewise.
12246 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12247 * print-tree.c (print_node): Likewise.
12248 * stor-layout.c (layout_type): Likewise.
12249 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12250 * tree-cfg.c (verify_gimple_comparison): Likewise.
12251 (verify_gimple_assign_binary): Likewise.
12252 (verify_gimple_assign_ternary): Likewise.
12253 (verify_gimple_assign_single): Likewise.
12254 * tree-pretty-print.c (dump_generic_node): Likewise.
12255 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12256 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12257 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12258 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12259 (vect_shift_permute_load_chain): Likewise.
12260 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12261 (expand_vector_condition, optimize_vector_constructor): Likewise.
12262 (lower_vec_perm, get_compute_type): Likewise.
12263 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12264 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12265 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12266 (vect_recog_mask_conversion_pattern): Likewise.
12267 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12268 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12269 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12270 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12271 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12272 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12273 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12274 (supportable_widening_operation): Likewise.
12275 (supportable_narrowing_operation): Likewise.
12276 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12278 * varasm.c (output_constant): Likewise.
12280 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12281 Alan Hayward <alan.hayward@arm.com>
12282 David Sherwood <david.sherwood@arm.com>
12284 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12285 so that both the length == 3 and length != 3 cases set up their
12286 own permute vectors. Add comments explaining why we know the
12287 number of elements is constant.
12288 (vect_permute_load_chain): Likewise.
12290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12291 Alan Hayward <alan.hayward@arm.com>
12292 David Sherwood <david.sherwood@arm.com>
12294 * machmode.h (mode_nunits): Change from unsigned char to
12296 (ONLY_FIXED_SIZE_MODES): New macro.
12297 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12298 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12299 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12301 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12302 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12303 or if measurement_type is not polynomial.
12304 * genmodes.c (ZERO_COEFFS): New macro.
12305 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12307 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12308 Use ZERO_COEFFS when emitting initializers.
12309 * data-streamer.h (bp_pack_poly_value): New function.
12310 (bp_unpack_poly_value): Likewise.
12311 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12312 for GET_MODE_NUNITS.
12313 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12314 for GET_MODE_NUNITS.
12315 * tree.c (make_vector_type): Remove temporary shim and make
12316 the real function take the number of units as a poly_uint64
12317 rather than an int.
12318 (build_vector_type_for_mode): Handle polynomial nunits.
12319 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12320 * emit-rtl.c (const_vec_series_p_1): Likewise.
12321 (gen_rtx_CONST_VECTOR): Likewise.
12322 * fold-const.c (test_vec_duplicate_folding): Likewise.
12323 * genrecog.c (validate_pattern): Likewise.
12324 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12325 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12326 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12327 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12328 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12329 * rtlanal.c (subreg_get_info): Likewise.
12330 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12331 (vect_grouped_load_supported): Likewise.
12332 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12333 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12334 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12335 (simplify_const_unary_operation, simplify_binary_operation_1)
12336 (simplify_const_binary_operation, simplify_ternary_operation)
12337 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12338 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12339 instead of CONST_VECTOR_NUNITS.
12340 * varasm.c (output_constant_pool_2): Likewise.
12341 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12342 explicit-encoded elements in the XVEC for variable-length vectors.
12344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12346 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12349 Alan Hayward <alan.hayward@arm.com>
12350 David Sherwood <david.sherwood@arm.com>
12352 * coretypes.h (fixed_size_mode): Declare.
12353 (fixed_size_mode_pod): New typedef.
12354 * builtins.h (target_builtins::x_apply_args_mode)
12355 (target_builtins::x_apply_result_mode): Change type to
12356 fixed_size_mode_pod.
12357 * builtins.c (apply_args_size, apply_result_size, result_vector)
12358 (expand_builtin_apply_args_1, expand_builtin_apply)
12359 (expand_builtin_return): Update accordingly.
12361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12363 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12364 * cselib.c (cselib_hash_rtx): Likewise.
12365 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12366 CONST_VECTOR encoding.
12368 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12369 Jeff Law <law@redhat.com>
12372 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12373 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12374 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12375 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12378 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12379 explicitly probe *sp in a noreturn function if there were any callee
12380 register saves or frame pointer is needed.
12382 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12385 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12386 BLKmode for ternary, binary or unary expressions.
12389 * var-tracking.c (delete_vta_debug_insn): New inline function.
12390 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12391 insns from get_insns () to NULL instead of each bb separately.
12392 Use delete_vta_debug_insn. No longer static.
12393 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12394 delete_vta_debug_insns callers.
12395 * rtl.h (delete_vta_debug_insns): Declare.
12396 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12397 instead of variable_tracking_main.
12399 2018-01-03 Martin Sebor <msebor@redhat.com>
12401 PR tree-optimization/83603
12402 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12403 arguments past the endof the argument list in functions declared
12404 without a prototype.
12405 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12406 Avoid checking when arguments are null.
