1 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
5 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
7 (armv2, armv2a, armv3, armv3m): Delete architectures.
8 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
9 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
10 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
12 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
13 (*mulsidi3adddi): Likewise.
15 (*mulsidi3_nov6): Likewise.
16 (umulsidi3): Likewise.
17 (umulsidi3_nov6): Likewise.
18 (umaddsidi4): Likewise.
19 (*umulsidi3adddi): Likewise.
20 (smulsi3_highpart): Likewise.
21 (*smulsi3_highpart_nov6): Likewise.
22 (umulsi3_highpart): Likewise.
23 (*umulsi3_highpart_nov6): Likewise.
24 * config/arm/arm.h (arm_arch3m): Delete.
25 * config/arm/arm.c (arm_arch3m): Delete.
26 (arm_option_override_internal): Update armv3-related comment.
27 (arm_configure_build_target): Delete use of isa_bit_mode32.
28 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
29 (arm_rtx_costs_internal): Delete check of arm_arch3m.
30 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
33 * config/arm/arm-protos.h (arm_arch3m): Delete.
34 * config/arm/arm-tables.opt: Regenerate.
35 * config/arm/arm-tune.md: Likewise.
36 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
37 deleted architectures.
39 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
41 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
42 (armv5t, armv5te): New features.
43 (ARMv5, ARMv5e): Delete fgroups.
44 (ARMv5t, ARMv5te): Adjust for above changes.
46 (armv5, armv5e): Delete arches.
47 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
49 (*call_reg_arm): Likewise.
50 (*call_value_reg_armv5): Likewise.
51 (*call_value_reg_arm): Likewise.
52 (*call_symbol): Likewise.
53 (*call_value_symbol): Likewise.
54 (*sibcall_insn): Likewise.
55 (*sibcall_value_insn): Likewise.
58 (define_split and define_peephole2 dependent on arm_arch5):
60 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
62 (TARGET_ARM_QBIT): Likewise.
63 (TARGET_DSP_MULTIPLY): Likewise.
64 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
65 (arm_arch5, arm_arch5e): Delete.
66 (arm_arch5t, arm_arch5te): Declare.
67 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
68 (arm_arch5t): Declare.
69 (arm_option_reconfigure_globals): Update for the above.
70 (arm_options_perform_arch_sanity_checks): Update comment, replace
71 use of arm_arch5 with arm_arch5t.
72 (use_return_insn): Likewise.
73 (arm_emit_call_insn): Likewise.
74 (output_return_instruction): Likewise.
75 (arm_final_prescan_insn): Likewise.
76 (arm_coproc_builtin_available): Likewise.
77 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
78 arm_arch5e with arm_arch5t and arm_arch5te.
79 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
80 (arm_arch5t, arm_arch5te): Declare.
81 * config/arm/arm-tables.opt: Regenerate.
82 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
83 * config/arm/t-multilib: Likewise.
84 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
86 (*call_reg_thumb1): Likewise.
87 (*call_value_reg_thumb1_v5): Likewise.
88 (*call_value_reg_thumb1): Likewise.
89 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
91 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
93 2018-05-18 Martin Liska <mliska@suse.cz>
96 * doc/gcov.texi: Document -t option of gcov tool.
98 2018-05-18 Martin Liska <mliska@suse.cz>
100 PR gcov-profile/84846
101 * gcov.c (print_usage): Add new -t option.
102 (process_args): Handle the option.
103 (generate_results): Use stdout as output when requested by
106 2018-05-18 Martin Liska <mliska@suse.cz>
108 PR gcov-profile/84846
109 * coverage.c (coverage_init): Write PWD to .gcno file.
110 * doc/gcov.texi: Document how working directory is printed.
111 * gcov-dump.c (dump_gcov_file): Print PWD.
112 * gcov.c (output_intermediate_file): Likewise.
113 (read_graph_file): Read PWD string.
114 (output_lines): Print PWD.
116 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
119 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
120 for retval and return false if all args to phi are zero.
122 2018-05-18 Richard Biener <rguenther@suse.de>
124 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
126 (evrp_dom_walker::before_dom_children): Call it.
128 2018-05-18 Richard Biener <rguenther@suse.de>
130 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
131 results when processing array refs with variable index.
133 2018-05-18 Toon Moene <toon@moene.org>
135 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
136 directly after that of -floop-interchange. Indicate that both
137 options are enabled by default when specifying -O3.
139 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
141 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
142 iterator. Delete separate integer-mode vec_set<mode> expander.
143 (aarch64_simd_vec_setv2di): Delete.
144 (vec_setv2di): Delete.
145 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
146 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
147 the "w, r" alternative.
149 2018-05-18 Martin Liska <mliska@suse.cz>
151 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
152 * tree-pass.h (make_pass_lower_switch_O0): New function.
153 * tree-switch-conversion.c (node_has_low_bound): Remove.
154 (node_has_high_bound): Likewise.
155 (node_is_bounded): Likewise.
156 (class pass_lower_switch): Make it a template type and create
158 (pass_lower_switch::execute): Add template argument.
159 (make_pass_lower_switch): New function.
160 (make_pass_lower_switch_O0): New function.
161 (do_jump_if_equal): Remove.
162 (emit_case_nodes): Simplify to just handle all 3 cases and leave
163 all the hard work to tree optimization passes.
165 2018-05-18 Martin Liska <mliska@suse.cz>
167 * dbgcnt.c (limit_low): Renamed from limit.
168 (limit_high): New variable.
169 (dbg_cnt_is_enabled): Check for upper limit.
170 (dbg_cnt): Adjust dumping.
171 (dbg_cnt_set_limit_by_index): Add new argument for high
173 (dbg_cnt_set_limit_by_name): Likewise.
174 (dbg_cnt_process_single_pair): Parse new format.
175 (dbg_cnt_process_opt): Use strtok.
176 (dbg_cnt_list_all_counters): Remove 'value' and add
178 * doc/invoke.texi: Document changes.
180 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
182 * doc/sourcebuild.texi (scalar_all_fma): Document.
183 * tree.def (FMA_EXPR): Delete.
184 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
185 * internal-fn.c (ternary_direct): New macro.
186 (expand_ternary_optab_fn): Likewise.
187 (direct_ternary_optab_supported_p): Likewise.
188 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
189 * builtins.c (fold_builtin_fma): Delete.
190 (fold_builtin_3): Don't call it.
191 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
192 * expr.c (expand_expr_real_2): Likewise.
193 * fold-const.c (operand_equal_p): Likewise.
194 (fold_ternary_loc): Likewise.
195 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
196 * gimple.c (DEFTREECODE): Likewise.
197 * gimplify.c (gimplify_expr): Likewise.
198 * optabs-tree.c (optab_for_tree_code): Likewise.
199 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
200 * tree-eh.c (operation_could_trap_p): Likewise.
201 (stmt_could_throw_1_p): Likewise.
202 * tree-inline.c (estimate_operator_cost): Likewise.
203 * tree-pretty-print.c (dump_generic_node): Likewise.
204 (op_code_prio): Likewise.
205 * tree-ssa-loop-im.c (stmt_cost): Likewise.
206 * tree-ssa-operands.c (get_expr_operands): Likewise.
207 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
208 * fold-const-call.h (fold_fma): Delete.
209 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
210 CFN_FNMA and CFN_FNMS.
212 * genmatch.c (combined_fn): New enum.
213 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
214 (commutative_op): New function.
215 (commutate): Use it. Handle more than 2 operands.
216 (dt_operand::gen_gimple_expr): Use commutative_op.
217 (parser::parse_expr): Allow :c to be used with non-binary
218 operators if the commutative operand is known.
219 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
220 CFN_FMS, CFN_FNMA and CFN_FNMS.
221 (backprop::process_assign_use): Remove FMA_EXPR handling.
222 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
223 (gen_hsa_fma): New function.
224 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
225 IFN_FNMA and IFN_FNMS.
226 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
227 * gimple-fold.h (follow_all_ssa_edges): Declare.
228 * gimple-fold.c (follow_all_ssa_edges): New function.
229 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
230 gimple_build interface and use follow_all_ssa_edges to fold the result.
231 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
232 instead of checking for optabs directly.
233 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
234 rather than FMA_EXPRs.
235 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
236 call to IFN_FMA instead of an FMA_EXPR.
238 2018-05-17 Jim Wilson <jimw@sifive.com>
240 * expr.c (do_tablejump): When converting index to Pmode, if we have a
241 sign extended promoted subreg, and the range does not have the sign bit
242 set, then do a sign extend.
244 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
245 test, check for sign extended subreg and/or constant operands, and
246 do a sign extend in that case.
248 2018-05-17 Steve Ellcey <sellcey@cavium.com>
250 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
251 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
253 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
254 Change logics_shift_reg to logics_shift_imm.
255 (thunderx2t99_fp_loadpair_basic): Delete.
256 (thunderx2t99_fp_storepair_basic): Delete.
257 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
258 (thunderx2t99_asimd_polynomial): Delete.
259 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
260 and neon_fp_mul_d_scalar_q.
261 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
262 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
263 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
264 (thunderx2t99_asimd_lut): Add missing tbl types.
265 (thunderx2t99_asimd_ext): Delete.
266 (thunderx2t99_asimd_load1_1_mult): Delete.
267 (thunderx2t99_asimd_load1_2_mult): Delete.
268 (thunderx2t99_asimd_load1_ldp): New.
269 (thunderx2t99_asimd_load1): New.
270 (thunderx2t99_asimd_load2): Add missing *load2* types.
271 (thunderx2t99_asimd_load3): New.
272 (thunderx2t99_asimd_load4): New.
273 (thunderx2t99_asimd_store1_1_mult): Delete.
274 (thunderx2t99_asimd_store1_2_mult): Delete.
275 (thunderx2t99_asimd_store2_mult): Delete.
276 (thunderx2t99_asimd_store2_onelane): Delete.
277 (thunderx2t99_asimd_store_stp): New.
278 (thunderx2t99_asimd_store1): New.
279 (thunderx2t99_asimd_store2): New.
280 (thunderx2t99_asimd_store3): New.
281 (thunderx2t99_asimd_store4): New.
283 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
285 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
286 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
288 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
289 Segher Boessenkool <segher@kernel.crashing.org>
292 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
295 2018-05-17 Richard Biener <rguenther@suse.de>
297 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
298 for pruning loop and prune defs feeding only already visited PHIs.
300 2018-05-17 Richard Biener <rguenther@suse.de>
302 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
304 2018-05-17 Bin Cheng <bin.cheng@arm.com>
305 Richard Biener <rguenther@suse.de>
307 PR tree-optimization/85793
308 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
309 for VMAT_ELEMENTWISE.
311 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
313 * internal-fn.h (lookup_internal_fn): Declare
314 * internal-fn.c (lookup_internal_fn): New function.
315 * gimple.c (gimple_build_call_from_tree): Handle calls to
317 * gimple-pretty-print.c (dump_gimple_call): Print "." before
318 internal function names.
319 * tree-pretty-print.c (dump_generic_node): Likewise.
320 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
322 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
324 * gimple-fold.h (gimple_build): Make the function forms take
325 combined_fn rather than built_in_function.
326 (gimple_simplify): Likewise.
327 * gimple-match-head.c (gimple_simplify): Likewise.
328 * gimple-fold.c (gimple_build): Likewise.
329 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
330 rather than gimple_build_call_internal.
331 (get_initial_defs_for_reduction): Likewise.
332 (vect_create_epilog_for_reduction): Likewise.
333 (vectorizable_live_operation): Likewise.
335 2018-05-17 Martin Liska <mliska@suse.cz>
337 * gimple-ssa-sprintf.c (format_directive): Do not use
338 space in between 'G_' and '('.
340 2018-05-17 Jakub Jelinek <jakub@redhat.com>
343 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
344 even if the mask is not all ones.
347 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
349 (ix86_gimple_fold_builtin): Likewise.
352 * config/i386/i386.c: Include tree-vector-builder.h.
353 (ix86_vector_shift_count): New function.
354 (ix86_fold_builtin): Fold shift builtins by scalar count.
355 (ix86_gimple_fold_builtin): Likewise.
357 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
358 _mm512_setzero): New intrinsics.
360 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
361 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
363 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
364 code generation for cases where splatting a value is not useful.
365 * simplify-rtx.c (simplify_ternary_operation): Simplify
366 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
367 mode to a vec_concat.
369 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
371 * config.gcc: Support "goldmont-plus".
372 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
374 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
375 PROCESSOR_GOLDMONT_PLUS.
376 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
377 (processor_target_table): Add "goldmont-plus".
378 (PTA_GOLDMONT_PLUS): Define.
379 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
380 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
381 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
382 (fold_builtin_cpu): Add "goldmont-plus".
383 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
384 (ix86_option_override_internal): Add "goldmont-plus".
385 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
386 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
387 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
388 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
390 2018-05-17 Richard Biener <rguenther@suse.de>
392 PR tree-optimization/85757
393 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
394 remove defs that only feed that PHI from further processing.
396 2018-05-16 Jim Wilson <jimw@sifive.com>
398 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
400 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
402 2018-05-16 Mark Wielaard <mark@klomp.org>
404 * dwarf2out.c (count_index_strings): New function.
405 (output_indirect_strings): Call count_index_strings and generate
406 header for dwarf_version >= 5.
408 2018-05-16 Mark Wielaard <mark@klomp.org>
410 * dwarf2out.c (dwarf_FORM): New function.
411 (set_indirect_string): Use dwarf_FORM.
412 (reset_indirect_string): Likewise.
413 (size_of_die): Likewise.
414 (value_format): Likewise.
415 (output_die): Likewise.
416 (add_skeleton_AT_string): Likewise.
417 (output_macinfo_op): Likewise.
418 (index_string): Likewise.
419 (output_index_string_offset): Likewise.
420 (output_index_string): Likewise.
421 (count_index_strings): Likewise.
423 2018-05-16 Carl Love <cel@us.ibm.com>
425 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
426 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
428 2018-05-16 Martin Jambor <mjambor@suse.cz>
430 * ipa-prop.c (ipa_free_all_edge_args): Remove.
431 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
433 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
435 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
436 (fnma<mode>4): Likewise.
437 (fms<mode>4): Likewise.
438 (fnms<mode>4): Likewise.
439 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
440 (aarch64_fnma<mode>4): Likewise.
441 (aarch64_fms<mode>4): Likewise.
442 (aarch64_fnms<mode>4): Likewise.
443 (aarch64_fnmadd<mode>4): Likewise.
445 2018-05-16 Jason Merrill <jason@redhat.com>
447 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
449 2018-05-16 Richard Biener <rguenther@suse.de>
451 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
452 (dump_stmt_cost): Declare.
453 (add_stmt_cost): Dump cost we add.
454 (add_stmt_costs): New function.
455 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
457 (vect_analyze_stmt): Adjust prototype.
458 (vectorizable_condition): Likewise.
459 (vectorizable_live_operation): Likewise.
460 (vectorizable_reduction): Likewise.
461 (vectorizable_induction): Likewise.
462 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
463 cost vector to pass to vectorizable_ and record afterwards.
464 (vect_model_reduction_cost): Take cost vector argument and adjust.
465 (vect_model_induction_cost): Likewise.
466 (vectorizable_reduction): Likewise.
467 (vectorizable_induction): Likewise.
468 (vectorizable_live_operation): Likewise.
469 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
470 SLP_TREE_NUMBER_OF_VEC_STMTS.
471 (vect_analyze_slp_cost_1): Remove.
472 (vect_analyze_slp_cost): Likewise.
473 (vect_slp_analyze_node_operations): Take visited args and
474 a target cost vector. Avoid processing already visited stmt sets.
475 (vect_slp_analyze_operations): Use a local cost vector to gather
476 costs and register those of non-discarded instances.
477 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
478 (vect_schedule_slp_instance): Remove copying of
479 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
481 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
482 adding cost. Record cost entry location.
483 (vect_prologue_cost_for_slp_op): Function to compute cost of
484 a constant or invariant generated for SLP vect in the prologue,
485 split out from vect_analyze_slp_cost_1.
486 (vect_model_simple_cost): Make static. Adjust for SLP costing.
487 (vect_model_promotion_demotion_cost): Likewise.
488 (vect_model_store_cost): Likewise, make static.
489 (vect_model_load_cost): Likewise.
490 (vectorizable_bswap): Add cost vector arg and adjust.
491 (vectorizable_call): Likewise.
492 (vectorizable_simd_clone_call): Likewise.
493 (vectorizable_conversion): Likewise.
494 (vectorizable_assignment): Likewise.
495 (vectorizable_shift): Likewise.
496 (vectorizable_operation): Likewise.
497 (vectorizable_store): Likewise.
498 (vectorizable_load): Likewise.
499 (vectorizable_condition): Likewise.
500 (vectorizable_comparison): Likewise.
501 (can_vectorize_live_stmts): Likewise.
502 (vect_analyze_stmt): Likewise.
503 (vect_transform_stmt): Adjust calls to vectorizable_*.
504 * tree-vectorizer.c: Include gimple-pretty-print.h.
505 (dump_stmt_cost): New function.
507 2018-05-16 Richard Biener <rguenther@suse.de>
509 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
510 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
511 * tree-ssa-dse.c: Include tree-ssa-loop.h.
512 (check_name): New callback.
513 (dse_classify_store): Track cycles via a visited bitmap of PHI
514 defs and simplify handling of in-loop and across loop dead stores
515 and properly fail for loop-variant refs. Handle byte-tracking with
516 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
519 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
521 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
522 (vect_get_mask_type_for_stmt): Likewise.
523 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
525 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
526 to determine the statement's vector type and the vector type that
527 should be used for calculating nunits. Deal with cases in which
528 the type has to be deferred.
529 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
530 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
531 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
532 (vect_determine_vf_for_stmt): New functions, split out from...
533 (vect_determine_vectorization_factor): ...here.
534 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
535 (vect_get_mask_type_for_stmt): New functions, split out from
536 vect_determine_vectorization_factor.
538 2018-05-16 Richard Biener <rguenther@suse.de>
540 * tree-cfg.c (verify_gimple_assign_ternary): Properly
541 verify the [VEC_]COND_EXPR embedded comparison.
543 2018-05-15 Martin Sebor <msebor@redhat.com>
545 PR tree-optimization/85753
546 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
547 RECORD_TYPE in addition to ARRAY_TYPE.
549 2018-05-15 Martin Sebor <msebor@redhat.com>
552 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
554 2018-05-15 Richard Biener <rguenther@suse.de>
556 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
557 add by_clobber_p one. Change algorithm to collect all defs
558 representing uses we need to walk and try reducing them to
559 a single one before failing.
560 (dse_dom_walker::dse_optimize_stmt): Adjust.
562 2018-05-13 Mark Wielaard <mark@klomp.org>
564 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
565 (size_of_loc_descr): Likewise.
566 (output_loc_operands): Likewise.
567 (output_loc_operands_raw): Likewise.
568 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
569 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
570 (hash_loc_operands): Likewise.
571 (compare_loc_operands): Likewise.
573 2018-05-14 Mark Wielaard <mark@klomp.org>
575 * dwarf2out.c (count_index_addrs): New function.
576 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
578 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
580 PR tree-optimization/83648
581 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
582 return value as malloc candidate.
584 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
587 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
588 as true in call to suggest_attribute.
590 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
592 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
595 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
597 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
599 (aarch64_crypto_aesd_fused): Likewise.
601 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
603 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
604 (movsi_aarch64): Likewise.
605 (load_pairsi): Likewise.
606 (load_pairdi): Likewise.
607 (store_pairsi): Likewise.
608 (store_pairdi): Likewise.
609 (load_pairsf): Likewise.
610 (load_pairdf): Likewise.
611 (store_pairsf): Likewise.
612 (store_pairdf): Likewise.
613 (zero_extend): Likewise.
614 (trunc): Swap alternatives.
615 (fcvt_target): Add '?' to prefer w over r.
617 2018-05-14 Jakub Jelinek <jakub@redhat.com>
620 * config/i386/i386.md: Disallow non-commutative arithmetics in
621 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
622 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
623 in the peephole2 before it.
625 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
627 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
628 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
629 (ix86_handle_option): Handle -mcldemote.
630 * config.gcc: New header.
631 * config/i386/cldemoteintrin.h: New file.
632 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
633 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
635 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
636 OPTION_MASK_ISA_CLDEMOTE.
637 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
638 (ix86_valid_target_attribute_inner_p): Ditto.
639 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
640 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
641 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
642 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
643 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
645 * config/i386/i386.opt: Add -mcldemote.
646 * config/i386/x86intrin.h: New header.
647 * doc/invoke.texi: Add -mcldemote.
649 2018-05-14 Richard Biener <rguenther@suse.de>
651 * doc/match-and-simplify.texi: Adjust :s documentation.
653 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
655 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
656 intended memcpy size.
657 (REORDER_45): Likewise.
659 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
662 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
663 * vec.c (qsort_chk): Use gcc_qsort.
664 * Makefile.in (OBJS-libcommon): Add sort.o.
665 (build/sort.o): New target. Use it...
666 (BUILD_RTL): ... here, and...
667 (build/gencfn-macros): ... here, and...
668 (build/genmatch): ... here.
670 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
671 Chung-Ju Wu <jasonwucj@gmail.com>
673 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
674 * config/nds32/nds32-graywolf.md: New file.
675 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
676 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
678 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
679 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
680 * config/nds32/nds32.md (pipeline_model): Add graywolf.
681 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
682 * config/nds32/pipelines.md: Include n15 settings.
684 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
685 Chung-Ju Wu <jasonwucj@gmail.com>
687 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
688 * config/nds32/nds32-n13.md: New file.
689 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
690 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
692 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
693 * config/nds32/nds32.md (pipeline_model): Add n13.
694 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
695 * config/nds32/pipelines.md: Include n13 settings.
697 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
698 Chung-Ju Wu <jasonwucj@gmail.com>
700 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
701 * config/nds32/nds32-n10.md: New file.
702 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
703 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
705 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
706 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
707 * config/nds32/nds32.md (pipeline_model): Add n10.
708 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
709 * config/nds32/pipelines.md: Include n10 settings.
711 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
712 Kito Cheng <kito.cheng@gmail.com>
713 Chung-Ju Wu <jasonwucj@gmail.com>
715 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
716 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
717 Add enum values for DSP extension instructions.
718 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
720 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
721 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
723 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
724 * config/nds32/nds32-dspext.md: New file for DSP implementation.
725 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
726 * config/nds32/nds32-intrinsic.md: Likewise.
727 * config/nds32/nds32_intrinsic.h: Likewise.
728 * config/nds32/nds32-md-auxiliary.c: Likewise.
729 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
730 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
731 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
732 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
733 * config/nds32/nds32-protos.h: New declarations for DSP extension.
734 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
735 TYPE_DMAC in switch statement.
736 * config/nds32/nds32.c: New checking and implementation for DSP
737 extension instructions.
738 * config/nds32/nds32.h: Likewise.
739 * config/nds32/nds32.md: Likewise.
740 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
741 * config/nds32/predicates.md: Implement new predicates for DSP
744 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
746 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
747 Reformat alternatives and attributes so it is easier to identify
748 which constraints/attributes go with which instruction.
749 (mov<mode>_hardfloat32, FMOVE64): Likewise.
750 (mov<mode>_softfloat32, FMOVE64): Likewise.
751 (mov<mode>_hardfloat64, FMOVE64): Likewise.
752 (mov<mode>_softfloat64, FMOVE64): Likewise.
754 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
756 * doc/extend.texi (PowerPC Built-in Functions): Rename this
758 (Basic PowerPC Built-in Functions): The new name of the
759 subsection previously known as "PowerPC Built-in Functions".
760 (Basic PowerPC Built-in Functions Available on all Configurations):
762 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
763 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
764 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
765 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
767 2018-05-11 Martin Jambor <mjambor@suse.cz>
770 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
773 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
776 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
778 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
780 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
781 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
782 (ix86_handle_option): Handle -mwaitpkg.
783 * config.gcc: New header.
784 * config/i386/cpuid.h (bit_WAITPKG): New bit.
785 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
786 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
788 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
789 OPTION_MASK_ISA_WAITPKG.
790 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
791 (ix86_option_override_internal): Add PTA_WAITPKG.
792 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
793 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
795 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
796 __builtin_ia32_umwait and __builtin_ia32_tpause.
797 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
798 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
799 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
800 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
801 UNSPECV_TPAUSE): New.
802 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
803 * config/i386/i386.opt: Add -mwaitpkg.
804 * config/i386/waitpkgintrin.h: New file.
805 * config/i386/x86intrin.h: New header.
806 * doc/invoke.texi: Add -mwaitpkg.
808 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
811 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
813 (cortex-m0): Use armv6s-m isa.
814 (cortex-m0plus): Likewise.
815 (cortex-m1): Likewise.
816 (cortex-m0.small-multiply): Likewise.
817 (cortex-m0plus.small-multiply): Likewise.
818 (cortex-m1.small-multiply): Likewise.
820 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
821 Jakub Jelinek <jakub@redhat.com>
823 PR tree-optimization/85692
824 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
825 source permute as well.
827 2018-05-11 Martin Liska <mliska@suse.cz>
830 * doc/extend.texi: Document LLVM style format for no_sanitize
833 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
835 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
836 mode_supports_vsx_dform_quad to mode_supports_dq_form.
837 (mode_supports_vsx_dform_quad): Likewise.
838 (mode_supports_vmx_dform): Move these functions to be next to the
839 other mode_supports functions.
840 (mode_supports_dq_form): Likewise.
841 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
842 mode_supports_dq_form.
843 (reg_offset_addressing_ok_p): Likewise.
844 (offsettable_ok_by_alignment): Likewise.
845 (rs6000_legitimate_offset_address_p): Likewise.
846 (legitimate_lo_sum_address_p): Likewise.
847 (rs6000_legitimize_address): Likewise.
848 (rs6000_legitimize_reload_address): Likewise.
849 (rs6000_secondary_reload_inner): Likewise.
850 (rs6000_preferred_reload_class): Likewise.
851 (rs6000_output_move_128bit): Likewise.
853 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
855 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
856 Generate SImode target register for null target.
857 <case IX86_BUILTIN_XGETBV>: Ditto.
858 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
859 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
861 2018-05-10 Carl Love <cel@us.ibm.com>
863 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
864 dcbtt and dcbtstt if operands[2] is 0.
866 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
869 * config/i386/sse.md (usadv64qi): New expander.
871 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
873 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
874 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
875 -maltivec=be support.
876 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
877 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
878 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
879 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
880 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
881 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
882 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
883 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
884 altivec_vsumsws): Adjust.
885 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
886 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
887 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
889 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
890 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
891 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
892 (altivec_lve<VI_char>x): Delete expand.
893 (*altivec_lve<VI_char>x_internal): Rename to...
894 (altivec_lve<VI_char>x): ... this.
895 (altivec_lvxl_<mode>): Delete expand.
896 (*altivec_lvxl_<mode>_internal): Rename to ...
897 (altivec_lvxl_<mode>): ... this.
898 (altivec_stvxl_<mode>): Delete expand.
899 (*altivec_stvxl_<mode>_internal): Rename to ...
900 (altivec_stvxl_<mode>): ... this.
901 (altivec_stve<VI_char>x): Delete expand.
902 (*altivec_stve<VI_char>x_internal): Rename to ...
903 (altivec_stve<VI_char>x): ... this.
904 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
905 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
906 reduc_plus_scal_<mode>): Adjust.
907 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
909 (rs6000_cpu_cpp_builtins): Adjust.
910 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
911 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
912 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
913 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
914 -maltivec=be support.
915 (rs6000_split_vec_extract_var): Adjust.
916 (rs6000_split_v4si_init): Adjust.
917 (swap_selector_for_mode): Delete.
918 (altivec_expand_lvx_be, altivec_expand_stvx_be,
919 altivec_expand_stvex_be): Delete.
920 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
921 -maltivec=be support.
922 (rs6000_gimple_fold_builtin): Ditto.
923 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
925 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
926 (TARGET_DIRECT_MOVE_64BIT): Adjust.
927 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
928 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
929 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
930 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
931 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
932 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
933 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
934 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
935 anonymous split): Adjust.
936 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
937 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
939 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
941 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
942 when --with-gxx-include-dir is also specified.
943 * configure: Regenerate.
945 2018-05-09 Jim Wilson <jimw@sifive.com>
948 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
949 * config/riscv/t-withmultilib: New.
950 * config/riscv/withmultilib.h: New.
951 * doc/install.texi: Document RISC-V --with-multilib-list support.
953 2018-05-09 Richard Biener <rguenther@suse.de>
955 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
957 (vect_bb_vectorization_profitable_p): Adjust. Compute
958 actual scalar cost using the cost vector and the add_stmt_cost
961 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
963 PR rtl-optimization/85645
964 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
965 in the REG_CFA_REGISTER note for LR, don't leave it empty.
967 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
969 PR rtl-optimization/85645
970 * shrink-wrap.c (spread_components): Return a boolean saying if
971 anything was changed.
972 (try_shrink_wrapping_separate): Iterate spread_components until
973 nothing changes anymore.
975 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
977 PR rtl-optimization/85645
978 * regrename.c (build_def_use): Also kill the chains that include the
979 destination of a REG_CFA_REGISTER note.
981 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
983 PR rtl-optimization/85645
984 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
985 insn that has a REG_CFA_REGISTER note.
987 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
989 * cfgexpand.c (expand_clobber): New function.
990 (expand_gimple_stmt_1): Use it.
991 * tree-vect-stmts.c (vect_clobber_variable): New function,
993 (vectorizable_simd_clone_call): ...here.
994 (vectorizable_store): Emit a clobber either side of an
995 IFN_STORE_LANES sequence.
996 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
998 2018-05-09 Tom de Vries <tom@codesourcery.com>
1001 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1002 (define_insn "trap_if_false"): Add exit after trap.
1004 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1006 PR rtl-optimization/85638
1007 * bb-reorder.c: Include common/common-target.h.
1008 (create_forwarder_block): New function extracted from...
1009 (fix_up_crossing_landing_pad): ...here. Rename into...
1010 (dw2_fix_up_crossing_landing_pad): ...this.
1011 (sjlj_fix_up_crossing_landing_pad): New function.
1012 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1013 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1014 from both partitions and exit the loop after one iteration.
1016 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1019 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1021 (Basic PowerPC Built-in Functions): The new name of the
1022 subsection previously known as "PowerPC Built-in Functions".
1023 (Basic PowerPC Built-in Functions Available on all Configurations):
1025 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1027 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1029 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1031 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1034 2018-05-08 Jim Wilson <jimw@sifive.com>
1036 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1037 (LD_EMUL_SUFFIX): New.
1038 (LINK_SPEC): Use it.
1040 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1042 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1044 (Basic PowerPC Built-in Functions): The new name of the
1045 subsection previously known as "PowerPC Built-in Functions".
1046 (Basic PowerPC Built-in Functions Available on all Configurations):
1048 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1050 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1052 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1054 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1057 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1060 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1061 after cmpelim optimization.
1063 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1065 * config.gcc: Support "goldmont".
1066 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1067 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1069 * config/i386/i386.c (m_GOLDMONT): Define.
1070 (processor_target_table): Add "goldmont".
1071 (PTA_GOLDMONT): Define.
1072 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1073 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1074 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1075 (fold_builtin_cpu): Add "goldmont".
1076 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1077 (ix86_option_override_internal): Add "goldmont".
1078 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1079 (processor_type): Add PROCESSOR_GOLDMONT.
1080 * config/i386/i386.md: Add CPU "glm".
1081 * config/i386/glm.md: New file.
1082 * config/i386/x86-tune.def: Add m_GOLDMONT.
1083 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1085 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1088 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1090 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1091 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1092 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1095 * config/i386/i386.c (ix86_fold_builtin): Handle
1096 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1099 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1100 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1102 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1105 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1106 (check_arch): Likewise.
1107 (check_fpu): Return the result rather than printing it.
1108 (end arch): Fix operator precedence.
1109 (end cpu): Likewise.
1110 (END): Print the result from check_fpu.
1112 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1113 Alan Hayward <alan.hayward@arm.com>
1114 David Sherwood <david.sherwood@arm.com>
1116 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1117 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1118 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1119 (*fcmuo<mode>_and): New patterns.
1121 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1123 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1124 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1125 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1126 (cmp_op, sve_imm_con): New code attributes.
1127 (SVE_COND_INT_CMP, imm_con): Delete.
1128 (cmp_op): Remove above unspecs from int attribute.
1129 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1131 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1132 comparison-specific unspecs.
1133 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1134 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1135 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1136 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1137 (*vec_fcm<cmp_op><mode>): Rename to...
1138 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1139 (*vec_fcmuo<mode>): Rename to...
1140 (*fcmuo<mode>): ...this and adjust likewise.
1141 (*pred_fcm<cmp_op><mode>): New pattern.
1142 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1143 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1145 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1147 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1148 (aarch64_emit_sve_predicated_cond): New function.
1149 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1150 (aarch64_emit_unspec_cond_or): Replace with...
1151 (aarch64_emit_sve_or_conds): ...this new function. Use
1152 aarch64_emit_sve_ptrue_op for the individual comparisons and
1153 aarch64_emit_binop to OR them together.
1154 (aarch64_emit_inverted_unspec_cond): Replace with...
1155 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1156 aarch64_emit_sve_ptrue_op for the comparison and
1157 aarch64_emit_unop to invert the result.
1158 (aarch64_expand_sve_vec_cmp_float): Update after the above
1159 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1161 2018-05-07 Nathan Sidwell <nathan@acm.org>
1163 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1164 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1165 (Backwards Compatibility): Likewise.
1167 2018-05-07 Luis Machado <luis.machado@linaro.org>
1171 2018-05-07 Luis Machado <luis.machado@linaro.org>
1173 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1174 <prefetch_dynamic_strides>: New const bool field.
1175 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1176 prefetch_dynamic_strides.
1177 (exynosm1_prefetch_tune): Likewise.
1178 (thunderxt88_prefetch_tune): Likewise.
1179 (thunderx_prefetch_tune): Likewise.
1180 (thunderx2t99_prefetch_tune): Likewise.
1181 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1182 (aarch64_override_options_internal): Update to set
1183 PARAM_PREFETCH_DYNAMIC_STRIDES.
1184 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1185 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1186 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1187 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1188 prefetch-dynamic-strides setting.
1190 2018-05-07 Luis Machado <luis.machado@linaro.org>
1192 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1193 <minimum_stride>: New const int field.
1194 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1195 minimum_stride field.
1196 (exynosm1_prefetch_tune): Likewise.
1197 (thunderxt88_prefetch_tune): Likewise.
1198 (thunderx_prefetch_tune): Likewise.
1199 (thunderx2t99_prefetch_tune): Likewise.
1200 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1201 (aarch64_override_options_internal): Update to set
1202 PARAM_PREFETCH_MINIMUM_STRIDE.
1203 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1204 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1205 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1206 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1207 stride is constant and is below the minimum stride threshold.
1209 2018-05-07 Luis Machado <luis.machado@linaro.org>
1211 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1214 2018-05-07 Luis Machado <luis.machado@linaro.org>
1216 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1217 <prefetch_dynamic_strides>: New const bool field.
1218 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1219 prefetch_dynamic_strides.
1220 (exynosm1_prefetch_tune): Likewise.
1221 (thunderxt88_prefetch_tune): Likewise.
1222 (thunderx_prefetch_tune): Likewise.
1223 (thunderx2t99_prefetch_tune): Likewise.
1224 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1225 (aarch64_override_options_internal): Update to set
1226 PARAM_PREFETCH_DYNAMIC_STRIDES.
1227 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1228 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1229 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1230 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1231 prefetch-dynamic-strides setting.
1233 2018-05-07 Luis Machado <luis.machado@linaro.org>
1235 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1236 <minimum_stride>: New const int field.
1237 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1238 minimum_stride field.
1239 (exynosm1_prefetch_tune): Likewise.
1240 (thunderxt88_prefetch_tune): Likewise.
1241 (thunderx_prefetch_tune): Likewise.
1242 (thunderx2t99_prefetch_tune): Likewise.
1243 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1244 (aarch64_override_options_internal): Update to set
1245 PARAM_PREFETCH_MINIMUM_STRIDE.
1246 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1247 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1248 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1249 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1250 stride is constant and is below the minimum stride threshold.
1252 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1255 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1256 the type is addressable. Don't force op into register if it has
1259 2018-05-05 Roland McGrath <mcgrathr@google.com>
1262 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1263 any section for which we don't know a specific type it should have,
1264 regardless of name. Previously this was done only for the exact
1265 names ".init_array", ".fini_array", and ".preinit_array".
1266 (default_elf_asm_named_section): Add comment about
1267 relationship with default_section_type_flags and SECTION_NOTYPE.
1268 (get_section): Don't consider it a type conflict if one side has
1269 SECTION_NOTYPE and the other doesn't, as long as neither has the
1270 SECTION_BSS et al used in the default_section_type_flags logic.
1272 2018-05-05 Tom de Vries <tom@codesourcery.com>
1275 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1276 (workaround_barsyncs): New function.
1277 (nvptx_reorg): Use workaround_barsyncs.
1278 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1279 (define_expand "nvptx_membar_cta"): New define_expand.
1280 (define_insn "*nvptx_membar_cta"): New insn.
1282 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1284 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1285 To improve optimization opportunities.
1286 * builtin-types.def: The new needed builtin types for the above.
1288 2018-05-04 Richard Biener <rguenther@suse.de>
1290 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1291 * gimple-ssa-store-merging.c
1292 (imm_store_chain_info::output_merged_store): Remove redundant create,
1293 release split_store vector contents on failure.
1294 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1295 scalar stmt vector on cache hit.
1297 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1299 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1301 * config.gcc (powerpc-xilinx-eabi*): Remove.
1302 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1304 (fusion_addis_mem_combo_load): Ditto.
1305 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1307 (rs6000_cpu_cpp_builtins): Ditto.
1308 * config/rs6000/rs6000-linux.c
1309 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1310 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1311 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1313 (rs6000_setup_reg_addr_masks): Ditto.
1314 (rs6000_init_hard_regno_mode_ok): Ditto.
1315 (rs6000_option_override_internal): Ditto.
1316 (legitimate_lo_sum_address_p): Ditto.
1317 (rs6000_legitimize_address): Ditto.
1318 (rs6000_legitimize_reload_address): Ditto.
1319 (rs6000_legitimate_address_p): Ditto.
1320 (abi_v4_pass_in_fpr): Ditto.
1321 (setup_incoming_varargs): Ditto.
1322 (rs6000_gimplify_va_arg): Ditto.
1323 (rs6000_split_multireg_move): Ditto.
1324 (rs6000_savres_strategy): Ditto.
1325 (rs6000_emit_prologue_components): Ditto.
1326 (rs6000_emit_epilogue_components): Ditto.
1327 (rs6000_emit_prologue): Ditto.
1328 (rs6000_emit_epilogue): Ditto.
1329 (rs6000_elf_file_end): Ditto.