12408 2018-01-03 Martin Sebor <msebor@redhat.com>
12411 * doc/extend.texi (attribute const): Fix a typo.
12412 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12413 issuing -Wsuggest-attribute for void functions.
12415 2018-01-03 Martin Sebor <msebor@redhat.com>
12417 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12418 offset_int::from instead of wide_int::to_shwi.
12419 (maybe_diag_overlap): Remove assertion.
12420 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12421 * gimple-ssa-sprintf.c (format_directive): Same.
12422 (parse_directive): Same.
12423 (sprintf_dom_walker::compute_format_length): Same.
12424 (try_substitute_return_value): Same.
12426 2018-01-03 Jeff Law <law@redhat.com>
12428 PR middle-end/83654
12429 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12430 non-constant residual for zero at runtime and avoid probing in
12431 that case. Reorganize code for trailing problem to mirror handling
12434 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12436 PR tree-optimization/83501
12437 * tree-ssa-strlen.c (get_string_cst): New.
12438 (handle_char_store): Call get_string_cst.
12440 2018-01-03 Martin Liska <mliska@suse.cz>
12442 PR tree-optimization/83593
12443 * tree-ssa-strlen.c: Include tree-cfg.h.
12444 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12445 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12446 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12448 (strlen_dom_walker::before_dom_children): Call
12449 gimple_purge_dead_eh_edges. Dump tranformation with details
12451 (strlen_dom_walker::before_dom_children): Update call by adding
12452 new argument cleanup_eh.
12453 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12455 2018-01-03 Martin Liska <mliska@suse.cz>
12458 * cif-code.def (VARIADIC_THUNK): New enum value.
12459 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12462 2018-01-03 Jan Beulich <jbeulich@suse.com>
12464 * sse.md (mov<mode>_internal): Tighten condition for when to use
12465 vmovdqu<ssescalarsize> for TI and OI modes.
12467 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12469 Update copyright years.
12471 2018-01-03 Martin Liska <mliska@suse.cz>
12474 * ipa-visibility.c (function_and_variable_visibility): Skip
12475 functions with noipa attribure.
12477 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12479 * gcc.c (process_command): Update copyright notice dates.
12480 * gcov-dump.c (print_version): Ditto.
12481 * gcov.c (print_version): Ditto.
12482 * gcov-tool.c (print_version): Ditto.
12483 * gengtype.c (create_file): Ditto.
12484 * doc/cpp.texi: Bump @copying's copyright year.
12485 * doc/cppinternals.texi: Ditto.
12486 * doc/gcc.texi: Ditto.
12487 * doc/gccint.texi: Ditto.
12488 * doc/gcov.texi: Ditto.
12489 * doc/install.texi: Ditto.
12490 * doc/invoke.texi: Ditto.
12492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12494 * vector-builder.h (vector_builder::m_full_nelts): Change from
12495 unsigned int to poly_uint64.
12496 (vector_builder::full_nelts): Update prototype accordingly.
12497 (vector_builder::new_vector): Likewise.
12498 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12499 (vector_builder::operator ==): Likewise.
12500 (vector_builder::finalize): Likewise.
12501 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12502 Take the number of elements as a poly_uint64 rather than an
12504 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12505 from unsigned int to poly_uint64.
12506 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12507 (vec_perm_indices::new_vector): Likewise.
12508 (vec_perm_indices::length): Likewise.
12509 (vec_perm_indices::nelts_per_input): Likewise.
12510 (vec_perm_indices::input_nelts): Likewise.
12511 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12512 number of elements per input as a poly_uint64 rather than an
12513 unsigned int. Use the original encoding for variable-length
12514 vectors, rather than clamping each individual element.
12515 For the second and subsequent elements in each pattern,
12516 clamp the step and base before clamping their sum.
12517 (vec_perm_indices::series_p): Handle polynomial element counts.
12518 (vec_perm_indices::all_in_range_p): Likewise.
12519 (vec_perm_indices_to_tree): Likewise.
12520 (vec_perm_indices_to_rtx): Likewise.
12521 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12522 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12523 (tree_vector_builder::new_binary_operation): Handle polynomial
12524 element counts. Return false if we need to know the number
12525 of elements at compile time.
12526 * fold-const.c (fold_vec_perm): Punt if the number of elements
12527 isn't known at compile time.
12529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12531 * vec-perm-indices.h (vec_perm_builder): Change element type
12532 from HOST_WIDE_INT to poly_int64.
12533 (vec_perm_indices::element_type): Update accordingly.
12534 (vec_perm_indices::clamp): Handle polynomial element_types.
12535 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12536 (vec_perm_indices::all_in_range_p): Likewise.
12537 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12539 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12540 polynomial vec_perm_indices element types.
12541 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12542 * fold-const.c (fold_vec_perm): Likewise.
12543 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12544 * tree-vect-generic.c (lower_vec_perm): Likewise.
12545 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12546 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12547 element type to HOST_WIDE_INT.