1330 (rs6000_function_value): Ditto.
1331 (rs6000_libcall_value): Ditto.
1332 * config/rs6000/rs6000.h: Ditto.
1333 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1334 (TARGET_MINMAX): ... this. New.
1335 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1336 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1337 (*movsi_internal1_single): Delete.
1338 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1339 mfpu=, mxilinx-fpu): Delete.
1340 * config/rs6000/singlefp.h: Delete.
1341 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1342 * config/rs6000/t-rs6000: Ditto.
1343 * config/rs6000/t-xilinx: Delete.
1344 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1345 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1346 (VStype_simple): Delete.
1347 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1348 * config/rs6000/xfpu.h: Delete.
1349 * config/rs6000/xfpu.md: Delete.
1350 * config/rs6000/xilinx.h: Delete.
1351 * config/rs6000/xilinx.opt: Delete.
1352 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1353 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1355 2018-05-04 Tom de Vries <tom@codesourcery.com>
1358 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1361 2018-05-04 Richard Biener <rguenther@suse.de>
1364 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1365 (expand_complex_div_wide): Likewise.
1366 (expand_complex_operations_1): Likewise.
1367 (expand_complex_libcall): Preserve EH info of the original stmt.
1368 (tree_lower_complex): Handle removed blocks.
1369 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1370 on complex multiplication and division libcall builtins.
1372 2018-05-04 Richard Biener <rguenther@suse.de>
1375 * fold-const.c (negate_expr_p): Restrict negation of operand
1376 zero of a division to when we know that can happen without
1378 (fold_negate_expr_1): Likewise.
1380 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1383 * real.h (real_nextafter): Declare.
1384 * real.c (real_nextafter): New function.
1385 * fold-const-call.c (fold_const_nextafter): New function.
1386 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1387 CASE_CFN_NEXTTOWARD.
1388 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1389 even when arg1_mode is different from arg0_mode.
1391 2018-05-03 Nathan Sidwell <nathan@acm.org>
1393 * doc/extend.texi (Deprecated Features): Remove
1395 (Backwards Compatibility): Likewise.
1396 * doc/invoke.texi (C++ Language Options): Likewise.
1397 (C++ Dialect Options): Likewise.
1399 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1402 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1403 _mm512_mask_mullox_epi64): New intrinsics.
1405 2018-05-03 Tom de Vries <tom@codesourcery.com>
1408 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1409 dump files): Add offload-tree.
1411 2018-05-03 Richard Biener <rguenther@suse.de>
1413 PR tree-optimization/85615
1414 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1415 to loops not nested in BBs loop father to avoid creating multi-entry
1418 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1420 PR tree-optimization/70291
1421 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1422 arguments. Change return type to tree. Emit libcall as a new
1423 statement rather than replacing existing one when inplace_p is true.
1424 (expand_complex_multiplication_components): New function.
1425 (expand_complex_multiplication): Expand floating-point complex
1426 multiplication using the above.
1427 (expand_complex_division): Rename inner_type parameter to type.
1428 Update expand_complex_libcall call-site.
1429 (expand_complex_operations_1): Update expand_complex_multiplication
1430 and expand_complex_division call-sites.
1432 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1435 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1436 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1437 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1438 the highest significant bit of the shift count mask is clear. In
1439 check whether and[sq]i3 is needed verify that all significant bits
1440 of the shift count other than the highest are set.
1442 2018-05-02 Tom de Vries <tom@codesourcery.com>
1445 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1446 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1447 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1448 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1449 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1450 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1451 __builtin_goacc_parlevel_size.
1453 2018-05-02 Richard Biener <rguenther@suse.de>
1455 PR tree-optimization/85597
1456 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1457 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1460 2018-05-02 Tom de Vries <tom@codesourcery.com>
1463 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1464 dump files): Add ltrans-tree.
1466 2018-05-02 Tom de Vries <tom@codesourcery.com>
1469 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1470 dump files): Add wpa-ipa.
1472 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1474 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1475 powerpc*-*-linux*paired* target.
1476 * config/rs6000/750cl.h: Delete.
1477 * config/rs6000/paired.h: Delete.
1478 * config/rs6000/paired.md: Delete.
1479 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1481 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1482 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1483 comment. Remove paired float support.
1484 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1485 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1487 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1488 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1490 * config/rs6000/rs6000.c: Remove paired float support.
1491 (paired_expand_vector_init, paired_expand_vector_move,
1492 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1493 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1494 paired_expand_builtin, paired_expand_predicate_builtin,
1495 paired_init_builtins): Delete.
1496 * config/rs6000/rs6000.h: Remove paired float support.
1497 * config/rs6000/rs6000.md: Remove paired float support.
1498 (move_from_CR_ov_bit): Delete.
1499 * config/rs6000/rs6000.opt (mpaired): Delete.
1500 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1501 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1503 2018-05-02 Richard Biener <rguenther@suse.de>
1506 * gimplify.c (gimplify_save_expr): When in SSA form allow
1507 SAVE_EXPRs to compute to SSA vars.
1509 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1512 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1513 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1514 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1515 clobber operands[2], instead use a new pseudo. Formatting fixes.
1517 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1519 PR tree-optimization/85586
1520 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1521 exit early for statements in the same group if the accesses are
1524 2018-05-02 Tom de Vries <tom@codesourcery.com>
1527 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1530 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1532 PR tree-optimization/85143
1533 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1535 2018-05-01 Tom de Vries <tom@codesourcery.com>
1538 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1539 not found" error message.
1541 2018-05-01 Tom de Vries <tom@codesourcery.com>
1544 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1545 * vec.c (test_ordered_remove_if): New function.
1546 (vec_c_tests): Call test_ordered_remove_if.
1547 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1548 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1549 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1550 VEC_ORDERED_REMOVE_IF.
1552 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1554 PR tree-optimization/82665
1555 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1556 pointer subtraction where arguments come from a memchr call.
1558 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1560 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1561 --push-state --as-needed and --pop-state instead of --as-needed and
1562 --no-as-needed if ld supports it.
1563 * configure: Regenerated.
1566 * doc/install.texi2html: Replace _002d with - and _002a with * in
1567 generated html files using sed.
1569 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1572 * gcc-rich-location.c (blank_line_before_p): New function.
1573 (use_new_line): New function.
1574 (gcc_rich_location::add_fixit_insert_formatted): New function.
1575 * gcc-rich-location.h
1576 (gcc_rich_location::add_fixit_insert_formatted): New function.
1578 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1580 * selftest.c (assert_streq): Rename "expected" and "actual" to
1581 "val1" and "val2". Extend NULL-handling to cover both inputs
1582 symmetrically, while still requiring both to be non-NULL for a pass.
1583 * selftest.h (assert_streq): Rename "expected" and "actual" to
1585 (ASSERT_EQ): Likewise.
1586 (ASSERT_EQ_AT): Likewise.
1587 (ASSERT_KNOWN_EQ): Likewise.
1588 (ASSERT_KNOWN_EQ_AT): Likewise.
1589 (ASSERT_NE): Likewise.
1590 (ASSERT_MAYBE_NE): Likewise.
1591 (ASSERT_MAYBE_NE_AT): Likewise.
1592 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1593 the assertion to pass.
1594 (ASSERT_STREQ_AT): Likewise.
1596 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1598 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1599 interaction with -pie.
1601 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1603 * selftest.h: Fix alphabetization of per-source-file selftest
1606 2018-04-30 Jason Merrill <jason@redhat.com>
1608 PR c++/61982 - dead stores to destroyed objects.
1609 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1612 2018-04-30 Jason Merrill <jason@redhat.com>
1614 * tree.c (build_clobber): New.
1615 * tree.h: Declare it.
1616 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1618 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1620 * diagnostic-show-locus.c (layout::layout): Update for
1621 location_get_source_line returning a char_span.
1622 (struct char_span): Move to input.h.
1623 (struct correction): Update for fields in char_span becoming
1625 (struct source_line): Update for location_get_source_line
1626 returning a char_span.
1627 (layout::print_line): Likewise.
1628 * edit-context.c (edited_file::print_content): Likewise.
1629 (edited_file::print_diff_hunk): Likewise.
1630 (edited_file::print_run_of_changed_lines): Likewise.
1631 (edited_file::get_num_lines): Likewise.
1632 (edited_line::edited_line): Likewise.
1633 * final.c (asm_show_source): Likewise.
1634 * input.c (location_get_source_line): Convert return type
1635 from const char * to char_span, losing the final "line_len"
1637 (dump_location_info): Update for the above.
1638 (get_substring_ranges_for_loc): Likewise. Use a char_span
1639 when handling the literal within the line.
1640 (test_reading_source_line): Update for location_get_source_line
1641 returning a char_span.
1642 * input.h (class char_span): Move here from
1643 diagnostic-show-locus.c, converting from a struct to a class.
1644 Make data members private.
1645 (char_span::operator bool): New.
1646 (char_span::length): New.
1647 (char_span::get_buffer): New.
1648 (char_span::operator[]): New.
1649 (char_span::subspan): Make const.
1650 (char_span::xstrdup): New.
1651 (location_get_source_line): Convert return type from const char *
1652 to char_span, losing the final "line_size" param.
1654 2018-04-30 Jan Hubicka <jh@suse.cz>
1656 * lto-wrapper.c (ltrans_priorities): New static var.
1657 (cmp_priority): New.
1658 (run_gcc): Read priorities and if doing parallel build order
1659 the Makefile by them.
1661 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1663 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1665 2018-04-30 Richard Biener <rguenther@suse.de>
1667 * tree-cfg.c (verify_address): Remove base argument, add
1668 flag whether to check TREE_ADDRESSABLE and do that.
1669 (verify_expr): Remove.
1670 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1671 (verify_gimple_assign_single): Likewise.
1672 (verify_gimple_switch): Likewise.
1673 (verify_expr_location_1): Dereference tp once. Add (disabled)
1674 piece from verify_expr.
1675 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1677 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1679 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1681 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1683 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1684 (small_data_pattern): Likewise.
1685 (arc_rewrite_small_data): Likewise.
1686 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1687 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1688 (get_symbol_alignment): New function.
1689 (legitimate_small_data_address_p): Likewise.
1690 (legitimate_scaled_address): Update, call
1691 legitimate_small_data_address_p.
1692 (output_sdata): New static variable.
1693 (arc_print_operand): Update how we handle small data operands.
1694 (arc_print_operand_address): Likewise.
1695 (arc_legitimate_address_p): Update, use
1696 legitimate_small_data_address_p.
1697 (arc_rewrite_small_data_p): Remove.
1698 (arc_rewrite_small_data_1): Likewise.
1699 (arc_rewrite_small_data): Likewise.
1700 (small_data_pattern): Likewise.
1701 (compact_sda_memory_operand): Update to use
1702 legitimate_small_data_address_p and get_symbol_alignment.
1703 (prepare_move_operands): Don't rewite sdata pattern.
1704 (prepare_extend_operands): Remove.
1705 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1707 (zero_extendqisi2): Likewise.
1708 (zero_extendhisi2): Likewise.
1709 (extendqihi2): Likewise.
1710 (extendqisi2): Likewise.
1711 (extendhisi2): Likewise.
1715 * config/arc/constraints.md (Usd): Change it to memory constraint.
1717 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1719 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1720 as source of std instructions.
1721 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1722 allow 6-bit constants as source for store instructions.
1723 (movdi_insn): Update instruction pattern to allow 6-bit constants
1724 as source for store instructions.
1726 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1728 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1730 2018-04-30 Nathan Sidwell <nathan@acm.org>
1731 Sandra Loosemore <sandra@codesourcery.com>
1733 * dumpfile.c (dump_open): Allow '-' for stdout.
1734 * doc/invoke.texi (Developer Options): Document dump filename
1735 determination early. Document stdin/stdout selection.
1737 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1739 Microblaze Target: PIC data text relative
1741 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1742 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1744 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1745 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1746 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1747 New addressing mode for data-text relative position indepenedent code.
1748 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1749 'ADDRESS_SYMBOLIC_TXT_REL'.
1750 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1751 (microblaze_legitimate_pic_operand): Exclude function calls from
1752 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1753 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1755 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1756 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1757 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1758 for 'address + offset'.
1759 (microblaze_expand_prologue): Add new function prologue call for
1761 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1762 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1763 table in case of TARGET_PIC_DATA_TEXT_REL.
1764 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1765 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1766 Add new macros 'UNSPEC_TEXT',
1767 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1768 + exclude function calls from 'UNSPEC_PLT' in case of data text
1770 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1771 new target hook for generating address diff vector tables in case of
1773 * doc/tm.texi : Regenerate.
1774 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1775 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1776 of addr diff vector generation.
1777 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1778 target hook definition.
1779 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1780 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1781 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1782 option for data text relative.
1784 2018-04-30 Richard Biener <rguenther@suse.de>
1786 * tree-chrec.h (evolution_function_is_constant_p): Remove
1788 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1790 2018-04-30 Richard Biener <rguenther@suse.de>
1793 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1795 2018-04-30 Richard Biener <rguenther@suse.de>
1797 PR tree-optimization/28364
1798 PR tree-optimization/85275
1799 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1800 copying first exit test.
1802 2018-04-28 Mark Wielaard <mark@klomp.org>
1804 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1806 (dwarf_AT): Handle DW_AT_addr_base.
1807 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1809 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1812 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1813 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1814 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1815 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1817 2018-04-28 Richard Biener <rguenther@suse.de>
1819 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1820 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1821 to reflect use. Only add interesting stmts.
1823 2018-04-27 Martin Jambor <mjambor@suse.cz>
1826 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1827 the jump function allows for passing through aggregate values.
1829 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1831 * input.h (in_system_header_at): Convert from macro to inline
1833 (from_macro_expansion_at): Likewise.
1834 (from_macro_definition_at): Likewise.
1836 2018-04-27 Jeff Law <law@redhat.com>
1838 * config.gcc: Mark tile* targets as deprecated/obsolete.
1840 2018-04-27 Richard Biener <rguenther@suse.de>
1842 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1845 2018-04-27 Richard Biener <rguenther@suse.de>
1847 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1849 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1851 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1852 with Yd constraint. Set "preferred_for_speed" attribute from
1853 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1855 (*movdi_internal): Ditto.
1856 (movti_interunit splitters): Remove
1857 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1858 (movdi_interunit splitters): Ditto.
1859 * config/i386/constraints.md (Ye): Remove.
1860 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1862 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1865 * config/aarch64/constraints.md (Usg): Limit to 31.
1868 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1870 PR tree-optimization/85529
1871 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1872 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1873 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1874 zero extension or masking of the MSB bit.
1875 (optimize_range_tests): Add FIRST_BB argument, pass it through
1876 to optimize_range_tests_var_bound.
1877 (maybe_optimize_range_tests, reassociate_bb): Adjust
1878 optimize_range_tests callers.
1880 2018-04-26 Richard Biener <rguenther@suse.de>
1881 Jakub Jelinek <jakub@redhat.com>
1883 * cgraph.h (symbol_table): Just declare debug method here.
1884 * symtab.c (symbol_table::debug): Define.
1886 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1888 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1890 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1892 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1893 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1894 (*movdi_internal): Substitute Yi and Yj constraint with x
1895 and Ym and Yn constraint with y constraint. Update "isa"
1896 attribute and set "preferred_for_speed" attribute from
1897 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1898 (*movsi_internal): Ditto.
1899 (*movdf_internal): Ditto.
1900 (*movsf_internal): Ditto.
1901 (*zero_extendsidi2): Ditto.
1902 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1903 (sse2_loadld): Ditto.
1904 (*vec_extract<ssevecmodelower>_0): Ditto.
1905 (*vec_extractv4si_0_zext_sse4): Ditto.
1906 (vec_concatv2di): Ditto.
1907 (*vec_dup<mode>): Ditto.
1908 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1909 * config/i386/constraints.md (Yi): Remove.
1914 2018-04-26 Nathan Sidwell <nathan@acm.org>
1916 * dumpfile.c (dump_open): New.
1917 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1918 (dump_finish): Detect stdio/stderr by value not name.
1920 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1922 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1924 2018-04-26 Tom de Vries <tom@codesourcery.com>
1927 * config/nvptx/nvptx.c (verify_neutering_jumps)
1928 (verify_neutering_labels): New function
1929 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1931 2018-04-26 Tom de Vries <tom@codesourcery.com>
1934 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1935 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1938 2018-04-26 Richard Biener <rguenther@suse.de>
1939 Tom de Vries <tom@codesourcery.com>
1942 * lto-streamer-out.c (output_function): Fixup loops if required to match
1943 discovery done in the reader.
1945 2018-04-26 Richard Biener <rguenther@suse.de>
1947 PR tree-optimization/85116
1948 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1949 have a loop exit from the single latch predecessor. Remove
1950 case of header with just condition.
1951 (ch_base::copy_headers): Exclude infinite loops from any
1953 (pass_ch::execute): Record exits.
1955 2018-04-26 Richard Biener <rguenther@suse.de>
1957 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1958 prologue cost vector and pass it to vect_get_load_cost.
1959 (vect_get_peeling_costs_all_drs): Likewise.
1960 (vect_peeling_hash_get_lowest_cost): Likewise.
1961 (vect_enhance_data_refs_alignment): Likewise.
1963 2018-04-26 Richard Biener <rguenther@suse.de>
1966 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1967 checking of integer<->pointer conversions.
1968 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1969 sign-/zero-extending pointer types.
1970 (expand_omp_for_static_chunk): Likewise.
1972 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1973 Jean Lee <xiaoyur347@gmail.com>
1975 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1976 (TARGET_ASAN_SHADOW_OFFSET): Define.
1977 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1978 true for -fsanitize=address.
1980 2018-04-25 Mark Wielaard <mark@klomp.org>
1982 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1985 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1987 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1988 than "alu", remove explicit "memory" and "imm_disp" attributes.
1989 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1992 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1993 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1994 gen_lowpart_no_emit.
1996 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1999 * config/i386/i386.c (ix86_expand_builtin): Change memory
2000 operand to XI, extend p0 to Pmode.
2001 * config/i386/i386.md: Change unspec volatile and operand
2002 1 mode to XI, change operand 0 mode to P.
2004 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2006 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2007 GET_MODE_MASK before any checking.
2008 (nds32_can_use_bset_p): Likewise.
2009 (nds32_can_use_btgl_p): Likewise.
2011 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2013 * config/nds32/nds32-doubleword.md: New define_split pattern for
2014 illegal register number.
2016 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2018 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2020 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2022 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2024 2018-04-25 Richard Biener <rguenther@suse.de>
2026 * lto-streamer.h (LTO_major_version): Bump to 8.
2028 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2030 * BASE-VER: Set to 9.0.0.
2032 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2034 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2035 in __abskf2 and __powikf2.
2037 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2040 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2041 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2042 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2043 Use the above on operand 2. Reindent.
2044 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2046 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2049 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2051 * config/i386/i386.opt (mcet): Removed.
2052 * doc/install.texi: Remove -mcet documentation.
2053 * doc/invoke.texi: Likewise.
2055 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2058 * doc/install.texi: Remove -mcet from bootstrap-cet.
2060 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2063 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2064 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2068 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2069 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2070 containing a CONST_VECTOR.
2072 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2074 * doc/install.texi: Update newlib dependency for nvptx.
2076 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2079 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2080 instead of INTVAL when shifting x left.
2082 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2084 PR tree-optimization/85478
2085 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2086 vect_grouped_store_supported for single element vectors.
2088 2018-04-24 Richard Biener <rguenther@suse.de>
2091 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2092 load cost increase to the case of non-constant step.
2094 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2097 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2098 destination if any_malformed_asm.
2100 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2103 * expr.c (store_field): In the bitfield case, if the value comes from
2104 a function call and is returned in registers by means of a PARALLEL,
2105 do not change the mode of the temporary unless BLKmode and VOIDmode.
2107 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2109 PR rtl-optimization/85423
2110 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2111 dependencies to debug insns when the previous insn is non-debug.
2113 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2115 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2116 enums into a single definition.
2117 (fls): Fix predicates and printing.
2120 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2122 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2123 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2124 and short u6 immediate.
2125 (check_if_valid_sleep_operand): Remove.
2126 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2128 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2130 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2131 flag_always_save_lp condition.
2132 * config/nds32/nds32.opt (malways-save-lp): New option.
2134 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2136 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2137 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2138 * config/nds32/nds32.h
2139 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2140 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2142 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2144 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2145 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2147 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2148 Chung-Ju Wu <jasonwucj@gmail.com>
2150 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2152 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2153 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2155 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2157 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2159 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2161 * config/nds32/nds32-protos.h (nds32_data_alignment,
2162 nds32_local_alignment): Declare.
2163 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2164 nds32_local_alignment): New functions.
2165 (TARGET_CONSTANT_ALIGNMENT): Define.
2166 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2168 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2170 * config/nds32/nds32.c
2171 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2172 (TARGET_MODES_TIEABLE_P): Likewise.
2174 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2176 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2179 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2180 Chung-Ju Wu <jasonwucj@gmail.com>
2182 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2183 for unaligned access.
2184 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2186 * config/nds32/nds32-intrinsic.md: Likewise.
2187 * config/nds32/nds32_intrinsic.h: Likewise.
2188 * config/nds32/nds32.h (nds32_builtins): Likewise.
2189 * config/nds32/nds32.opt (munaligned-access): New option.
2190 * config/nds32/nds32.c (nds32_asm_file_start): Display
2191 flag_unaligned_access status.
2193 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2195 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2196 -mno-relax is present.
2197 * config/riscv/linux.h (LINK_SPEC): Ditto.
2199 2018-04-20 Martin Sebor <msebor@redhat.com>
2202 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2204 (gimple_fold_builtin_stxcpy_chk): Same.
2205 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2207 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2210 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2211 __powikf2 when long double is IEEE 128-bit.
2213 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2215 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2216 step to make sure stack always aligned.
2218 2018-04-20 Carl Love <cel@us.ibm.com>
2221 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2222 size check for arg0.
2224 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2225 Tom de Vries <tom@codesourcery.com>
2228 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2229 Emit insns for calls too.
2230 (nvptx_find_par): Always look for worker-level predecessor insn.
2231 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2233 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2234 (nvptx_process_pars): Propagate frames for calls.
2236 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2239 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2241 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2242 (ix86_handle_option): Don't handle OPT_mibt.
2243 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2245 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2247 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2248 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2249 (ix86_target_macros): Define __CET__ with flag_cf_protection
2250 for -fcf-protection.
2251 * config/i386/i386.c (isa2_opts): Remove -mibt.
2252 * config/i386/i386.h (TARGET_IBT): Removed.
2253 (TARGET_IBT_P): Likewise.
2254 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2255 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2256 * config/i386/i386.opt (mcet): Update help message.
2259 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2260 -mcet as an alias for -mshstk.
2262 2018-04-20 Richard Biener <rguenther@suse.de>
2265 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2266 complexity by forcing a single use of the multiply operand.
2268 2018-04-20 Martin Jambor <mjambor@suse.cz>
2271 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2272 recursion dependency to only apply to non-clones.
2274 2018-04-20 Martin Jambor <mjambor@suse.cz>
2277 * ipa-cp.c (create_specialized_node): Check that clones of
2278 self-recursive edges exist during IPA-CP.
2280 2018-04-19 Toon Moene <toon@moene.org>
2282 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2285 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2287 PR tree-optimization/85467
2288 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2289 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2290 VECTOR_CST element to type.
2292 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2295 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2296 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2297 (builtin_longjmp): Likewise.
2298 (save_stack_nonlocal): New pattern.
2299 (restore_stack_nonlocal): Likewise.
2301 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2304 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2305 Replace ASM_OUTPUT_LABEL with fprintf.
2307 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2310 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2311 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2312 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2313 define __IBT__ and __SHSTK__ for -fcf-protection.
2314 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2316 (ix86_trampoline_init): Likewise.
2317 (x86_output_mi_thunk): Likewise.
2318 (ix86_notrack_prefixed_insn_p): Likewise.
2319 (ix86_option_override_internal): Don't disallow -fcf-protection.
2320 * config/i386/i386.md (rdssp<mode>): Also enable for
2322 (incssp<mode>): Likewise.
2323 (nop_endbr): Likewise.
2324 * config/i386/i386.opt (mcet): Change help message to built-in
2328 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2329 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2330 enable CET built-in functions.
2332 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2334 * common/config/i386/i386-common.c
2335 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2336 OPTION_MASK_ISA_MOVDIRI_UNSET,
2337 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2338 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2339 * config.gcc (movdirintrin.h): New header.
2340 * config/i386/cpuid.h (bit_MOVDIRI,
2341 bit_MOVDIR64B): New bits.
2342 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2344 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2345 (VOID, PVOID, PCVOID)): New function types.
2346 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2347 __builtin_ia32_directstoreu_u64,
2348 __builtin_ia32_movdir64b): New builtins.
2349 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2350 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2352 (ix86_valid_target_attribute_inner_p): Ditto.
2353 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2354 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2355 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2356 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2357 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2358 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2359 (movdiri<mode>, movdir64b_<mode>): New.
2360 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2361 * config/i386/immintrin.h: Include movdirintrin.h.
2362 * config/i386/movdirintrin.h: New file.
2363 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2365 2018-04-19 Richard Biener <rguenther@suse.de>
2368 * cfg.c (clear_bb_flags): When loop state says we have
2369 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2371 2018-04-19 Richard Biener <rguenther@suse.de>
2373 PR tree-optimization/84737
2374 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2375 copying restrict info.
2376 (vect_setup_realignment): Use it.
2377 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2378 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2379 the first DR to all generated stores.
2380 (vectorizable_load): Likewise for loads.
2382 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2384 PR tree-optimization/85446
2385 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2386 the integral and pointer types to have the same precision.
2388 * doc/install.texi: Document --disable-cet being the default and
2391 2018-04-18 Martin Liska <mliska@suse.cz>
2393 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2396 2018-04-18 Martin Liska <mliska@suse.cz>
2399 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2402 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2403 arguments if they are comparable.
2405 2018-04-18 Martin Liska <mliska@suse.cz>
2408 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2411 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2414 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2417 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2418 ENDBR after calling __morestack.
2420 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2423 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2424 by using gcc_base_ver to generate a gcc_driver_version, and use
2425 it when generating GCC_DRIVER_NAME.
2426 * configure: Regenerate.
2428 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2431 * config.gcc: Obsolete powerpc*-*-*spe*.
2433 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2436 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2437 (stabstr_D): Change type of unum from unsigned int to
2438 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2441 2018-04-17 Jim Wilson <jimw@sifive.com>
2444 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2445 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2446 Set arg_pointer_offset after using pretend_args_size.
2448 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2450 PR rtl-optimization/85431
2451 * dse.c (record_store): Ignore zero width stores.
2454 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2455 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2456 __builtin_stack_restore rather than after it.
2457 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2458 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2459 argument instead of virtual_dynamic_stack_rtx.
2461 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2463 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2465 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2466 Add note to error message to explain internal mapping of overloaded
2467 built-in function name to non-overloaded built-in function name.
2468 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2471 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2474 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2475 where the inputs overlap with the output.
2477 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2479 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2480 (=v, v) alternative and explicit "memory" attribute.
2481 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2482 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2484 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2485 "sselog1" type instead of "sselog".
2486 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2487 "sselog". Remove explicit "memory" attribute.
2488 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2489 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2491 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2492 "sselog1" type instead of "sselog". Remove explicit "memory"
2494 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2495 use "sselog1" type instead of "sselog". Remove explicit "memory"
2497 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2498 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2500 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2501 "sselog1" type instead of "sselog". Remove explicit "memory"
2503 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2504 use "sselog1" type instead of "sselog". Remove explicit "memory"
2508 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2511 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2514 2018-04-17 Martin Jambor <mjambor@suse.cz>
2517 * ipa-cp.c (create_specialized_node): Call
2518 expand_all_artificial_thunks if necessary.
2520 2018-04-17 Martin Liska <mliska@suse.cz>
2523 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2524 in message, remote space in between '_G' and '('.
2526 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2529 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2530 avx512f_vmcmp<mode>3<round_saeonly_name>,
2531 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2532 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2533 avx512f_rndscale<mode><round_saeonly_name>,
2534 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2535 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2536 Use %<iptr>2 instead of %2 for -masm=intel.
2537 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2538 avx512f_vcvttss2usi<round_saeonly_name>,
2539 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2541 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2542 avx512f_vcvttsd2usi<round_saeonly_name>,
2543 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2544 Use %q1 instead of %1 for -masm=intel.
2545 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2546 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2547 of %3 for -masm=intel.
2548 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2550 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2552 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2554 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2555 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2557 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2559 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2560 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2561 %g1 and one with %0 and %1.
2562 (avx512er_vmrcp28<mode><round_saeonly_name>,
2563 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2565 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2566 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2567 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2568 of %0 and %{%4%} for -masm=intel.
2569 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2570 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2571 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2572 order of %0 and %{%5%}%{z%} for -masm=intel.
2574 2018-04-17 Jan Hubicka <jh@suse.cz>
2577 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2579 2018-04-17 Martin Liska <mliska@suse.cz>
2582 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2583 for target_clone error message. Make default implementation
2584 clone to be a local declaration.
2585 (separate_attrs): Add new argument and check for an empty
2587 (expand_target_clones): Handle it.
2588 (ipa_target_clone): Make redirection just for target_clones
2591 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2592 Tom de Vries <tom@codesourcery.com>
2595 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2596 tiled basic blocks without omp continue statements.
2598 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2601 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2602 vec_extract expression as having side effects to make sure it gets
2605 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2608 * config/i386/i386.c (get_builtin_code_for_version): Check
2611 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2614 * gcc/config.gcc: Support "skylake".
2615 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2617 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2618 (processor_target_table): Add "skylake".
2619 (ix86_option_override_internal): Add "skylake".
2620 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2621 PROCESSOR_CANNONLAKE.
2622 (get_builtin_code_for_version): Fix priority for
2623 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2624 PROCESSOR_SKYLAKE-AVX512.
2625 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2626 (processor_type): Add PROCESSOR_SKYLAKE.
2628 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2629 Jason Merrill <jason@redhat.com>
2632 * convert.c (convert_to_integer_1): Use direct recursion for
2633 enumeral types and types with a precision less than the number
2634 of bits in their mode.
2636 2018-04-16 Julia Koval <julia.koval@intel.com>
2639 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2640 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2642 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2645 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2646 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2647 and -mno-direct-move.
2649 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2652 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2653 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2654 Ensure negative shifts result in {0}.
2656 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2658 PR rtl-optimization/79916
2659 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2660 regs (if any) to define how to gnerate SD moves when LRA is in
2663 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2665 PR rtl-optimization/85393
2666 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2667 * except.c (expand_dw2_landing_pad_for_region): Make static.
2668 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2669 a label and unconditional jump to old_bb, rather than
2670 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2673 PR rtl-optimization/85376
2674 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2675 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2676 instead of a specific value.
2678 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2679 Bin Cheng <bin.cheng@arm.com>
2681 PR tree-optimization/82965
2682 PR tree-optimization/83991
2683 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2684 by_profile_only parameter.
2685 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2686 information if the loop was predicted to iterate too many times.
2687 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2689 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2692 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2695 2018-04-13 Martin Liska <mliska@suse.cz>
2696 Jakub Jelinek <jakub@redhat.com>
2699 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2700 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2701 * builtins.c (expand_builtin_memory_copy_args): Use
2702 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2703 handle dest_addr == pc_rtx.
2705 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2708 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2709 asked to not generate direct moves.
2710 (fix_trunc<mode>si2_stfiwx): Similar.
2711 (fix_trunc<mode>si2_internal): Similar.
2713 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2716 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2717 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2718 lookup if dest in some wider mode is known to be const0_rtx and
2719 if so, record permanent equivalence for it to be ZERO_EXTEND of
2720 the narrower mode destination.
2722 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2724 * lto-streamer-out.c (output_function): Revert 259346.
2725 * omp-expand.c (expand_oacc_for): Likewise.
2727 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2729 PR rtl-optimization/85354
2730 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2731 * sel-sched.c (sel_global_init): ... here.
2733 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2736 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2737 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2738 mode for PE-COFF targets.
2739 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2740 (i386_pe_asm_lto_end): Likewise.
2741 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2742 (TARGET_ASM_LTO_END): Likewise.
2743 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2744 (i386_pe_asm_lto_start): New function.
2745 (i386_pe_asm_lto_end): Likewise.
2747 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2748 Richard Biener <rguenther@suse.de>
2751 * lto-streamer-out.c (output_function): Fix CFG loop state before
2753 * omp-expand.c (expand_oacc_for): Handle calls to internal
2754 functions like regular functions.
2756 2018-04-12 Richard Biener <rguenther@suse.de>
2759 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2760 for the early LTO debug to properly generate references to it
2761 during DIE emission. Do not re-use that for the skeleton for
2763 (dwarf2out_early_finish): Likewise.
2765 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2768 * config/i386/sse.md
2769 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2770 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2771 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2772 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2773 and output is a reg, avoid creating invalid lowpart subreg, but
2774 instead split into a 512-bit move. Don't split if not AVX512VL,
2775 input is xmm16+ reg and output is a mem.
2776 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2777 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2778 xmm16+ reg and output is a mem.
2780 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2782 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2783 also for flag_dwarf2_cfi_asm.
2785 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2787 PR rtl-optimization/85342
2788 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2789 a bool scalar var inside of the loop instead. Don't try to update
2790 recog_data.operand after failed apply_change_group.
2792 2018-04-12 Tom de Vries <tom@codesourcery.com>
2795 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2796 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2797 array with flexible array member as array without given dimension.
2798 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2799 argument for undefined param to true.
2801 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2804 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2805 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2806 from PowerPC section.
2807 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2808 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2809 help text that is too long.
2810 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2811 help text that is too long.
2812 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2813 help text that is too long.
2815 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2817 * config/alpha/alpha.md (stack_probe_internal): Rename
2818 from "probe_stack". Update all callers.
2820 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2822 PR rtl-optimization/84566
2823 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2824 sched_macro_fuse_insns.
2826 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2829 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2830 (compute_block_dependences): ... from here.
2832 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2834 PR tree-optimization/85331
2835 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2836 from int to HOST_WIDE_INT.
2838 2018-04-11 Martin Jambor <mjambor@suse.cz>
2841 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2842 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2843 not the same as the source val.
2844 (cgraph_edge_brings_value_p): New parameter.
2845 (gather_edges_for_value): Pass destination value to
2846 cgraph_edge_brings_value_p.
2847 (perhaps_add_new_callers): Likewise.
2848 (get_info_about_necessary_edges): Likewise and exclude values brought
2849 only by self-recursive edges.
2850 (create_specialized_node): Redirect only clones of self-calling edges.
2851 (+self_recursive_pass_through_p): New function.
2852 (find_more_scalar_values_for_callers_subset): Use it.
2853 (find_aggregate_values_for_callers_subset): Likewise.
2854 (known_aggs_to_agg_replacement_list): Removed.
2855 (decide_whether_version_node): Re-calculate known constants for all
2856 remaining context clones.
2858 2018-04-11 Richard Biener <rguenther@suse.de>
2861 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2862 from early DWARF output.
2863 (dwarf2out_early_finish): Output line info unconditionally into
2864 early DWARF and add reference to it.
2866 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2869 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2870 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2871 other than V2DFmode using iptr mode attribute.
2872 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2874 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2876 PR rtl-optimization/84659
2877 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2879 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2882 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2884 (output_loc_list): Pass address of a dummy size variable even in the
2885 locview handling loop.
2886 (index_location_lists): Add comment on why skip_loc_list_entry can't
2889 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2892 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2895 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2898 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2899 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2900 and -mstring-compare-inline-limit.
2902 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2905 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2906 for stack clash protection in a register whenever we need it to be in
2909 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2911 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2912 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2914 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2917 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2919 (mlong-double-): Ditto.
2920 * config/rs6000/sysv4.opt (msdata=): Ditto.
2921 (mtls-size=): Ditto.
2923 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2925 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2926 erroneous entries for
2927 "vector int vec_ldl (int, long int *)", and
2928 "vector unsigned int vec_ldl (int, unsigned long int *)".
2929 Add comments and entries for
2930 "vector bool char vec_ldl (int, bool char *)",
2931 "vector bool short vec_ldl (int, bool short *)",
2932 "vector bool int vec_ldl (int, bool int *)",
2933 "vector bool long long vec_ldl (int, bool long long *)",
2934 "vector pixel vec_ldl (int, pixel *)",
2935 "vector long long vec_ldl (int, long long *)",
2936 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2937 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2938 type tree bool_long_long_type_node and correct definition of
2939 bool_V2DI_type_node to make reference to this new type tree.
2940 (rs6000_mangle_type): Replace erroneous reference to
2941 bool_long_type_node with bool_long_long_type_node.
2942 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2943 comments to emphasize sign distinctions for char and int types and
2944 replace RS6000_BTI_bool_long constant with
2945 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2946 use of RS6000_BTI_pixel.
2947 (bool_long_type_node): Remove this macro definition.
2948 (bool_long_long_type_node): New macro definition
2950 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2952 PR rtl-optimization/85300
2953 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2954 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2955 simplify_unary_operation fails.
2957 2018-04-10 Martin Liska <mliska@suse.cz>
2959 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2960 cgraph_edge and ipa_ref.
2962 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2966 * config/i386/sse.md
2967 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2968 computation of the VEC_MERGE selector from mask.
2969 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2970 Fix decoding of the VEC_MERGE selector into mask.
2972 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2974 PR tree-optimization/85286
2975 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2977 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2979 * final.c (final_1): Set insn_last_address as well as
2980 insn_current_address.
2982 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2985 * explow.c (emit_stack_probe): Call validize_mem on memory location
2986 before passing it to gen_probe_stack. Create address operand and
2987 legitimize it for the probe_stack_address case.
2989 2018-04-09 Jan Hubicka <jh@suse.cz>
2992 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2993 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2994 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2996 (free_lang_data): Rebuild type inheritance graph.
2998 2018-04-09 Martin Sebor <msebor@redhat.com>
3000 * invoke.texi (-finline-small-functions): Mention other optimization
3002 (-findirect-inlining, -fpartial-inlining): Same.
3003 (-finline-functions-called-once): Same.
3004 (-freorder-blocks-and-partition): Same.
3006 2018-04-09 Jan Hubicka <jh@suse.cz>
3009 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3010 jumps; choose last target that matches the criteria (i.e.
3011 no partition changes for non-crossing jumps).
3012 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3013 support for redirecting crossing jumps to non-crossing.
3015 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3017 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3018 also for naked functions.
3020 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3022 * config/arc/arc.md (add_shift): New pattern.
3023 (add_shift2): Likewise.
3024 (sub_shift): Likewise.
3025 (sub_shift_cmp0_noout): Likewise.
3026 (compare_si_ashiftsi): Likewise.
3027 (xbfu_cmp0_noout): New combine pattern.
3028 (xbfu_cmp0"): Likewise.
3029 (movsi_set_cc_insn): Place the predicable variant first.