12549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12550 Alan Hayward <alan.hayward@arm.com>
12551 David Sherwood <david.sherwood@arm.com>
12553 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12554 rather than an int. Use plus_constant.
12555 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12556 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12558 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12559 Alan Hayward <alan.hayward@arm.com>
12560 David Sherwood <david.sherwood@arm.com>
12562 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12563 a HOST_WIDE_INT to a poly_int64.
12565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12566 Alan Hayward <alan.hayward@arm.com>
12567 David Sherwood <david.sherwood@arm.com>
12569 * calls.c (load_register_parameters): Cope with polynomial
12570 mode sizes. Require a constant size for BLKmode parameters
12571 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12572 forces a parameter to be padded at the lsb end in order to
12573 fill a complete number of words, require the parameter size
12574 to be ordered wrt UNITS_PER_WORD.
12576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12577 Alan Hayward <alan.hayward@arm.com>
12578 David Sherwood <david.sherwood@arm.com>
12580 * reload1.c (spill_stack_slot_width): Change element type
12581 from unsigned int to poly_uint64_pod.
12582 (alter_reg): Treat mode sizes as polynomial.
12584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12585 Alan Hayward <alan.hayward@arm.com>
12586 David Sherwood <david.sherwood@arm.com>
12588 * reload.c (complex_word_subreg_p): New function.
12589 (reload_inner_reg_of_subreg, push_reload): Use it.
12591 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12592 Alan Hayward <alan.hayward@arm.com>
12593 David Sherwood <david.sherwood@arm.com>
12595 * lra-constraints.c (process_alt_operands): Reject matched
12596 operands whose sizes aren't ordered.
12597 (match_reload): Refer to this check here.
12599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12600 Alan Hayward <alan.hayward@arm.com>
12601 David Sherwood <david.sherwood@arm.com>
12603 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12604 that the mode size is in the set {1, 2, 4, 8, 16}.
12606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12607 Alan Hayward <alan.hayward@arm.com>
12608 David Sherwood <david.sherwood@arm.com>
12610 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12611 Use plus_constant instead of gen_rtx_PLUS.
12613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12614 Alan Hayward <alan.hayward@arm.com>
12615 David Sherwood <david.sherwood@arm.com>
12617 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12618 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12619 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12620 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12621 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12622 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12623 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12624 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12625 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12626 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12628 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12629 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12630 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12631 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12632 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12633 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12634 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12635 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12636 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12637 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12639 * expr.c (emit_move_resolve_push): Treat the input and result
12640 of PUSH_ROUNDING as a poly_int64.
12641 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12642 (emit_push_insn): Likewise.
12643 * lra-eliminations.c (mark_not_eliminable): Likewise.
12644 * recog.c (push_operand): Likewise.
12645 * reload1.c (elimination_effects): Likewise.
12646 * rtlanal.c (nonzero_bits1): Likewise.
12647 * calls.c (store_one_arg): Likewise. Require the padding to be
12648 known at compile time.
12650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12651 Alan Hayward <alan.hayward@arm.com>
12652 David Sherwood <david.sherwood@arm.com>
12654 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12655 Use plus_constant instead of gen_rtx_PLUS.
12657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12658 Alan Hayward <alan.hayward@arm.com>
12659 David Sherwood <david.sherwood@arm.com>
12661 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12662 rather than an int.
12664 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12665 Alan Hayward <alan.hayward@arm.com>
12666 David Sherwood <david.sherwood@arm.com>
12668 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12669 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12670 via stack temporaries. Treat the mode size as polynomial too.
12672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12673 Alan Hayward <alan.hayward@arm.com>
12674 David Sherwood <david.sherwood@arm.com>
12676 * expr.c (expand_expr_real_2): When handling conversions involving
12677 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12678 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12679 as a poly_uint64 too.
12681 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12682 Alan Hayward <alan.hayward@arm.com>
12683 David Sherwood <david.sherwood@arm.com>
12685 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12687 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12688 Alan Hayward <alan.hayward@arm.com>
12689 David Sherwood <david.sherwood@arm.com>
12691 * combine.c (can_change_dest_mode): Handle polynomial
12692 REGMODE_NATURAL_SIZE.
12693 * expmed.c (store_bit_field_1): Likewise.
12694 * expr.c (store_constructor): Likewise.
12695 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12696 and polynomial REGMODE_NATURAL_SIZE.
12697 (gen_lowpart_common): Likewise.
12698 * reginfo.c (record_subregs_of_mode): Likewise.
12699 * rtlanal.c (read_modify_subreg_p): Likewise.
12701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12702 Alan Hayward <alan.hayward@arm.com>
12703 David Sherwood <david.sherwood@arm.com>
12705 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12706 numbers of elements.
12708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12709 Alan Hayward <alan.hayward@arm.com>
12710 David Sherwood <david.sherwood@arm.com>
12712 * match.pd: Cope with polynomial numbers of vector elements.