3030 (commutative_binary_cmp0_noout): Remove clobber.
3031 (commutative_binary_cmp0): New pattern.
3032 (noncommutative_binary_cmp0): Likewise.
3033 (noncommutative_binary_cmp0_noout): Likewise.
3034 (noncommutative_binary_comparison_result_used): Removed.
3035 (rsub_cmp0): New pattern.
3036 (rsub_cmp0_noout): Likewise.
3037 (extzvsi): Changed, keep only meaningful variants.
3038 (SQH, SEZ): New iterators.
3039 (SQH_postfix): New mode attribute.
3040 (SEZ_prefix): New code attribute.
3041 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3042 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3043 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3045 (noncommutative_operator): Check the availability of barrel
3048 2018-04-09 Richard Biener <rguenther@suse.de>
3050 PR tree-optimization/85284
3051 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3052 Only use the niter constraining form of simple_iv when the exit
3055 2018-04-09 Tom de Vries <tom@codesourcery.com>
3058 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3059 (define_expand "*memory_barrier"): New define_expand.
3060 (define_insn "memory_barrier"): New insn.
3062 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3064 PR rtl-optimization/80463
3065 PR rtl-optimization/83972
3066 PR rtl-optimization/83480
3068 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3069 correct producer for the insn.
3070 (tidy_control_flow): Fixup seqnos in case of debug insns.
3072 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3074 PR rtl-optimization/83913
3076 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3077 different sched-times when merging exprs.
3079 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3081 PR rtl-optimization/83962
3083 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3084 tidy_fallthru_edge and tidy_control_flow.
3086 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3088 PR rtl-optimization/83530
3090 * sel-sched.c (force_next_insn): New global variable.
3091 (remove_insn_for_debug): When force_next_insn is true, also leave only
3092 next insn in the ready list.
3093 (sel_sched_region): When the region wasn't scheduled, make another pass
3094 over it with force_next_insn set to 1.
3096 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3098 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3100 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3101 for interrupt control.
3102 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3103 functions for interrupt control.
3104 * config/nds32/nds32-intrinsic.md: Likewise.
3105 * config/nds32/nds32_intrinsic.h: Likewise.
3106 * config/nds32/nds32.h (nds32_builtins): Likewise.
3108 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3110 * config/nds32/nds32.c (nds32_init_machine_status,
3111 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3112 strict_aligned_p field.
3113 (nds32_expand_to_rtl_hook): New function.
3114 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3115 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3117 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3118 Chung-Ju Wu <jasonwucj@gmail.com>
3120 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3121 * config/nds32/nds32-n7.md: New file.
3122 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3123 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3125 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3126 * config/nds32/nds32.md (pipeline_model): Add n7.
3127 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3128 * config/nds32/pipelines.md: Include n7 settings.
3130 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3131 Chung-Ju Wu <jasonwucj@gmail.com>
3133 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3134 * config/nds32/nds32-e8.md: New file.
3135 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3136 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3138 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3139 * config/nds32/nds32.md (pipeline_model): Add e8.
3140 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3141 * config/nds32/pipelines.md: Include e8 settings.
3143 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3144 Chung-Ju Wu <jasonwucj@gmail.com>
3146 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3147 * config/nds32/nds32-n8.md: New file.
3148 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3149 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3151 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3152 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3153 * config/nds32/nds32.md (pipeline_model): Add n8.
3154 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3155 * config/nds32/pipelines.md: Include n8 settings.
3157 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3158 Chung-Ju Wu <jasonwucj@gmail.com>
3160 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3161 * config/nds32/nds32-n9-2r1w.md: New file.
3162 * config/nds32/nds32-n9-3r2w.md: New file.
3163 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3164 nds32_register_ports): New or modify for cpu n9.
3165 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3167 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3168 * config/nds32/nds32-utils.c: New file.
3169 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3170 TARGET_MUL_SLOW): Define.
3171 * config/nds32/nds32.md (pipeline_model): New attribute.
3172 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3173 New options that support cpu n9.
3174 * config/nds32/pipelines.md: Include n9 settings.
3175 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3177 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3179 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3180 information if necessary.
3181 (output_cond_branch_compare_zero): Likewise.
3182 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3183 (nds32_target_alignment): Refine for alignment.
3184 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3185 (FUNCTION_BOUNDARY): Modify.
3186 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3188 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3190 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3192 * config/nds32/constants.md (unspec_volatile_element): Add values for
3193 TLB operation and data prefetch.
3194 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3195 functions for TLB operation and data prefetch.
3196 * config/nds32/nds32-intrinsic.md: Likewise.
3197 * config/nds32/nds32_intrinsic.h: Likewise.
3198 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3199 (nds32_print_operand): Likewise.
3200 * config/nds32/nds32.h (nds32_builtins): Likewise.
3202 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3203 Andrew Pinski <pinsika@gcc.gnu.org>
3206 * match.pd: Use constant_boolean_node of correct type instead of
3207 boolean_true_node or boolean_false_node for simplifying
3208 pointer comparisons to zero.
3210 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3212 PR tree-optimization/80021
3213 * tree.c (verify_type_variant): Make error call in verify_variant_match
3214 translatable and remove final full stop.
3216 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3218 * config/nds32/constants.md (unspec_volatile_element): Add
3219 UNSPEC_VOLATILE_EH_RETURN.
3220 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3221 nds32_output_stack_pop): Support dwarf exception handling process.
3222 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3223 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3224 exception handling process.
3225 (nds32_compute_stack_frame): Likewise.
3226 (nds32_return_addr_rtx): Likewise.
3227 (nds32_initial_elimination_offset): Likewise.
3228 (nds32_expand_prologue): Likewise.
3229 (nds32_expand_epilogue): Likewise.
3230 (nds32_dynamic_chain_address): New function.
3231 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3233 (DYNAMIC_CHAIN_ADDRESS): Define.
3234 (EH_RETURN_DATA_REGNO): Define.
3235 (EH_RETURN_STACKADJ_RTX): Define.
3236 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3237 patterns for dwarf exception handling.
3239 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3241 * config/nds32/nds32.h: Clean up obsolete macros.
3243 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3245 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3246 Add enum values for particular instructions.
3247 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3248 particular intrinsic functions.
3249 * config/nds32/nds32-intrinsic.md: Likewise.
3250 * config/nds32/nds32_intrinsic.h: Likewise.
3251 * config/nds32/nds32.h (nds32_builtins): Likewise.
3252 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3253 (btst, ave): New patterns for particular instructions.
3255 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3257 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3258 Add enum values for atomic load/store and memory sync.
3259 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3261 * config/nds32/nds32-intrinsic.md: Likewise.
3262 * config/nds32/nds32_intrinsic.h: Likewise.
3263 * config/nds32/nds32.h (nds32_builtins): Likewise.
3265 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3267 PR tree-optimization/85257
3268 * fold-const.c (native_encode_vector): If not all elts could fit
3269 and off is -1, return 0 rather than offset.
3270 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3271 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3272 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3273 adjust buffer in native_interpret_expr call.
3275 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3277 * config/nds32/constants.md (unspec_volatile_element): Add cache
3278 control enum values.
3279 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3280 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3281 * config/nds32/nds32.c (nds32_cctl_names): New.
3282 (nds32_print_operand): Handle cache control register names.
3283 * config/nds32/nds32.h (nds32_builtins): New enum values.
3284 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3286 * config/nds32/nds32.md (type): Add mmu.
3287 * config/nds32/pipelines.md (simple_insn): Add mmu.
3289 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3291 * config/nds32/nds32.md (type): Remove call.
3292 * config/nds32/pipelines.md (simple_insn): Likewise.
3294 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3296 * config/nds32/constants.md (unspec_volatile_element): Add
3297 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3298 UNSPEC_VOLATILE_FMFCFG.
3299 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3300 description for fmfcfg and fmfcsr.
3301 (bdesc_1arg): Add fmtcsr.
3302 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3303 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3304 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3305 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3306 unspec_fmfcfg): New patterns.
3307 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3308 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3309 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3310 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3311 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3312 __nds32__fmfcfg): Define.
3314 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3316 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3317 intrinsic register names.
3318 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3319 intrinsic register enum values and macros.
3321 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3323 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3324 for load/store addressing form.
3325 (nds32_print_operand_address): Likewise.
3327 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3330 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3331 based on LABEL_REF. Remove useless assertion.
3332 (pic_address_needs_scratch): Fix formatting.
3333 (sparc_legitimize_pic_address): Minor tweaks.
3334 (sparc_delegitimize_address): Adjust assertion accordingly.
3335 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3336 into symbolic_operand.
3337 (movsi_high_pic_label_ref): Likewise.
3338 (movsi_lo_sum_pic_label_ref): Likewise.
3339 (movdi_pic_label_ref): Likewise.
3340 (movdi_high_pic_label_ref): Likewise.
3341 (movdi_lo_sum_pic_label_ref): Likewise.
3343 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3345 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3346 custom LIB_SPEC setup.
3348 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3349 Kito Cheng <kito.cheng@gmail.com>
3351 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3352 * config/riscv/freebsd.h: New.
3354 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3356 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3357 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3360 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3361 Kito Cheng <kito.cheng@gmail.com>
3363 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3364 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3365 * config/nds32/nds32-protos.h (nds32_output_call,
3366 nds32_output_return): Declare.
3367 * config/nds32/nds32.md: Refine all the call and return patterns.
3369 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3372 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3373 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3375 PR rtl-optimization/84872
3376 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3377 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3380 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3382 * expr.c (copy_blkmode_to_reg): Revert 254862.
3383 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3385 2018-04-06 Richard Biener <rguenther@suse.de>
3388 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3389 after seeing a component reference with an adjacent field. Treat
3390 refs to arrays at struct end of external decls similar to
3391 refs to unconstrained commons.
3393 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3396 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3397 look through SAVE_EXPRs with non-side-effects argument. Adjust
3399 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3402 2018-04-06 Richard Biener <rguenther@suse.de>
3405 * alias.c (find_base_term): New wrapper around find_base_term
3406 unwinding CSELIB_VAL_PTR changes.
3407 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3410 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3412 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3414 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3415 constant definitions.
3416 ("nop"): lr 0,0 -> nopr r0
3417 ("nop_lr0", "nop_lr1"): New insn definitions.
3419 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3421 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3422 NDS32_V3PUSH_AVAILABLE_P macro.
3424 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3425 Chung-Ju Wu <jasonwucj@gmail.com>
3427 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3428 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3429 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3430 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3431 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3432 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3433 * config/nds32/constraints.md: New constraints and checking for hard
3434 float configuration.
3435 * config/nds32/iterators.md: New mode iterator and attribute for hard
3436 float configuration.
3437 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3439 * config/nds32/nds32-fpu.md: New file.
3440 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3441 deal with hard float code generation.
3442 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3444 (abi_type, float_reg_number): New enum type.
3445 * config/nds32/nds32-predicates.c: New predicates for hard float.
3446 * config/nds32/nds32-protos.h: Declare functions for hard float.
3447 * config/nds32/nds32.c: Implementation for hard float configuration.
3448 * config/nds32/nds32.h: Definitions for hard float configuration.
3449 * config/nds32/nds32.md: Include hard float machine description and
3450 modify patterns for hard float configuration.
3451 * config/nds32/nds32.opt: New options for hard float configuration.
3452 * config/nds32/predicates.md: New predicates for hard float
3455 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3457 * common/config/nds32/nds32-common.c
3458 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3460 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3463 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3464 CONSTRUCTOR_ELT (ctor, ...)->value.
3466 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3469 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3471 2018-04-05 Tom de Vries <tom@codesourcery.com>
3474 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3477 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3478 Kito Cheng <kito.cheng@gmail.com>
3480 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3481 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3482 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3483 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3485 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3486 Kito Cheng <kito.cheng@gmail.com>
3488 * config/nds32/constraints.md (Ufe): New memory constraint.
3489 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3490 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3491 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3493 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3494 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3496 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3498 * config/nds32/nds32.md: Use optimize_size in the condition for
3499 alu-shift instructions.
3501 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3503 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3505 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3507 * config/nds32/nds32.md (negsi2): Refine pattern.
3509 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3510 Chung-Ju Wu <jasonwucj@gmail.com>
3512 * config/nds32/iterators.md (shift_rotate): New code iterator.
3513 (shift): New code attribute.
3514 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3515 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3516 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3517 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3518 bit-wise operations.
3519 (andsi3, *andsi3): Ditto.
3520 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3521 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3522 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3523 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3524 nds32_ior_operand, nds32_xor_operand): New predicates.
3526 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3528 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3529 (addsi3, subsi3): ... this.
3531 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3533 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3535 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3537 * config/nds32/nds32.md: Adjust indention.
3539 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3541 * config/nds32/nds32.md (feature): New attribute.
3543 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3545 * config/nds32/nds32.md (subtype): New attribute.
3547 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3550 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3551 expansion to perform a bitwise AND of the argument followed by a
3552 boolean negation of the result.
3554 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3556 PR rtl-optimization/84878
3557 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3558 the basic block. Assert the use reference is not artificial and that
3559 it has an associated insn.
3561 2018-04-04 Michael Matz <matz@suse.de>
3563 * builtins.c (compute_objsize): Pass correct operand
3564 to array_at_struct_end_p.
3566 2018-04-04 Richard Biener <rguenther@suse.de>
3569 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3570 from contexts for DINFO_LEVEL_TERSE and below.
3572 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3574 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3575 resiter_operand condition.
3576 * config/nds32/nds32.md (*move<mode>): Ditto.
3578 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3579 Monk Chiang <sh.chiang04@gmail.com>
3581 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3583 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3585 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3587 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3588 Kito Cheng <kito.cheng@gmail.com>
3590 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3591 nds32_cond_code_str, output_cond_branch,
3592 output_cond_branch_compare_zero, nds32_expand_cbranch,
3593 nds32_expand_cstore, nds32_expand_movcc,
3594 nds32_output_cbranchsi4_equality_zero,
3595 nds32_output_cbranchsi4_equality_reg,
3596 nds32_output_cbranchsi4_equality_reg_or_const_int,
3597 nds32_output_cbranchsi4_greater_less_zero: New functions.
3598 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3599 nds32_expand_cstore, nds32_expand_movcc,
3600 nds32_output_cbranchsi4_equality_zero,
3601 nds32_output_cbranchsi4_equality_reg,
3602 nds32_output_cbranchsi4_equality_reg_or_const_int,
3603 nds32_output_cbranchsi4_greater_less_zero): Declare.
3604 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3605 nds32_rimm11s_operand): New predicates.
3606 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3607 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3610 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3612 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3613 * config/nds32/nds32.md: Ditto.
3614 * config/nds32/pipelines.md: Ditto.
3616 2018-04-04 Richard Biener <rguenther@suse.de>
3618 PR tree-optimization/85168
3619 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3620 propagating abnormals.
3622 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3624 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3626 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3627 Kito Cheng <kito.cheng@gmail.com>
3629 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3630 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3631 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3632 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3633 * config/nds32/nds32.md (sibcall_internal): New.
3634 (sibcall_register): Remove.
3635 (sibcall_immediate): Remove.
3636 (sibcall_value_internal): New.
3637 (sibcall_value_register): Remove.
3638 (sibcall_value_immediate): Remove.
3639 * config/nds32/predicates.md (nds32_general_register_operand): New.
3640 (nds32_call_address_operand): New.
3642 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3644 PR rtl-optimization/85167
3645 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3646 bb_defs if *split_p, instead preinitialize it to NULL.
3648 PR tree-optimization/85156
3649 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3650 evaluating the argument multiple times.
3652 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3654 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3656 (_mm_cvtpd_ps): Likewise.
3657 (_mm_cvttpd_epi32): Likewise.
3658 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3659 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3660 vector, pixel, and bool following altivec.h include.
3662 2018-04-03 Martin Sebor <msebor@redhat.com>
3664 * doc/extend.texi (Common Function Attributes): Clarify.
3665 (const attribute): Likewise.
3666 (pure attribute): Likewise.
3668 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3671 * config/i386/i386.c (ix86_expand_vector_set): Use
3672 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3674 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3676 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3677 instructions when changing rounding bits to preserve precision bits
3678 in the x87 control word.
3680 2018-04-03 Martin Liska <mliska@suse.cz>
3682 PR tree-optimization/82491
3683 * rtl.h (strip_offset_and_add): Replace += suboffset with
3684 poly_uint64 () + suboffset.
3686 2018-03-29 Martin Liska <mliska@suse.cz>
3687 Martin Jambor <mjambor@suse.cz>
3690 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3691 param_type is not an integral or pointer type.
3693 2018-04-03 Richard Biener <rguenther@suse.de>
3695 * sese.h (recompute_all_dominators): Remove.
3697 2018-04-02 Martin Sebor <msebor@redhat.com>
3699 * doc/invoke.texi (-Wrestrict): Fix typos.
3701 2018-04-02 Jim Wilson <jimw@sifive.com>
3703 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3704 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3705 (<optab>di3, <optab>si3_extend): Likewise.
3706 (<optab>si3_mask, <optab>si3_mask_1): New.
3707 (<optab>di3_mask, <optab>di3_mask_1): New.
3708 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3709 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3710 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3712 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3714 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3717 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3719 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3720 (nds32_canonicalize_comparison): New function.
3722 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3723 Kito Cheng <kito.cheng@gmail.com>
3724 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3726 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3727 * config/nds32/constants.md (unspec_volatile_element): Add
3728 UNSPEC_VOLATILE_RELAX_GROUP.
3729 * config/nds32/nds32-relax-opt.c: New file.
3730 * config/nds32/nds32-predicates.c
3731 (nds32_symbol_load_store_p): New function.
3732 * config/nds32/nds32-protos.h
3733 (nds32_symbol_load_store_p): Declare function.
3734 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3735 * config/nds32/nds32.c
3736 (nds32_register_pass): New function to register pass.
3737 (nds32_register_passes): New function to register passes.
3738 * config/nds32/nds32.md (relax_group): New pattern.
3739 * config/nds32/nds32.opt (mrelax-hint): New option.
3740 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3742 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3744 * config/nds32/t-nds32: Modify files dependency.
3746 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3748 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3749 (PROFILE_HOOK): Define its implementation.
3751 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3753 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3754 type and 32-bit size.
3756 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3759 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3760 (V_128_256): New mode iterator.
3761 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3762 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3763 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3765 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3766 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3768 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3771 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3772 NaN inputs correctly.
3774 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3777 * config/rs6000/vsx.md (??r): New mode attribute.
3778 (*vsx_mov<mode>_64bit): Use it.
3779 (*vsx_mov<mode>_32bit): Likewise.
3781 2018-03-30 Martin Sebor <msebor@redhat.com>
3783 PR tree-optimization/84818
3784 * builtins.c (check_access): Use warning_n.
3786 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3789 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3791 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3794 2018-03-30 Julia Koval <julia.koval@intel.com>
3797 * x86-tune.def (movx, partial_reg_dependency): Enable for
3800 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3803 * lra-constraints.c (process_alt_operands): Move setting
3804 this_alternative_matches below.
3806 2018-03-29 Martin Liska <mliska@suse.cz>
3809 * doc/invoke.texi: Document how LTO works with debug info.
3810 Describe auto-load support of binutils. Mention 'x86-64'
3811 as valid option value of -march option.
3813 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3815 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3818 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3819 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3820 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3823 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3826 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3827 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3828 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3829 for RS6000_BTM_POWERPC64.
3830 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3831 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3832 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3837 2018-03-28 Carl Love <cel@us.ibm.com>
3840 2017-09-27 Carl Love <cel@us.ibm.com>
3842 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3843 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3844 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3847 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3849 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3850 instead of __vector bool.
3851 (_mm_max_pu8): Likewise.
3852 (_mm_min_pi16): Likewise.
3854 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3857 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3858 (DIVWEUO): Likewise.
3860 (DIVDEUO): Likewise.
3861 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3862 DIVWEUO and DIVDEUO.
3863 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3864 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3865 (div_extend): Likewise.
3866 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3868 (__builtin_divweuo): Likewise.
3869 (__builtin_divdeo): Likewise.
3870 (__builtin_divdeuo): Likewise.
3872 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3875 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3876 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3878 PR tree-optimization/82004
3879 * gimple-match-head.c (optimize_pow_to_exp): New function.
3880 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3881 Don't fold to exp if optimize_pow_to_exp is false.
3883 2018-03-28 Martin Liska <mliska@suse.cz>
3886 * calls.c (initialize_argument_information): Fix trailing space.
3887 * common.opt: Fix typo and provide better explanation for
3888 -fsanitize-coverage option.
3889 * config/i386/i386.opt: Fix typo.
3891 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3892 Martin Liska <mliska@suse.cz>
3895 * gimplify.c (asan_poison_variable): Don't do the check for
3896 gimplify_omp_ctxp here.
3897 (gimplify_decl_expr): Do it here.
3898 (gimplify_target_expr): Likewise.
3900 2018-03-28 Martin Liska <mliska@suse.cz>
3903 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3904 chkp_type_bounds_count if MPX is not enabled.
3906 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3908 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3910 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3913 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3914 function to create the function decl for complex long double
3915 multiply and divide for -mabi=ieeelongdouble.
3916 (init_float128_ieee): Call it.
3918 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3921 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3922 -fcf-protection=branch -mibt.
3923 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3925 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3928 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3930 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3933 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3934 extern array declarations.
3936 2018-03-27 Richard Biener <rguenther@suse.de>
3939 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3940 explicit single_use checks.
3942 2018-03-27 Richard Biener <rguenther@suse.de>
3944 PR tree-optimization/85082
3945 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3948 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3950 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3951 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3952 Turn on fasynchronous-unwind-tables and funwind-tables.
3954 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3957 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3958 (*bmi_blsr_<mode>_ccz): Ditto.
3960 2018-03-26 Tom de Vries <tom@codesourcery.com>
3962 PR tree-optimization/85063
3963 * omp-general.c (offloading_function_p): New function. Factor out
3965 * omp-offload.c (pass_omp_target_link::gate): ... here.
3966 * omp-general.h (offloading_function_p): Declare.
3967 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3968 with attribute omp declare target for offloading functions.
3970 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3972 PR tree-optimization/84005
3973 * tree-data-ref.h (get_base_for_alignment): Declare.
3974 * tree-data-ref.c (get_base_for_alignment_1): New function.
3975 (get_base_for_alignment): Likewise.
3976 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3977 get_base_for_alignment to find a suitable base object, instead
3978 of always using drb->base_address.
3980 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3983 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3984 known size by default.
3986 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3989 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3990 and non BLKmode operands.
3992 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3995 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3996 Clean up attributes.
3998 2018-03-23 Richard Biener <rguenther@suse.de>
4001 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4002 we are going to emit early debug for LTO.
4004 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4007 * function.c (match_asm_constraints_1): Don't optimize if input
4008 doesn't satisfy general_operand predicate for output's mode.
4011 * alias.c (write_dependence_p): Don't require for x_canonicalized
4012 non-VOIDmode if x has VOIDmode.
4015 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4016 just don't try to optimize it rather than assert it never happens.
4018 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4020 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4021 macro expansions for definition of ST_INTERNAL_<mode> and
4022 LD_INTERNAL_<mode> builtins.
4023 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4025 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4027 (altivec_expand_st_builtin): Likewise.
4028 (altivec_expand_builtin): Remove calls to deleted functions.
4029 (rs6000_address_for_altivec): Delete this function.
4030 * config/rs6000/vector.md: Remove expands for
4031 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4033 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4036 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4037 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4038 re-computing once computed.
4039 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4040 (arm_init_machine_status): Initialize
4041 machine->static_chain_stack_bytes.
4043 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4046 * doc/extend.texi: Add four new prototypes for vec_ld.
4047 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4048 definitions for more logical presentation.
4049 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4050 entries for V1TI variants of __builtin_altivec_ld builtin.
4051 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4052 handling of V1TI variant of LVX icode pattern.
4053 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4054 (rs6000_gimple_fold_builtin): Likewise.
4055 (altivec_init_builtins): Add code to define
4056 __builtin_altivec_lvx_v1ti function.
4058 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4061 * function.c (match_asm_constraints_1): Don't do the optimization
4062 if input isn't a REG, SUBREG, MEM or constant.
4064 2018-03-22 Tom de Vries <tom@codesourcery.com>
4066 PR tree-optimization/84956
4067 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4068 bb_has_abnormal_pred.
4070 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4073 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4074 DECL_INITIAL (decl) to decl at the end.
4075 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4078 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4080 * doc/extend.texi (__builtin_tgmath): Document when complex
4081 integer types are treated as _Complex _Float64.
4083 2018-03-21 Tom de Vries <tom@codesourcery.com>
4085 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4087 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4089 PR tree-optimization/84960
4090 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4091 if it is ENTRY block, move them into single succ of ENTRY in that case.
4093 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4095 PR tree-optimization/84811
4096 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4097 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4098 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4099 (known_subrange_p): Update accordingly. Cast each value involved
4100 in the size comparison, rather than casting the result of the
4103 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4105 PR tree-optimization/84982
4106 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4107 by flipping the least significant bit rather than all bits from
4108 bitpos to bitpos + bitsize - 1.
4110 2018-03-21 Nathan Sidwell <nathan@acm.org>
4112 * doc/extend.texi (Deprecated Features): Remove mention of
4113 long-deleted deprecations.
4115 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4118 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4119 * configure: Regenerate.
4121 2018-03-21 Tom de Vries <tom@codesourcery.com>
4123 PR tree-optimization/83126
4124 * tree-parloops.c (num_phis): New function.
4125 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4127 2018-03-21 Nathan Sidwell <nathan@acm.org>
4129 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4130 mention anon-struct/union members and trailing attributes.
4132 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4134 PR tree-optimization/84969
4135 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4136 builtin memset partitions if they set different rhs values.
4138 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4140 PR rtl-optimization/84989
4141 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4142 VEC_DUPLICATE with scalar result mode.
4144 2018-03-21 Martin Liska <mliska@suse.cz>
4147 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4148 not intended return statement.
4150 2018-03-21 Martin Liska <mliska@suse.cz>
4153 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4154 (chkp_find_bound_slots_1): Limit number of iterations.
4156 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4159 * Minor grammar fixes for x86 options.
4161 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4164 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4165 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4167 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4170 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4171 (altivec_lvx_<mode>_1op): Likewise.
4172 (altivec_stvx_<mode>_2op): Likewise.
4173 (altivec_stvx_<mode>_1op): Likewise.
4174 (altivec_lvx_<VM2:mode>): New define_expand.
4175 (altivec_stvx_<VM2:mode>): Likewise.
4176 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4177 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4178 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4179 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4180 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4181 (rs6000_gen_lvx): Likewise.
4182 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4183 (altivec_expand_stv_builtin): Likewise.
4184 (altivec_expand_builtin): Likewise.
4185 * config/rs6000/vector.md: Likewise.
4187 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4190 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4193 2018-03-20 Richard Biener <rguenther@suse.de>
4196 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4197 sign-conversions as zero, fall back to standard scalar_stmt
4200 2018-03-20 Martin Liska <mliska@suse.cz>
4203 * predict.c (rebuild_frequencies): Handle case when we have
4204 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4206 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4209 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4210 flag_section_anchors.
4211 * varasm.c (use_blocks_for_decl_p): Remove hack for
4212 dw2_force_const_mem.
4215 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4217 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4218 be created, use lowpart_subreg of operands[0] rather than operands[0]
4220 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4221 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4222 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4223 and n constraint instead of aarch64_shift_imm_di and Usd.
4224 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4225 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4227 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4230 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4231 to favor GPR over NEON registers.
4232 (<shift>di3_neon): Likewise.
4234 2018-03-20 Tom de Vries <tom@codesourcery.com>
4237 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4238 (nvptx_process_pars): Emit bar.sync asap and alap.
4240 2018-03-20 Tom de Vries <tom@codesourcery.com>
4243 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4244 seen_label if seen_label is already set.
4246 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4249 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4250 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4251 Use 1U instead of 1. Formatting fixes.
4254 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4255 instead of TREE_TYPE (s1) for the return value.
4257 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4259 PR tree-optimization/84946
4260 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4261 bitsize + bitsize in poly_uint64 rather than poly_int64.
4264 * dwarf2asm.c: Include fold-const.c.
4265 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4266 of decl rather than decl itself.
4268 PR rtl-optimization/84643
4269 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4271 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4274 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4275 calling assemble_variable.
4277 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4280 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4281 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4283 2018-03-19 Jim Wilson <jimw@sifive.com>
4286 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4287 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4288 (riscv_first_stack_step): Likewise.
4289 (riscv_option_override): Use STACK_BOUNDARY instead of
4291 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4293 (BIGGEST_ALIGNMENT): Set to 128.
4294 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4295 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4298 2018-03-19 Richard Biener <rguenther@suse.de>
4300 PR tree-optimization/84933
4301 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4302 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4304 2018-03-19 Richard Biener <rguenther@suse.de>
4306 PR tree-optimization/84859
4307 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4308 (cond_if_else_store_replacement): Perform sinking operation on
4309 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4310 Generalize what a BB with a single eligible store is.
4312 2018-03-19 Richard Biener <rguenther@suse.de>
4314 PR tree-optimization/84929
4315 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4316 chrec_is_positive against non-chrec arg.
4318 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4321 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4323 2018-03-18 Martin Liska <mliska@suse.cz>
4325 PR rtl-optimization/84635
4326 * regrename.c (build_def_use): Use matches_mode only when
4329 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4331 PR tree-optimization/84913
4332 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4333 vectorize chains of COND_EXPRs.
4335 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4337 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4339 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4341 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4343 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4345 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4347 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4348 Kito Cheng <kito.cheng@gmail.com>
4350 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4351 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4352 (nds32_adjust_reg_alloc_order): New function.
4353 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4355 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4357 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4358 nds32_print_operand, nds32_print_operand_address): Use
4359 HOST_WIDE_INT_PRINT_DEC instead.
4361 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4363 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4365 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4368 * config/i386/i386.c (initial_ix86_tune_features,
4369 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4371 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4372 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4373 rather than 1u << ix86_tune. Formatting fix.
4374 (ix86_option_override_internal): Change ix86_arch_mask from
4375 unsigned int to unsigned HOST_WIDE_INT, initialize to
4376 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4377 (ix86_function_specific_restore): Likewise.
4379 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4382 * postreload.c (reload_combine_recognize_pattern): Perform
4383 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4384 truncate_int_for_mode the result for the destination's mode.
4387 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4388 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4390 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4393 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4396 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4397 regno_allocno_class_array and sorted_pseudos.
4398 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4399 insns where regno is used.
4401 2018-03-16 Martin Liska <mliska@suse.cz>
4404 * multiple_target.c (create_dispatcher_calls): Redirect
4405 reference in the symbol table.
4407 2018-03-16 Martin Liska <mliska@suse.cz>
4410 * multiple_target.c (create_dispatcher_calls): Redirect also
4413 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4417 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4418 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4419 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4421 2018-03-16 Julia Koval <julia.koval@intel.com>
4423 * doc/invoke.texi (Skylake Server): Add CLWB.
4424 Cannonlake): Remove CLWB.
4426 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4428 PR tree-optimization/84841
4429 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4431 (FLOAT_ONE_CONST_TYPE): Define.
4432 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4433 (sort_by_operand_rank): Put entries with higher constant_type last
4434 rather than first to match comments.
4436 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4438 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4441 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4444 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4445 check and use address of target always.
4447 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4450 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4451 (indirect_thunk_bnd_needed): Likewise.
4452 (indirect_thunks_used): Likewise.
4453 (indirect_thunks_bnd_used): Likewise.
4454 (indirect_return_needed): New.
4455 (indirect_return_bnd_needed): Likewise.
4456 (output_indirect_thunk_function): Add a bool argument for
4458 (output_indirect_thunk_function): Don't generate alias for
4459 function return thunk.
4460 (ix86_code_end): Call output_indirect_thunk_function to generate
4461 function return thunks.
4462 (ix86_output_function_return): Set indirect_return_bnd_needed
4463 and indirect_return_needed instead of indirect_thunk_bnd_needed
4464 and indirect_thunk_needed.
4466 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4468 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4469 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4470 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4472 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4473 Paul Hua <paul.hua.gm@gmail.com>
4476 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4478 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4480 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4481 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4484 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4487 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4488 instead of GET_MODE_SIZE when comparing Units.
4490 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4493 * varasm.c (hash_section): Return an unchangeble hash value
4494 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4495 Return !aarch64_can_use_per_function_literal_pools_p ().
4497 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4500 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4501 rather than address of cmode as last argument to prepare_cmp_insn.
4503 2018-03-15 Julia Koval <julia.koval@intel.com>
4505 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4506 F_AVX512VNNI, F_AVX512BITALG): New.
4508 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4511 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4512 insn for floating-point loads and stores.
4514 2018-03-14 Carl Love <cel@us.ibm.com>
4516 * config/rs6000/rs6000-c.c: Add macro definitions for
4517 ALTIVEC_BUILTIN_VEC_PERMXOR.
4518 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4519 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4520 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4521 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4523 * config/doc/extend.texi: Add prototypes for vec_permxor.
4525 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4528 * diagnostic-show-locus.c (class layout_point): Convert m_line
4529 from int to linenum_type.
4530 (line_span::comparator): Use linenum "compare" function when
4531 comparing line numbers.
4532 (test_line_span): New function.
4533 (layout_range::contains_point): Convert param "row" from int to
4535 (layout_range::intersects_line_p): Likewise.
4536 (layout::will_show_line_p): Likewise.
4537 (layout::print_source_line): Likewise.
4538 (layout::should_print_annotation_line_p): Likewise.
4539 (layout::print_annotation_line): Likewise.
4540 (layout::print_leading_fixits): Likewise.
4541 (layout::annotation_line_showed_range_p): Likewise.
4542 (struct line_corrections): Likewise for field m_row.
4543 (line_corrections::line_corrections): Likewise for param "row".
4544 (layout::print_trailing_fixits): Likewise.
4545 (layout::get_state_at_point): Likewise.
4546 (layout::get_x_bound_for_row): Likewise.
4547 (layout::print_line): Likewise.
4548 (diagnostic_show_locus): Likewise for locals "last_line" and
4550 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4551 * input.c (selftest::test_linenum_comparisons): New function.
4552 (selftest::input_c_tests): Call it.
4553 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4554 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4555 * selftest.h (ASSERT_GT): New macro.
4556 (ASSERT_GT_AT): New macro.
4557 (ASSERT_LT): New macro.
4558 (ASSERT_LT_AT): New macro.
4560 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4562 PR rtl-optimization/84780
4563 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4565 2018-03-14 Martin Liska <mliska@suse.cz>
4567 * tree.c (record_node_allocation_statistics): Use
4568 get_stats_node_kind.
4569 (get_stats_node_kind): New function extracted from
4570 record_node_allocation_statistics.
4571 (free_node): Use get_stats_node_kind.
4573 2018-03-14 Richard Biener <rguenther@suse.de>
4575 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4576 that the value-set of ANTIC_IN doesn't grow.
4579 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4581 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4582 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4584 2018-03-14 Julia Koval <julia.koval@intel.com>
4586 * config.gcc (icelake-client, icelake-server): New.
4588 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4589 (initial_ix86_arch_features): Ditto.
4590 (PTA_SKYLAKE): Add SGX.
4591 (PTA_ICELAKE): Remove.
4592 (PTA_ICELAKE_CLIENT): New.
4593 (PTA_ICELAKE_SERVER): New.
4594 (ix86_option_override_internal): Split up icelake on icelake client and
4596 (get_builtin_code_for_version): Ditto.
4597 (fold_builtin_cpu): Ditto.
4598 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4599 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4600 * config/i386/i386.h (processor_type): Ditto.
4601 * doc/invoke.texi: Ditto.
4603 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4606 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4607 INTEGER_CST offset, add it together with bitpos / 8 and
4608 sign extend based on POINTER_SIZE.
4612 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4615 * config/i386/constraints.md (Yc): New register constraint.
4616 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4617 Use Yc constraint for alternative 2 of operand 0. Remove
4618 preferred_for_speed attribute.
4620 2018-03-14 Richard Biener <rguenther@suse.de>
4622 PR tree-optimization/84830
4623 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4624 with the old one to avoid oscillations.
4626 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4629 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4631 (assign_by_spills): Return a flag of reload assignment failure.
4632 Do not process the reload assignment failures. Do not spill other
4633 reload pseudos if they has the same reg class. Update n if
4635 (lra_assign): Add a return arg. Set up from the result of
4636 assign_by_spills call.
4637 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4638 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4639 usage_insns if it is not NULL.
4640 (spill_hard_reg_in_range): New function.
4641 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4642 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4643 function prototypes.
4644 (lra_assign): Change prototype.
4645 * lra.c (lra): Add code to deal with fails by splitting hard reg
4648 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4650 * config/riscv/riscv.opt (mrelax): New option.
4651 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4652 "norelax" when riscv_mrelax is disabled.
4653 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4655 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4658 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4659 reassociation for int modes.
4661 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4663 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4664 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4666 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4667 * config/aarch64/aarch64-sve.md
4668 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4669 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4670 (*extend<mode><Vwide>2): Rename to...
4671 (aarch64_sve_extend<mode><Vwide>2): ...this.
4672 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4673 renaming the old pattern to...
4674 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4676 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4677 define_expand, renaming the old pattern to...
4678 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4679 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4680 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4681 account when deciding which SVE instruction the optab should use.
4682 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4684 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4686 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4687 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4688 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4689 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4690 (tlsdesc_small_<mode>): Turn a define_expand and use
4691 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4692 (tlsdesc_small_advsimd_<mode>): ...this.
4693 (tlsdesc_small_sve_<mode>): New pattern.
4695 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4697 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4698 (UNSPEC_UMUL_HIGHPART): New constants.
4699 (MUL_HIGHPART): New int iteraor.
4700 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4701 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4703 (*<su>mul<mode>3_highpart): New define_insn.
4705 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4708 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4711 2018-03-13 Martin Liska <mliska@suse.cz>
4714 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4716 (sem_item_optimizer::~sem_item_optimizer): Release it.
4717 (sem_item_optimizer::merge_classes): Register variable aliases.
4718 (sem_item_optimizer::fixup_pt_set): New function.
4719 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4720 * ipa-icf.h: Declare new variables and functions.
4722 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4725 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4726 integer_pow2p@2 and test integer_pow2p in condition.
4727 (A < 0 ? C : 0): Similarly for @1.
4730 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4731 characters starting at p contain '\0' character, don't look beyond
4735 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4736 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4739 * reg-stack.c (change_stack): Change update_end var from int to
4740 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4741 also call set_block_for_insn on the newly added insns and rescan.
4744 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4745 on the last operand.
4748 * tree.c (stabilize_reference_1): Return save_expr (e) for
4749 STATEMENT_LIST even if it doesn't have side-effects.
4751 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4753 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4755 2018-03-12 Renlin Li <renlin.li@arm.com>
4757 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4758 aarch64_output_scalar_simd_mov_immediate.