12714 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12715 Alan Hayward <alan.hayward@arm.com>
12716 David Sherwood <david.sherwood@arm.com>
12718 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12719 in a POINTER_PLUS_EXPR.
12721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12722 Alan Hayward <alan.hayward@arm.com>
12723 David Sherwood <david.sherwood@arm.com>
12725 * omp-simd-clone.c (simd_clone_subparts): New function.
12726 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12727 (ipa_simd_modify_function_body): Likewise.
12729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12730 Alan Hayward <alan.hayward@arm.com>
12731 David Sherwood <david.sherwood@arm.com>
12733 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12734 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12735 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12736 (expand_vector_condition, vector_element): Likewise.
12737 (subparts_gt): New function.
12738 (get_compute_type): Use subparts_gt.
12739 (count_type_subparts): Delete.
12740 (expand_vector_operations_1): Use subparts_gt instead of
12741 count_type_subparts.
12743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12744 Alan Hayward <alan.hayward@arm.com>
12745 David Sherwood <david.sherwood@arm.com>
12747 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12748 (vect_compile_time_alias): ...this new function. Do the calculation
12749 on poly_ints rather than trees.
12750 (vect_prune_runtime_alias_test_list): Update call accordingly.
12752 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12753 Alan Hayward <alan.hayward@arm.com>
12754 David Sherwood <david.sherwood@arm.com>
12756 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12758 (vect_schedule_slp_instance): Likewise.
12760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12761 Alan Hayward <alan.hayward@arm.com>
12762 David Sherwood <david.sherwood@arm.com>
12764 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12765 constant and extern definitions for variable-length vectors.
12766 (vect_get_constant_vectors): Note that the number of units
12767 is known to be constant.
12769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12770 Alan Hayward <alan.hayward@arm.com>
12771 David Sherwood <david.sherwood@arm.com>
12773 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12774 of units as polynomial. Choose between WIDE and NARROW based
12777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12778 Alan Hayward <alan.hayward@arm.com>
12779 David Sherwood <david.sherwood@arm.com>
12781 * tree-vect-stmts.c (simd_clone_subparts): New function.
12782 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12785 Alan Hayward <alan.hayward@arm.com>
12786 David Sherwood <david.sherwood@arm.com>
12788 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12789 vectors as polynomial. Use build_index_vector for
12790 IFN_GOMP_SIMD_LANE.
12792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12793 Alan Hayward <alan.hayward@arm.com>
12794 David Sherwood <david.sherwood@arm.com>
12796 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12797 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12798 for variable-length vectors.
12799 (vectorizable_mask_load_store): Treat the number of units as
12800 polynomial, asserting that it is constant if the condition has
12801 already been enforced.
12802 (vectorizable_store, vectorizable_load): Likewise.
12804 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12805 Alan Hayward <alan.hayward@arm.com>
12806 David Sherwood <david.sherwood@arm.com>
12808 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12809 of units as polynomial. Punt if we can't tell at compile time
12810 which vector contains the final result.
12812 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12813 Alan Hayward <alan.hayward@arm.com>
12814 David Sherwood <david.sherwood@arm.com>
12816 * tree-vect-loop.c (vectorizable_induction): Treat the number
12817 of units as polynomial. Punt on SLP inductions. Use an integer
12818 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12819 cast of such a series for variable-length floating-point
12822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12823 Alan Hayward <alan.hayward@arm.com>
12824 David Sherwood <david.sherwood@arm.com>
12826 * tree.h (build_index_vector): Declare.
12827 * tree.c (build_index_vector): New function.
12828 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12829 of units as polynomial, forcibly converting it to a constant if
12830 vectorizable_reduction has already enforced the condition.
12831 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12832 to create a {1,2,3,...} vector.
12833 (vectorizable_reduction): Treat the number of units as polynomial.
12834 Choose vectype_in based on the largest scalar element size rather
12835 than the smallest number of units. Enforce the restrictions
12838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12839 Alan Hayward <alan.hayward@arm.com>
12840 David Sherwood <david.sherwood@arm.com>
12842 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12843 number of units as polynomial.
12845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12846 Alan Hayward <alan.hayward@arm.com>
12847 David Sherwood <david.sherwood@arm.com>
12849 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12850 * target.def (autovectorize_vector_sizes): Return the vector sizes
12851 by pointer, using vector_sizes rather than a bitmask.
12852 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12853 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12854 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12856 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12857 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12858 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12859 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12860 * omp-general.c (omp_max_vf): Likewise.
12861 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12862 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12863 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12864 * tree-vect-slp.c (vect_slp_bb): Likewise.
12865 * doc/tm.texi: Regenerate.
12866 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12868 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12869 the vector size as a poly_uint64 rather than an unsigned int.
12870 (current_vector_size): Change from an unsigned int to a poly_uint64.
12871 (get_vectype_for_scalar_type): Update accordingly.
12872 * tree.h (build_truth_vector_type): Take the size and number of
12873 units as a poly_uint64 rather than an unsigned int.
12874 (build_vector_type): Add a temporary overload that takes
12875 the number of units as a poly_uint64 rather than an unsigned int.
12876 * tree.c (make_vector_type): Likewise.
12877 (build_truth_vector_type): Take the number of units as a poly_uint64
12878 rather than an unsigned int.
12880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12881 Alan Hayward <alan.hayward@arm.com>
12882 David Sherwood <david.sherwood@arm.com>
12884 * target.def (get_mask_mode): Take the number of units and length
12885 as poly_uint64s rather than unsigned ints.
12886 * targhooks.h (default_get_mask_mode): Update accordingly.
12887 * targhooks.c (default_get_mask_mode): Likewise.
12888 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12889 * doc/tm.texi: Regenerate.
12891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12892 Alan Hayward <alan.hayward@arm.com>
12893 David Sherwood <david.sherwood@arm.com>
12895 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12896 * omp-general.c (omp_max_vf): Likewise.
12897 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12898 (expand_omp_simd): Handle polynomial safelen.
12899 * omp-low.c (omplow_simd_context): Add a default constructor.
12900 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12901 (lower_rec_simd_input_clauses): Update accordingly.
12902 (lower_rec_input_clauses): Likewise.
12904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12905 Alan Hayward <alan.hayward@arm.com>
12906 David Sherwood <david.sherwood@arm.com>
12908 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12909 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12910 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12911 (vect_analyze_slp_cost): Likewise.
12912 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12913 (vect_model_load_cost): Likewise.
12915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12916 Alan Hayward <alan.hayward@arm.com>
12917 David Sherwood <david.sherwood@arm.com>
12919 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12920 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12921 from an unsigned int * to a poly_uint64_pod *.
12922 (calculate_unrolling_factor): New function.
12923 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12926 Alan Hayward <alan.hayward@arm.com>
12927 David Sherwood <david.sherwood@arm.com>
12929 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12930 from an unsigned int to a poly_uint64.
12931 (_loop_vec_info::slp_unrolling_factor): Likewise.
12932 (_loop_vec_info::vectorization_factor): Change from an int
12934 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12935 (vect_get_num_vectors): New function.
12936 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12937 (vect_get_num_copies): Use vect_get_num_vectors.
12938 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12939 to an unsigned int *.
12940 (vect_analyze_data_refs): Change min_vf from an int * to a
12942 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12943 than an unsigned HOST_WIDE_INT.
12944 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12945 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12946 to an unsigned int *.
12947 (vect_analyze_data_ref_dependences): Likewise.
12948 (vect_compute_data_ref_alignment): Handle polynomial vf.
12949 (vect_enhance_data_refs_alignment): Likewise.
12950 (vect_prune_runtime_alias_test_list): Likewise.
12951 (vect_shift_permute_load_chain): Likewise.
12952 (vect_supportable_dr_alignment): Likewise.
12953 (dependence_distance_ge_vf): Take the vectorization factor as a
12954 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12955 (vect_analyze_data_refs): Change min_vf from an int * to a
12957 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12958 vfm1 as a poly_uint64 rather than an int. Make the same change
12959 for the returned bound_scalar.
12960 (vect_gen_vector_loop_niters): Handle polynomial vf.
12961 (vect_do_peeling): Likewise. Update call to
12962 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12963 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12965 * tree-vect-loop.c (vect_determine_vectorization_factor)
12966 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12967 (vect_get_known_peeling_cost): Likewise.
12968 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12969 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12970 (vect_transform_loop): Likewise. Use the lowest possible VF when
12971 updating the upper bounds of the loop.
12972 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12973 rather than an int.
12974 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12975 polynomial unroll factors.
12976 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12977 (vect_make_slp_decision): Likewise.
12978 (vect_supported_load_permutation_p): Likewise, and polynomial
12980 (vect_analyze_slp_cost): Handle polynomial vf.
12981 (vect_slp_analyze_node_operations): Likewise.
12982 (vect_slp_analyze_bb_1): Likewise.
12983 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12984 than an unsigned HOST_WIDE_INT.
12985 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12986 (vectorizable_load): Handle polynomial vf.
12987 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12989 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12992 Alan Hayward <alan.hayward@arm.com>
12993 David Sherwood <david.sherwood@arm.com>
12995 * match.pd: Handle bit operations involving three constants
12996 and try to fold one pair.
12998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13000 * tree-vect-loop-manip.c: Include gimple-fold.h.
13001 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13002 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13003 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13004 Add a path that uses a step of VF instead of 1, but disable it
13006 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13007 and niters_no_overflow parameters. Update calls to
13008 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13009 Create a new SSA name if the latter choses to use a ste other
13010 than zero, and return it via niters_vector_mult_vf_var.
13011 * tree-vect-loop.c (vect_transform_loop): Update calls to
13012 vect_do_peeling, vect_gen_vector_loop_niters and
13013 slpeel_make_loop_iterate_ntimes.