4760 2018-03-12 Martin Sebor <msebor@redhat.com>
4762 PR tree-optimization/83456
4763 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4764 for perfectly overlapping calls to memcpy.
4765 (gimple_fold_builtin_memory_chk): Same.
4766 (gimple_fold_builtin_strcpy): Handle no-warning.
4767 (gimple_fold_builtin_stxcpy_chk): Same.
4768 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4770 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4772 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4773 parameter. Use it for SFmode.
4774 (rs6000_function_arg_advance_1): Adjust.
4775 (rs6000_function_arg): Adjust.
4776 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4778 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4780 PR rtl-optimization/84169
4781 PR rtl-optimization/84780
4782 * combine.c (can_combine_p): Check for a 2-insn combination whether
4783 the destination register is used between the two insns, too.
4785 2018-03-12 Richard Biener <rguenther@suse.de>
4787 PR tree-optimization/84803
4788 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4789 for refs DR analysis didn't process.
4791 2018-03-12 Richard Biener <rguenther@suse.de>
4793 PR tree-optimization/84777
4794 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4795 force-vectorize loops ignore whether we are optimizing for size.
4797 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4799 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4800 (TARGET_MD_ASM_ADJUST): Define.
4802 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4803 Kito Cheng <kito.cheng@gmail.com>
4804 Chung-Ju Wu <jasonwucj@gmail.com>
4806 * config/nds32/nds32.c (nds32_compute_stack_frame,
4807 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4808 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4809 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4810 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4811 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4812 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4813 * config/nds32/nds32.md (prologue, epilogue): Use macro
4814 NDS32_V3PUSH_AVAILABLE_P to do checking.
4816 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4819 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4820 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4821 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4822 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4823 addition of most attributes on !orig_type_die or the attribute not
4824 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4826 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4827 Chung-Ju Wu <jasonwucj@gmail.com>
4829 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4831 * config/nds32/nds32.opt (mvh): New option.
4833 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4834 Chung-Ju Wu <jasonwucj@gmail.com>
4836 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4838 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4839 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4842 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4843 Chung-Ju Wu <jasonwucj@gmail.com>
4845 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4847 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4848 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4850 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4851 Kito Cheng <kito.cheng@gmail.com>
4852 Chung-Ju Wu <jasonwucj@gmail.com>
4854 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4855 UNSPEC_FFMISM and UNSPEC_FLMISM.
4856 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4857 for ffb, ffmism and flmism.
4858 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4859 (unspec_ffmism): Ditto.
4860 (unspec_flmism): Ditto.
4861 (nds32_expand_builtin_impl): Check if string extension is available.
4862 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4863 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4865 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4868 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4871 * lra-assigns.c (assign_by_spills): Return a flag of reload
4872 assignment failure. Do not process the reload assignment
4873 failures. Do not spill other reload pseudos if they has the same
4875 (lra_assign): Add a return arg. Set up from the result of
4876 assign_by_spills call.
4877 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4878 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4879 usage_insns if it is not NULL.
4880 (spill_hard_reg_in_range): New function.
4881 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4882 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4883 function prototypes.
4884 (lra_assign): Change prototype.
4885 * lra.c (lra): Add code to deal with fails by splitting hard reg
4888 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4891 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4893 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4896 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4897 (dw_val_node): Add val_symbolic_view.
4898 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4899 (symview_upper_bound): New.
4900 (new_line_info_table): Initialize symviews_since_reset.
4901 (dwarf2out_source_line): Count symviews_since_reset and set
4902 symview_upper_bound.
4903 (dw_val_equal_p): Handle symview.
4904 (add_AT_symview): New.
4905 (print_dw_val): Handle symview.
4906 (attr_checksum, attr_checksum_ordered): Likewise.
4907 (same_dw_val_p, size_of_die): Likewise.
4908 (value_format, output_die): Likewise.
4909 (add_high_low_attributes): Use add_AT_symview for entry_view.
4910 (dwarf2out_finish): Reset symview_upper_bound, clear
4913 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4916 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4917 Add strict argument and use it.
4918 (rs6000_split_multireg_move): Update for new strict argument.
4919 (mem_operand_gpr): Disallow all non-offsettable addresses.
4920 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4922 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4925 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4926 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4927 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4930 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4931 decl, use remap_type if we want to use the type.
4933 2018-03-09 Martin Sebor <msebor@redhat.com>
4935 PR tree-optimization/84526
4936 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4938 (builtin_access::generic_overlap): Be prepared to handle non-array
4941 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4943 PR rtl-optimization/84682
4944 * lra-constraints.c (process_address_1): Check is_address flag
4945 for address constraints.
4946 (process_alt_operands): Likewise.
4947 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4948 preprocess_constraints.
4949 * recog.h (preprocess_constraints): Add oploc parameter.
4951 * recog.c (preprocess_constraints): Test address_operand for
4952 CT_ADDRESS constraints.
4954 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4957 * lra-assigns.c (assign_by_spills): Return a flag of reload
4958 assignment failure. Do not process the reload assignment
4959 failures. Do not spill other reload pseudos if they has the same
4961 (lra_assign): Add a return arg. Set up from the result of
4962 assign_by_spills call.
4963 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4964 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4965 usage_insns if it is not NULL.
4966 (spill_hard_reg_in_range): New function.
4967 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4968 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4969 function prototypes.
4970 (lra_assign): Change prototype.
4971 * lra.c (lra): Add code to deal with fails by splitting hard reg
4974 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4977 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4978 Accept complain bool parameter. Only emit errors if it is true.
4979 (arm_parse_cpu_option_name): Likewise.
4980 (arm_target_thumb_only): Adjust callers of the above.
4981 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4982 prototype to take a default true bool parameter.
4983 (arm_parse_arch_option_name): Likewise.
4985 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4986 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4990 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4991 * configure: Regenerate.
4992 * configure.ac ("linker --version-script option"): New.
4993 ("linker soname option"): New.
4995 2018-03-09 Richard Biener <rguenther@suse.de>
4997 PR tree-optimization/84775
4998 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4999 immediate uses of predicate stmts and mark them modified.
5002 PR tree-optimization/84178
5003 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5005 (version_loop_for_if_conversion): Delay update_ssa call.
5006 (tree_if_conversion): Delay update_ssa until after predicate
5009 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5012 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5013 when the function accesses prior frames.
5015 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5018 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5019 gen_llsym, otherwise call maybe_gen_llsym.
5022 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5023 has ',' character inside of it.
5025 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5028 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5029 as clobbering CC_REGNUM.
5031 2018-03-08 Richard Biener <rguenther@suse.de>
5034 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5035 (follow_copies_to_constant): Do not follow SSA names registered
5038 2018-03-08 Richard Biener <rguenther@suse.de>
5040 PR tree-optimization/84178
5041 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5043 (version_loop_for_if_conversion): Delay update_ssa call.
5044 (tree_if_conversion): Delay update_ssa until after predicate
5047 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5049 PR tree-optimization/84178
5050 * tree-if-conv.c (release_bb_predicate): Remove the
5051 the assertion that the stmts have NULL use_ops.
5052 Discard the statements, asserting that they haven't
5053 yet been added to a BB.
5055 2018-03-08 Richard Biener <rguenther@suse.de>
5057 PR tree-optimization/84746
5058 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5059 (phi_translate): Pass in destination ANTIC_OUT set.
5060 (phi_translate_1): Likewise. For a simplified result lookup
5061 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5062 (phi_translate_set): Adjust.
5063 (do_pre_regular_insertion): Likewise.
5064 (do_pre_partial_partial_insertion): Likewise.
5066 2018-03-08 Martin Liska <mliska@suse.cz>
5068 PR gcov-profile/84735
5069 * doc/gcov.texi: Document usage of profile files.
5070 * gcov-io.h: Document changes in the format.
5072 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5076 * dwarf2out.c (struct dw_line_info_table): Update comments for
5078 (FORCE_RESET_NEXT_VIEW): New.
5079 (FORCE_RESETTING_VIEW_P): New.
5080 (RESETTING_VIEW_P): Check for -1 too.
5081 (ZERO_VIEW_P): Likewise.
5082 (new_line_info_table): Force-reset next view.
5083 (dwarf2out_begin_function): Likewise.
5084 (dwarf2out_source_line): Simplify zero_view_p initialization.
5085 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5086 view directly. Omit view when omitting .loc at line 0.
5088 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5090 PR tree-optimization/84740
5091 * tree-switch-conversion.c (process_switch): Call build_constructors
5092 only if info.phi_count is non-zero.
5094 PR tree-optimization/84739
5095 * tree-tailcall.c (find_tail_calls): Check call arguments against
5096 DECL_ARGUMENTS (current_function_decl) rather than
5097 DECL_ARGUMENTS (func) when checking for tail recursion.
5099 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5101 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5102 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5103 Volker Reichelt's entry and add entries for people that perform
5104 GCC fuzzy testing and report numerous bugs.
5106 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5109 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5110 readonly data in sdata, if that is disabled.
5111 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5112 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5113 -mreadonly-in-sdata option.
5115 2018-03-07 Martin Sebor <msebor@redhat.com>
5117 PR tree-optimization/84468
5118 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5119 basic block when looking for nul assignment.
5121 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5124 * except.h (output_function_exception_table): Adjust prototype.
5125 * except.c (output_function_exception_table): Remove FNNAME parameter
5126 and add SECTION parameter. Ouput one part of the table at a time.
5127 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5128 the first part of the exception table and emit unwind directives.
5129 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5130 (i386_pe_seh_cold_init): Likewise.
5131 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5132 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5133 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5134 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5135 * config/i386/winnt.c: Include except.h.
5136 (struct seh_frame_state): Add reg_offset, after_prologue and
5137 in_cold_section fields.
5138 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5139 (i386_pe_seh_cold_init): New function.
5140 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5141 to seh->in_cold_section.
5142 (seh_emit_push): Record the offset of the push.
5143 (seh_emit_save): Record the offet of the save.
5144 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5145 Test seh->after_prologue to disregard the epilogue.
5146 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5147 (i386_pe_end_cold_function): New function.
5149 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5152 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5153 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5156 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5158 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5159 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5162 * multiple_target.c: Include tree-inline.h and intl.h.
5163 (expand_target_clones): Diagnose and fail if node->definition and
5164 !tree_versionable_function_p (node->decl).
5166 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5168 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5170 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5171 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5172 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5174 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5177 * combine.c (try_combine): Use reg_or_subregno instead of handling
5178 just paradoxical SUBREGs and REGs.
5180 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5182 * config/arc/arc.c (arc_finalize_pic): Remove function.
5183 (arc_must_save_register): We use single base PIC register, remove
5184 checks to save/restore the PIC register.
5185 (arc_expand_prologue): Likewise.
5186 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5188 (arc_verify_short): Likewise.
5189 (arc_attr_type): Likewise.
5190 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5191 (walk_stores): Likewise.
5192 (arc_address_cost): Make it static.
5193 (arc_verify_short): Likewise.
5194 (branch_dest): Likewise.
5195 (arc_attr_type): Likewise.
5196 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5197 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5198 (arc_final_prescan_insn): Remove inserting the nops due to
5199 hardware hazards. It is done in reorg step.
5200 (insn_length_variant_t): Remove.
5201 (insn_length_parameters_t): Likewise.
5202 (arc_insn_length_parameters): Likewise.
5203 (arc_get_insn_variants): Likewise.
5204 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5206 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5209 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5212 PR tree-optimization/84687
5213 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5215 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5217 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5219 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5220 Rename to ppc_speculation_barrier.
5221 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5222 __builtin_ppc_speculation_barrier.
5224 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5227 * combine.c (combine_simplify_rtx): Don't try to simplify if
5228 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5231 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5233 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5234 to 32 bytes when compiling for POWER9.
5236 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5239 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5240 regparm >= 3 with no arg reg available also for calls with
5241 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5244 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5246 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5248 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5251 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5253 2018-03-05 Richard Biener <rguenther@suse.de>
5255 PR tree-optimization/84486
5256 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5257 When inserting a __builtin_assume_aligned call set the LHS
5258 SSA name alignment info accordingly.
5260 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5262 PR tree-optimization/84114
5263 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5264 Avoid reassociation of FLOAT_MODE addition.
5266 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5268 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5269 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5270 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5271 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5272 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5273 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5274 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5276 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5277 __builtin_ia32_wbinvd): New builtins.
5278 (SPECIAL_ARGS2): New.
5279 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5280 (SPECIAL_ARGS2): New.
5281 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5282 (ix86_valid_target_attribute_inner_p): Ditto.
5283 (ix86_init_mmx_sse_builtins): Add special_args2.
5284 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5285 TARGET_WBNOINVD_P): New.
5286 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5287 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5288 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5289 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5290 * config/i386/pconfigintrin.h: New file.
5291 * config/i386/wbnoinvdintrin.h: Ditto.
5292 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5293 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5295 2018-03-05 Richard Biener <rguenther@suse.de>
5297 PR tree-optimization/84670
5298 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5300 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5301 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5302 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5303 doesn't grow if all successors (recursively) were visited at least
5306 2018-03-05 Richard Biener <rguenther@suse.de>
5308 PR tree-optimization/84650
5309 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5310 if executed in the loop pipeline.
5312 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5314 * doc/configfiles.texi (Configuration Files): Move info about
5315 conditionalizing $target-protos.h to...
5316 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5317 differs from $target-protos.h.
5319 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5320 Chung-Ju Wu <jasonwucj@gmail.com>
5322 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5323 * config/nds32/nds32-multiple.md (setmemsi): Define.
5324 * config/nds32/nds32-memory-manipulation.c
5325 (nds32_gen_dup_4_byte_to_word_value): New.
5326 (emit_setmem_word_loop): New.
5327 (emit_setmem_byte_loop): New.
5328 (nds32_expand_setmem_loop): New.
5329 (nds32_expand_setmem_loop_v3m): New.
5330 (nds32_expand_setmem_unroll): New.
5331 (nds32_expand_setmem): New.
5333 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5334 Chung-Ju Wu <jasonwucj@gmail.com>
5336 * config/nds32/nds32-memory-manipulation.c
5337 (nds32_emit_load_store): New.
5338 (nds32_emit_post_inc_load_store): New.
5339 (nds32_emit_mem_move): New.
5340 (nds32_emit_mem_move_block): New.
5341 (nds32_expand_movmemsi_loop_unknown_size): New.
5342 (nds32_expand_movmemsi_loop_known_size): New.
5343 (nds32_expand_movmemsi_loop): New.
5344 (nds32_expand_movmemsi_unroll): New.
5345 (nds32_expand_movmemqi): Rename ...
5346 (nds32_expand_movmemsi): ... to this.
5347 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5348 (movmemsi): ... to this.
5349 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5350 (nds32_expand_movmemsi): ... to this.
5352 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5353 Monk Chiang <sh.chiang04@gmail.com>
5354 Chung-Ju Wu <jasonwucj@gmail.com>
5356 * config/nds32/nds32-protos.h
5357 (nds32_expand_load_multiple): New arguments.
5358 (nds32_expand_store_multiple): Ditto.
5359 (nds32_valid_multiple_load_store): Rename ...
5360 (nds32_valid_multiple_load_store_p): ... to this.
5361 * config/nds32/nds32-memory-manipulation.c
5362 (nds32_expand_load_multiple): Refine implementation.
5363 (nds32_expand_store_multiple): Ditto.
5364 * config/nds32/nds32-multiple.md
5365 (load_multiple): Update nds32_expand_load_multiple interface.
5366 (store_multiple): Update nds32_expand_store_multiple interface.
5367 * config/nds32/nds32-predicates.c
5368 (nds32_valid_multiple_load_store): Rename ...
5369 (nds32_valid_multiple_load_store_p): ... to this and refine
5371 * config/nds32/predicates.md
5372 (nds32_load_multiple_and_update_address_operation): New predicate.
5373 (nds32_store_multiple_and_update_address_operation): New predicate.
5375 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5376 Chung-Ju Wu <jasonwucj@gmail.com>
5378 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5379 (combo): New attribute.
5380 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5382 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5384 * config/nds32/nds32.opt: Change -mcmodel= default value.
5386 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5387 Monk Chiang <sh.chiang04@gmail.com>
5388 Chung-Ju Wu <jasonwucj@gmail.com>
5390 * config/nds32/constants.md (unspec_element): New enum.
5391 * config/nds32/constraints.md (Umw): New constraint.
5392 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5393 * config/nds32/nds32-intrinsic.md: Likewise.
5394 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5395 (nds32_valid_smw_lwm_base_p): New.
5396 (nds32_output_smw_single_word): New.
5397 (nds32_output_lmw_single_word): New.
5398 (nds32_expand_unaligned_load): New.
5399 (nds32_expand_unaligned_store): New.
5400 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5401 (nds32_output_smw_single_word): Declare.
5402 (nds32_output_lmw_single_word): Declare.
5403 (nds32_expand_unaligned_load): Declare.
5404 (nds32_expand_unaligned_store): Declare.
5405 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5406 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5407 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5408 NDS32_BUILTIN_UASTORE_DW.
5409 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5412 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5413 Kito Cheng <kito.cheng@gmail.com>
5414 Chung-Ju Wu <jasonwucj@gmail.com>
5416 * config/nds32/nds32-intrinsic.c
5417 (nds32_expand_builtin_null_ftype_reg): Delete.
5418 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5419 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5420 (nds32_read_argument): New.
5421 (nds32_legitimize_target): Ditto.
5422 (nds32_legitimize_argument): Ditto.
5423 (nds32_check_constant_argument): Ditto.
5424 (nds32_expand_unop_builtin): Ditto.
5425 (nds32_expand_unopimm_builtin): Ditto.
5426 (nds32_expand_binop_builtin): Ditto.
5427 (nds32_builtin_decl_impl): Ditto.
5428 (builtin_description): Ditto.
5429 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5430 (nds32_init_builtins_impl): Ditto.
5431 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5432 (nds32_builtin_decl): New.
5433 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5434 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5436 2018-03-02 Jeff Law <law@redhat.com>
5438 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5439 (redundant_insn, fill_simple_delay_slots): Likewise.
5440 (fill_slots_from_thread): Likewise.
5441 * resource.c (mark_referenced_resources): Likewise.
5442 (mark_set_resources, find_dead_or_set_registers): Likewise.
5444 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5446 * substring-locations.h (format_warning_va): Formatting fix for
5448 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5450 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5451 * substring-locations.c: Include intl.h.
5452 (format_warning_va): Turned into small wrapper around
5453 format_warning_n_va, renamed to ...
5454 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5455 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5457 (format_warning_at_substring_n): New function.
5458 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5459 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5460 format_warning_at_substring with just a shorter name instead of
5461 const function pointer.
5462 (fmtwarn_n): New function.
5463 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5464 appropriate, get rid of all the fmtstr temporaries, move conditionals
5465 with G_() wrapped string literals directly into fmtwarn arguments,
5466 cast dir.len to (int), formatting fixes.
5468 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5470 * doc/invoke.texi: Remove "Cilk Plus" references.
5472 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5473 Richard Biener <rguenther@suse.de>
5476 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5477 for error or warning attributes if CALL_FROM_THUNK_P is set.
5480 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5483 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5484 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5487 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5488 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5490 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5491 instead of -1U in last predictors element's probability member.
5493 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5496 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5497 arguments if they are comparable.
5499 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5501 PR tree-optimization/84634
5502 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5503 masks and masked_loop_p with a single loop_masks, making sure it's
5504 null for bb vectorization.
5506 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5508 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5509 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5510 loop->force_vectorize to check whether there is no alias.
5512 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5515 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5517 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5519 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5520 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5521 instead of a loop around prev_real_insn.