13014 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13015 (vect_gen_vector_loop_niters): Update declarations after above changes.
13017 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13019 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13020 128-bit round to integer instructions.
13021 (ceil<mode>2): Likewise.
13022 (btrunc<mode>2): Likewise.
13023 (round<mode>2): Likewise.
13025 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13027 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13028 unaligned VSX load/store on P8/P9.
13029 (expand_block_clear): Allow the use of unaligned VSX
13030 load/store on P8/P9.
13032 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13034 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13036 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13037 swap associated with both a load and a store.
13039 2018-01-02 Andrew Waterman <andrew@sifive.com>
13041 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13042 * config/riscv/riscv.md (clear_cache): Use it.
13044 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13046 * web.c: Remove out-of-date comment.
13048 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13050 * expr.c (fixup_args_size_notes): Check that any existing
13051 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13052 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13053 (emit_single_push_insn): ...here.
13055 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13057 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13058 (const_vector_encoded_nelts): New function.
13059 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13060 (const_vector_int_elt, const_vector_elt): Declare.
13061 * emit-rtl.c (const_vector_int_elt_1): New function.
13062 (const_vector_elt): Likewise.
13063 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13064 of CONST_VECTOR_ELT.
13066 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13068 * expr.c: Include rtx-vector-builder.h.
13069 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13070 directly on the tree encoding.
13071 (const_vector_from_tree): Likewise.
13072 * optabs.c: Include rtx-vector-builder.h.
13073 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13074 sequence of "u" values.
13075 * vec-perm-indices.c: Include rtx-vector-builder.h.
13076 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13077 directly on the vec_perm_indices encoding.
13079 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13081 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13082 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13083 * rtx-vector-builder.h: New file.
13084 * rtx-vector-builder.c: Likewise.
13085 * rtl.h (rtx_def::u2): Add a const_vector field.
13086 (CONST_VECTOR_NPATTERNS): New macro.
13087 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13088 (CONST_VECTOR_DUPLICATE_P): Likewise.
13089 (CONST_VECTOR_STEPPED_P): Likewise.
13090 (CONST_VECTOR_ENCODED_ELT): Likewise.
13091 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13092 (unwrap_const_vec_duplicate): Likewise.
13093 (const_vec_series_p): Check for a non-duplicated vector encoding.
13094 Say that the function only returns true for integer vectors.
13095 * emit-rtl.c: Include rtx-vector-builder.h.
13096 (gen_const_vec_duplicate_1): Delete.
13097 (gen_const_vector): Call gen_const_vec_duplicate instead of
13098 gen_const_vec_duplicate_1.
13099 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13100 (gen_const_vec_duplicate): Use rtx_vector_builder.
13101 (gen_const_vec_series): Likewise.
13102 (gen_rtx_CONST_VECTOR): Likewise.
13103 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13104 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13105 Build a new vector rather than modifying a CONST_VECTOR in-place.
13106 (handle_special_swappables): Update call accordingly.
13107 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13108 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13109 Build a new vector rather than modifying a CONST_VECTOR in-place.
13110 (handle_special_swappables): Update call accordingly.
13112 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13114 * simplify-rtx.c (simplify_const_binary_operation): Use
13115 CONST_VECTOR_ELT instead of XVECEXP.
13117 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13119 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13120 the selector elements to be different from the data elements
13121 if the selector is a VECTOR_CST.
13122 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13123 ssizetype for the selector.
13125 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13127 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13128 before testing each element individually.
13129 * tree-vect-generic.c (lower_vec_perm): Likewise.
13131 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13133 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13134 * selftest-run-tests.c (selftest::run_tests): Call it.
13135 * vector-builder.h (vector_builder::operator ==): New function.
13136 (vector_builder::operator !=): Likewise.
13137 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13138 (vec_perm_indices::all_from_input_p): New function.
13139 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13140 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13141 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13142 instead of reading the VECTOR_CST directly. Detect whether both
13143 vector inputs are the same before constructing the vec_perm_indices,
13144 and update the number of inputs argument accordingly. Use the
13145 utility functions added above. Only construct sel2 if we need to.
13147 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13149 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13150 the broadcast of the low byte.
13151 (expand_mult_highpart): Use an explicit encoding for the permutes.
13152 * optabs-query.c (can_mult_highpart_p): Likewise.
13153 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13154 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13155 (vectorizable_bswap): Likewise.
13156 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13157 explicit encoding for the power-of-2 permutes.
13158 (vect_permute_store_chain): Likewise.
13159 (vect_grouped_load_supported): Likewise.
13160 (vect_permute_load_chain): Likewise.
13162 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13164 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13165 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13166 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13167 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13168 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13169 (vect_gen_perm_mask_any): Likewise.
13171 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13173 * int-vector-builder.h: New file.
13174 * vec-perm-indices.h: Include int-vector-builder.h.
13175 (vec_perm_indices): Redefine as an int_vector_builder.
13176 (auto_vec_perm_indices): Delete.
13177 (vec_perm_builder): Redefine as a stand-alone class.
13178 (vec_perm_indices::vec_perm_indices): New function.
13179 (vec_perm_indices::clamp): Likewise.
13180 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13181 (vec_perm_indices::new_vector): New function.
13182 (vec_perm_indices::new_expanded_vector): Update for new
13183 vec_perm_indices class.
13184 (vec_perm_indices::rotate_inputs): New function.
13185 (vec_perm_indices::all_in_range_p): Operate directly on the
13186 encoded form, without computing elided elements.
13187 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13188 encoding. Update for new vec_perm_indices class.
13189 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13190 the given vec_perm_builder.
13191 (expand_vec_perm_var): Update vec_perm_builder constructor.
13192 (expand_mult_highpart): Use vec_perm_builder instead of
13193 auto_vec_perm_indices.
13194 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13195 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13196 or double series encoding as appropriate.
13197 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13198 vec_perm_indices instead of auto_vec_perm_indices.
13199 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13200 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13201 (vect_permute_store_chain): Likewise.
13202 (vect_grouped_load_supported): Likewise.
13203 (vect_permute_load_chain): Likewise.
13204 (vect_shift_permute_load_chain): Likewise.
13205 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13206 (vect_transform_slp_perm_load): Likewise.
13207 (vect_schedule_slp_instance): Likewise.
13208 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13209 (vectorizable_mask_load_store): Likewise.
13210 (vectorizable_bswap): Likewise.
13211 (vectorizable_store): Likewise.
13212 (vectorizable_load): Likewise.
13213 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13214 vec_perm_indices instead of auto_vec_perm_indices. Use
13215 tree_to_vec_perm_builder to read the vector from a tree.
13216 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13217 vec_perm_builder instead of a vec_perm_indices.
13218 (have_whole_vector_shift): Use vec_perm_builder and
13219 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13220 truncation to calc_vec_perm_mask_for_shift.
13221 (vect_create_epilog_for_reduction): Likewise.
13222 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13223 from auto_vec_perm_indices to vec_perm_indices.
13224 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13225 instead of changing individual elements.
13226 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13227 the vector in d.perm.
13228 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13229 from auto_vec_perm_indices to vec_perm_indices.
13230 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13231 instead of changing individual elements.
13232 (arm_vectorize_vec_perm_const): Use new_vector to install
13233 the vector in d.perm.
13234 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13235 Update vec_perm_builder constructor.
13236 (rs6000_expand_interleave): Likewise.
13237 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13238 (rs6000_expand_interleave): Likewise.
13240 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13242 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13243 to qimode could truncate the indices.
13244 * optabs.c (expand_vec_perm_var): Likewise.
13246 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13248 * Makefile.in (OBJS): Add vec-perm-indices.o.
13249 * vec-perm-indices.h: New file.
13250 * vec-perm-indices.c: Likewise.
13251 * target.h (vec_perm_indices): Replace with a forward class
13253 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13254 * optabs.h: Include vec-perm-indices.h.
13255 (expand_vec_perm): Delete.
13256 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13257 (expand_vec_perm_const): Declare.
13258 * target.def (vec_perm_const_ok): Replace with...
13259 (vec_perm_const): ...this new hook.
13260 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13261 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13262 * doc/tm.texi: Regenerate.
13263 * optabs.def (vec_perm_const): Delete.
13264 * doc/md.texi (vec_perm_const): Likewise.
13265 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13266 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13267 expand_vec_perm for constant permutation vectors. Assert that
13268 the mode of variable permutation vectors is the integer equivalent
13269 of the mode that is being permuted.
13270 * optabs-query.h (selector_fits_mode_p): Declare.
13271 * optabs-query.c: Include vec-perm-indices.h.
13272 (selector_fits_mode_p): New function.
13273 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13274 is defined, instead of checking whether the vec_perm_const_optab
13275 exists. Use targetm.vectorize.vec_perm_const instead of
13276 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13277 fit in the vector mode before using a variable permute.
13278 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13279 vec_perm_indices instead of an rtx.
13280 (expand_vec_perm): Replace with...
13281 (expand_vec_perm_const): ...this new function. Take the selector
13282 as a vec_perm_indices rather than an rtx. Also take the mode of
13283 the selector. Update call to shift_amt_for_vec_perm_mask.
13284 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13285 Use vec_perm_indices::new_expanded_vector to expand the original
13286 selector into bytes. Check whether the indices fit in the vector
13287 mode before using a variable permute.
13288 (expand_vec_perm_var): Make global.
13289 (expand_mult_highpart): Use expand_vec_perm_const.
13290 * fold-const.c: Includes vec-perm-indices.h.
13291 * tree-ssa-forwprop.c: Likewise.
13292 * tree-vect-data-refs.c: Likewise.
13293 * tree-vect-generic.c: Likewise.
13294 * tree-vect-loop.c: Likewise.
13295 * tree-vect-slp.c: Likewise.
13296 * tree-vect-stmts.c: Likewise.
13297 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13299 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13300 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13301 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13302 (aarch64_vectorize_vec_perm_const): ...this new function.
13303 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13304 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13305 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13306 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13307 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13308 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13309 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13311 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13312 check for NEON modes.
13313 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13314 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13315 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13316 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13318 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13319 the old VEC_PERM_CONST conditions.
13320 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13321 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13322 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13323 (ia64_vectorize_vec_perm_const_ok): Merge into...
13324 (ia64_vectorize_vec_perm_const): ...this new function.
13325 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13326 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13327 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13328 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13329 * config/mips/mips.c (mips_expand_vec_perm_const)
13330 (mips_vectorize_vec_perm_const_ok): Merge into...
13331 (mips_vectorize_vec_perm_const): ...this new function.
13332 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13333 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13334 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13335 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13336 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13337 (rs6000_expand_vec_perm_const): Delete.
13338 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13340 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13341 (altivec_expand_vec_perm_const_le): Take each operand individually.
13342 Operate on constant selectors rather than rtxes.
13343 (altivec_expand_vec_perm_const): Likewise. Update call to
13344 altivec_expand_vec_perm_const_le.
13345 (rs6000_expand_vec_perm_const): Delete.
13346 (rs6000_vectorize_vec_perm_const_ok): Delete.
13347 (rs6000_vectorize_vec_perm_const): New function.
13348 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13349 an element count and rtx array.
13350 (rs6000_expand_extract_even): Update call accordingly.
13351 (rs6000_expand_interleave): Likewise.
13352 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13353 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13354 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13355 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13356 (rs6000_expand_vec_perm_const): Delete.
13357 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13358 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13359 (altivec_expand_vec_perm_const_le): Take each operand individually.
13360 Operate on constant selectors rather than rtxes.
13361 (altivec_expand_vec_perm_const): Likewise. Update call to
13362 altivec_expand_vec_perm_const_le.
13363 (rs6000_expand_vec_perm_const): Delete.
13364 (rs6000_vectorize_vec_perm_const_ok): Delete.
13365 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13366 reference to the SPE evmerge intructions.
13367 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13368 an element count and rtx array.
13369 (rs6000_expand_extract_even): Update call accordingly.
13370 (rs6000_expand_interleave): Likewise.
13371 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13372 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13374 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13376 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13378 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13379 vector mode and that that mode matches the mode of the data
13381 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13382 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13383 directly using expand_vec_perm_1 when forcing selectors into
13385 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13387 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13389 * optabs-query.h (can_vec_perm_p): Delete.
13390 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13391 * optabs-query.c (can_vec_perm_p): Split into...
13392 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13393 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13394 particular selector is valid.
13395 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13396 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13397 (vect_grouped_load_supported): Likewise.
13398 (vect_shift_permute_load_chain): Likewise.
13399 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13400 (vect_transform_slp_perm_load): Likewise.
13401 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13402 (vectorizable_bswap): Likewise.
13403 (vect_gen_perm_mask_checked): Likewise.
13404 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13405 implementations of variable permutation vectors into account
13406 when deciding which selector to use.
13407 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13408 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13409 with a false third argument.
13410 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13411 to test whether the constant selector is valid and can_vec_perm_var_p
13412 to test whether a variable selector is valid.
13414 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13416 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13417 * optabs-query.c (can_vec_perm_p): Likewise.
13418 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13419 instead of vec_perm_indices.
13420 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13421 (vect_gen_perm_mask_checked): Likewise,
13422 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13423 (vect_gen_perm_mask_checked): Likewise,
13425 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13427 * optabs-query.h (qimode_for_vec_perm): Declare.
13428 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13429 (qimode_for_vec_perm): ...this new function.
13430 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13432 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13434 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13435 does not have a conditional at the top.
13437 2018-01-02 Richard Biener <rguenther@suse.de>
13439 * ipa-inline.c (big_speedup_p): Fix expression.
13441 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13444 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13447 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13451 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13452 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13453 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13454 cond_taken_branch_cost 3->4.
13456 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13458 PR tree-optimization/83581
13459 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13460 TODO_cleanup_cfg if any changes have been made.
13462 PR middle-end/83608
13463 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13464 convert_modes if target mode has the right side, but different mode
13467 PR middle-end/83609
13468 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13469 last argument when extracting from CONCAT. If either from_real or
13470 from_imag is NULL, use expansion through memory. If result is not
13471 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13472 the parts directly to inner mode, if even that fails, use expansion
13475 PR middle-end/83623
13476 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13477 check for bswap in mode rather than HImode and use that in expand_unop
13480 Copyright (C) 2018 Free Software Foundation, Inc.
13482 Copying and distribution of this file, with or without modification,
13483 are permitted in any medium without royalty provided the copyright
13484 notice and this notice are preserved.