5522 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5526 * config/i386/i386.c (ix86_print_operand): Use conditional
5527 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5530 2018-03-02 Richard Biener <rguenther@suse.de>
5532 PR tree-optimization/84427
5533 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5534 (bitmap_set_subtract_values): Rewrite to handle multiple
5537 (prune_clobbered_mems): Likewise.
5538 (phi_translate): Take edge instead of pred/phiblock.
5539 (phi_translate_1): Likewise.
5540 (phi_translate_set): Likewise. Insert all translated
5541 exprs for a value into the set, keeping possibly multiple
5542 expressions per value.
5543 (compute_antic_aux): Adjust for phi_translate changes.
5544 When intersecting union the expressions and prune those
5545 not in the final value set, keeping possibly multiple
5546 expressions per value. Do not use value-insertion
5547 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5548 all expressions. Add verification that the value-sets
5549 only shrink during iteration.
5550 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5551 (do_pre_regular_insertion): Likewise.
5552 (do_pre_partial_partial_insertion): Likewise.
5554 2018-03-02 Richard Biener <rguenther@suse.de>
5557 * config/darwin.c (saved_debug_info_level): New static global.
5558 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5559 (darwin_asm_lto_end): Restore debug info generation settings.
5561 2018-03-01 Martin Liska <mliska@suse.cz>
5564 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5567 2018-03-01 Richard Biener <rguenther@suse.de>
5570 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5571 type annotation with LTO when debug was disabled at compile-time.
5573 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5575 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5577 (mips_final_postscan_insn): Likewise.
5579 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5581 PR rtl-optimization/84528
5582 * alias.c (init_alias_target): Add commentary.
5583 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5584 a unique base value if the frame pointer is not eliminated
5585 to the stack pointer.
5587 2018-03-01 Tom de Vries <tom@codesourcery.com>
5589 PR rtl-optimization/83327
5590 * lra-int.h (hard_regs_spilled_into): Declare.
5591 * lra.c (hard_regs_spilled_into): Define.
5592 (init_reg_info): Init hard_regs_spilled_into.
5593 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5594 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5595 (process_bb_lives): Handle hard_regs_spilled_into.
5596 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5597 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5599 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5601 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5602 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5603 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5604 * config/rs6000/aix72.h: New file.
5606 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5608 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5609 instead of warning_at with conditional singular and plural messages
5613 * stor-layout.c (update_alignment_for_field): For
5614 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5615 && !DECL_PACKED (field), do the alignment update, just use
5616 only desired_align instead of MAX (type_align, desired_align)
5618 (place_field): Don't do known_align < desired_align handling
5619 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5620 is non-NULL, instead do it after rli->prev_field handling and
5621 only if not within a bitfield word. For DECL_PACKED (field)
5622 use type_align of BITS_PER_UNIT.
5624 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5626 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5627 superfluous parentheses and trailing spaces.
5629 2018-02-28 Richard Biener <rguenther@suse.de>
5631 PR tree-optimization/84584
5632 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5633 SCoPs with fake exit edge.
5635 2018-02-28 Martin Liska <mliska@suse.cz>
5638 * timevar.c (timer::print): Fix format to properly print 100%
5641 2018-02-28 Richard Biener <rguenther@suse.de>
5644 * genmatch.c (capture_info::walk_match): Do not mark
5645 captured expressions without operands as expr_p given
5646 they act more like predicates and should be subject to
5647 "lost tail" side-effect preserving.
5649 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5651 PR rtl-optimization/81611
5652 * auto-inc-dec.c (attempt_change): Move dead note from
5653 mem_insn if it's the next use of regno
5654 (find_address): Take address use of reg holding
5655 non-incremented value. Add parm to limit search to the named
5657 (merge_in_block): Attempt to use a mem insn that is the next
5658 use of the original regno.
5660 2018-02-27 Martin Sebor <msebor@redhat.com>
5663 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5664 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5666 2018-02-27 Martin Sebor <msebor@redhat.com>
5668 PR translation/84207
5669 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5670 n argument to unsigned HOST_WIDE_INT.
5671 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5672 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5673 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5674 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5676 2018-02-27 Richard Biener <rguenther@suse.de>
5678 PR tree-optimization/84512
5679 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5680 Do not use the estimate returned from record_stmt_cost for
5681 the scalar iteration cost but sum properly using add_stmt_cost.
5683 2018-02-27 Richard Biener <rguenther@suse.de>
5685 PR tree-optimization/84466
5686 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5687 Adjust last change to less strictly validate use operands.
5689 2018-02-27 Martin Liska <mliska@suse.cz>
5691 PR gcov-profile/84548
5692 * gcov.c (process_file): Allow partial overlap and consider it
5693 also as group functions.
5694 (output_lines): Properly calculate range of lines for a group.
5696 2018-02-27 Martin Liska <mliska@suse.cz>
5698 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5699 'ggc' suffixes. Change first column width.
5700 (timer::print): Fix formatting of the column.
5702 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5704 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5705 preserve inline entry blocks for the sake of debug inline
5706 entry point markers alone.
5707 (remove_unused_locals): Suggest in comments a better place to
5708 force the preservation of inline entry blocks that are
5709 otherwise unused, but do not preserve them.
5711 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5713 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5715 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5718 * config/i386/constraints.md (Bs): Replace
5719 ix86_indirect_branch_register with
5720 TARGET_INDIRECT_BRANCH_REGISTER.
5722 * config/i386/i386.md (indirect_jump): Likewise.
5723 (tablejump): Likewise.
5724 (*sibcall_memory): Likewise.
5725 (*sibcall_value_memory): Likewise.
5726 Peepholes of indirect call and jump via memory: Likewise.
5727 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5728 (*sibcall_value_GOT_32): Likewise.
5729 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5730 (GOT_memory_operand): Likewise.
5731 (call_insn_operand): Likewise.
5732 (sibcall_insn_operand): Likewise.
5733 (GOT32_symbol_operand): Likewise.
5734 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5736 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5738 PR rtl-optimization/83496
5739 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5740 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5741 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5742 redundant insn, if any.
5743 (relax_delay_slots): Likewise.
5744 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5746 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5748 PR tree-optimization/83965
5749 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5750 that grouped statements are part of a reduction chain. Return
5751 true if the statement is not marked as a reduction itself but
5753 (vect_recog_dot_prod_pattern): Don't check whether the statement
5754 is part of a group here.
5755 (vect_recog_sad_pattern): Likewise.
5756 (vect_recog_widen_sum_pattern): Likewise.
5758 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5761 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5763 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5766 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5768 (ix86_output_indirect_function_return): New prototype.
5769 (ix86_split_simple_return_pop_internal): Likewise.
5770 * config/i386/i386.c (indirect_return_via_cx): New.
5771 (indirect_return_via_cx_bnd): Likewise.
5772 (indirect_thunk_name): Handle return va CX_REG.
5773 (output_indirect_thunk_function): Create alias for
5774 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5775 (ix86_output_indirect_jmp): Remove the bool argument.
5776 (ix86_output_indirect_function_return): New function.
5777 (ix86_split_simple_return_pop_internal): Likewise.
5778 * config/i386/i386.md (*indirect_jump): Don't pass false
5779 to ix86_output_indirect_jmp.
5780 (*tablejump_1): Likewise.
5781 (simple_return_pop_internal): Change it to define_insn_and_split.
5782 Call ix86_split_simple_return_pop_internal to split it for
5784 (simple_return_indirect_internal): Call
5785 ix86_output_indirect_function_return instead of
5786 ix86_output_indirect_jmp.
5788 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5791 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5792 memset and value initialization afterwards.
5794 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5796 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5798 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5801 * common/config/aarch64/aarch64-common.c
5802 (aarch_option_optimization_table[]): Switch
5803 off fomit-frame-pointer
5805 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5806 Chung-Ju Wu <jasonwucj@gmail.com>
5808 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5810 (store_multiple): Ditto.
5812 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5814 * config.gcc: Add --with-cpu support for nds32 target.
5815 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5816 * config/nds32/nds32.opt: Add -mcpu= option.
5818 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5820 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5821 isel=yes): Warn for these deprecated options.
5823 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5825 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5826 ISA_2_5_MASKS_EMBEDDED.
5828 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5830 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5831 p->max as pointers rather than using iterative_hash_expr.
5833 2018-02-23 Carl Love <cel@us.ibm.com>
5835 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5836 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5838 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5839 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5840 P8V_BUILTIN_VEC_VUNSIGNED2.
5842 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5845 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5846 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5848 * lra-constraints.c (curr_insn_transform): Set up
5849 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5851 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5852 * lra-eliminations.c (spill_pseudos): Ditto.
5853 (process_insn_for_elimination): Ditto.
5854 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5855 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5856 LRA_NON_CLOBBERED_ALT.
5858 2018-02-22 Martin Sebor <msebor@redhat.com>
5860 PR tree-optimization/84480
5861 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5862 to maybe_diag_stxncpy_trunc. Call it.
5863 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5864 from gimple_fold_builtin_strcpy. Print inlining stack.
5865 (handle_builtin_stxncpy): Print inlining stack.
5866 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5868 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5871 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5872 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5873 and -fcheck-pointer-bounds are used together.
5874 (indirect_thunk_prefix): New enum.
5875 (indirect_thunk_need_prefix): New function.
5876 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5877 "_nt" instead of "_bnd" for NOTRACK prefix.
5878 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5879 (output_indirect_thunk_function): Likewise.
5881 (ix86_code_end): Update output_indirect_thunk_function calls.
5882 (ix86_output_indirect_branch_via_reg): Replace
5883 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5884 (ix86_output_indirect_branch_via_push): Likewise.
5885 (ix86_output_function_return): Likewise.
5886 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5887 incompatible with -fcf-protection=branch and
5888 -fcheck-pointer-bounds.
5890 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5893 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5894 Change gcc_assert call to output_operand_lossage.
5896 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5898 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5900 2018-02-22 DJ Delorie <dj@redhat.com>
5901 Sebastian Perta <sebastian.perta@renesas.com>
5902 Oleg Endo <olegendo@gcc.gnu.org>
5904 * config/rx/rx.c (rx_rtx_costs): New function.
5905 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5907 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5909 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5911 2018-02-22 Martin Liska <mliska@suse.cz>
5914 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5915 Add "native" as a possible value.
5917 2018-02-22 Martin Liska <mliska@suse.cz>
5920 * config/i386/i386.c (ix86_option_override_internal):
5921 Add "native" as a possible value for -march and -mtune.
5923 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5926 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5927 to all type variants.
5929 PR tree-optimization/84503
5930 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5931 width as info->bitpos + info->bitsize - start.
5932 (merged_store_group::merge_overlapping): Simplify width computation.
5933 (check_no_overlap): New function.
5934 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5935 start + width and last_order of the group, fail if check_no_overlap
5937 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5938 to group if check_no_overlap fails.
5940 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5942 * config/rs6000/altivec.md: Delete contraint arguments to
5943 define_expand, define_split, and define_peephole2, and in
5944 define_insn_and_split if always unused.
5945 * config/rs6000/darwin.md: Ditto.
5946 * config/rs6000/dfp.md: Ditto.
5947 * config/rs6000/rs6000.md: Ditto.
5948 * config/rs6000/sync.md: Ditto.
5949 * config/rs6000/vector.md: Ditto.
5950 * config/rs6000/vsx.md: Ditto.
5952 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5954 * config/rs6000/altivec.md: Write output control strings as braced
5955 blocks instead of double-quoted strings.
5956 * config/rs6000/darwin.md: Ditto.
5957 * config/rs6000/rs6000.md: Ditto.
5958 * config/rs6000/vector.md: Ditto.
5959 * config/rs6000/vsx.md: Ditto.
5961 2018-02-21 Jason Merrill <jason@redhat.com>
5963 PR c++/84314 - ICE with templates and fastcall attribute.
5964 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5966 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5968 * ipa-cp.c (determine_versionability): Fix comment typos.
5970 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5973 * ipa-cp.c (determine_versionability): Do not version functions caling
5976 2018-02-21 Martin Liska <mliska@suse.cz>
5979 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5980 Add "native" as a possible value.
5981 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5982 the macro when native cpu detection is available.
5984 2018-02-21 Martin Liska <mliska@suse.cz>
5987 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5988 Add "native" as a possible value.
5989 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5990 when native cpu detection is available.
5992 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5993 Martin Sebor <msebor@redhat.com>
5995 PR tree-optimization/84478
5996 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5998 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5999 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6000 support which is conservatively correct, for 2 only stay conservative
6001 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6002 argument to the 2 argument get_range_strlen, adjust 6 arg
6003 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6005 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6006 (gimple_fold_builtin_strlen): Pass true as last argument to
6009 2018-02-20 Martin Sebor <msebor@redhat.com>
6012 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6013 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6014 (builtin_memref::builtin_memref): Factor out parts into
6015 set_base_and_offset and call it.
6017 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6020 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6021 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6022 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6023 search at the associated MODE_INT.
6025 2018-02-20 Jeff Law <law@redhat.com>
6028 PR tree-optimization/81592
6030 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6031 for range data rather than using global data.
6032 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6033 range data rather than using global data.
6034 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6035 pass it to children as needed.
6036 (struct directive::fmtresult): Similarly.
6037 (struct directive::set_width): Similarly.
6038 (struct directive::set_precision): Similarly.
6039 (format_integer, format_directive, parse_directive): Similarly.
6040 (format_none): Accept unnamed vr_values parameter.
6041 (format_percent, format_floating, format_character): Similarly.
6042 (format_string, format_plain): Similarly.
6043 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6044 the EVRP range analyzer for range data rather than using global data.
6045 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6046 gimple-ssa-evrp-analyze.h
6047 (class sprintf_dom_walker): Add after_dom_children member function.
6048 Add evrp_range_analyzer member.
6049 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6050 range analyzer as needed.
6051 (sprintf_dom_walker::after_dom_children): New member function.
6052 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6054 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6055 (evrp_range_analyzer::pop_to_marker): Likewise.
6057 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6059 PR tree-optimization/84419
6060 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6061 with the required type if its current type is compatible but
6064 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6067 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6068 after vectorization.
6070 2018-02-20 Martin Liska <mliska@suse.cz>
6073 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6074 possible values if we don't have a hint.
6076 2018-02-20 Martin Liska <mliska@suse.cz>
6080 * final.c (shorten_branches): Build align_tab array with one
6082 * opts.c (finish_options): Add alignment option limit check.
6083 (MAX_CODE_ALIGN): Likewise.
6084 (MAX_CODE_ALIGN_VALUE): Likewise.
6085 * doc/invoke.texi: Document maximum allowed option value for
6086 all -falign-* options.
6088 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6091 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6092 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6093 * var-tracking.c (emit_note_insn_var_location): Remove all references
6094 to NOTE_INSN_CALL_ARG_LOCATION.
6095 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6096 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6097 Use copy_rtx_if_shared.
6098 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6099 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6100 (dwarf2out_var_location): Remove handling of
6101 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6103 * final.c (final_scan_insn): Remove all references to
6104 NOTE_INSN_CALL_ARG_LOCATION.
6105 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6106 before dumping final insns.
6107 * except.c (emit_note_eh_region_end): Remove all references to
6108 NOTE_INSN_CALL_ARG_LOCATION.
6109 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6110 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6111 * config/arc/arc.c (hwloop_optimize): Likewise.
6112 * config/arm/arm.c (create_fix_barrier): Likewise.
6113 * config/s390/s390.c (s390_chunkify_start): Likewise.
6114 * config/sh/sh.c (find_barrier): Likewise.
6115 * config/i386/i386.c (rest_of_insert_endbranch,
6116 ix86_seh_fixup_eh_fallthru): Likewise.
6117 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6118 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6119 * config/frv/frv.c (frv_function_prologue): Likewise.
6120 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6122 (note_outside_basic_block_p): Remove all references to
6123 NOTE_INSN_CALL_ARG_LOCATION.
6124 * gengtype.c (adjust_field_rtx_def): Likewise.
6125 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6127 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6128 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6131 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6134 PR tree-optimization/84452
6135 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6136 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6139 2018-02-19 Martin Liska <mliska@suse.cz>
6142 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6144 2018-02-19 Martin Liska <mliska@suse.cz>
6145 Richard Sandiford <richard.sandiford@linaro.org>
6147 PR tree-optimization/82491
6148 * gimple-fold.c (get_base_constructor): Make earlier bail out
6151 2018-02-19 Carl Love <cel@us.ibm.com>
6153 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6154 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6156 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6157 P8V_BUILTIN_VEC_NEG.
6159 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6161 * config/rl78/rl78.md (movdf): New define expand.
6163 2018-02-19 Martin Liska <mliska@suse.cz>
6166 * doc/invoke.texi: Fix typo.
6167 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6169 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6171 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6172 handle rs6000_single_float and rs6000_double_float specially for
6175 2018-02-16 Jeff Law <law@redhat.com>
6177 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6180 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6183 * ipa-inline.c (inline_small_functions): Fix a typo.
6185 2018-02-16 Nathan Sidwell <nathan@acm.org>
6187 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6189 2018-02-16 Carl Love <cel@us.ibm.com>
6191 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6192 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6193 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6194 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6195 expansion to P8V_BUILTIN_VEC_FLOAT2.
6197 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6199 PR rtl-optimization/70023
6200 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6201 src_regno into account.
6203 2018-02-16 Carl Love <cel@us.ibm.com>
6205 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6206 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6207 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6208 * config/rs6000/rs6000.c: Remove case statements for
6209 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6210 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6211 and P9V_BUILTIN_VEC_VINSERT4B.
6212 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6213 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6214 * config/rs6000/vsx.md:
6215 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6218 2018-02-16 Carl Love <cel@us.ibm.com>
6220 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6222 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6224 * config/rs6000/rs6000-c.c: Add the definitions for
6225 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6226 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6227 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6228 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6229 definition for insert4b and define insn *insert3b_internal.
6230 * doc/extend.texi: Add documentation for vec_extract4b.
6232 2018-02-16 Nathan Sidwell <nathan@acm.org>
6234 * doc/extend.texi (Backwards Compatibility): Mention friend
6235 injection. Note for-scope is deprecated.
6236 * doc/invoke.texi (-ffriend-injection): Deprecate.
6238 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6240 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6241 that moved to I2, also allow destinations that are a paradoxical
6242 subreg (instead of a normal reg).
6244 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6247 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6250 2018-02-16 Richard Biener <rguenther@suse.de>
6252 PR tree-optimization/84037
6253 PR tree-optimization/84016
6255 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6256 Adjust vec_construct for the fact we need additional higher latency
6257 128bit inserts for AVX256 and AVX512 vector builds.
6258 (ix86_add_stmt_cost): Scale vector construction cost for
6261 2018-02-16 Richard Biener <rguenther@suse.de>
6263 PR tree-optimization/84417
6264 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6265 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6266 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6268 2018-02-16 Martin Liska <mliska@suse.cz>
6271 * internal-fn.def (ASAN_CHECK): Set proper flags.
6272 (ASAN_MARK): Likewise.
6274 2018-02-16 Julia Koval <julia.koval@intel.com>
6276 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6277 from PTA_CANNONLAKE.
6279 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6282 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6283 Use ++iter rather than iter++ for std::list iterators.
6284 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6285 defer deleting them until all nodes in the forest are processed. Do
6286 free even leaf nodes. Change to_process into auto_vec.
6289 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6290 * vec.h (vec_default_construct): Use memset instead of placement new
6291 if BROKEN_VALUE_INITIALIZATION is defined.
6292 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6293 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6296 PR rtl-optimization/83723
6297 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6298 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6299 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6301 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6303 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6305 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6307 PR rtl-optimization/81443
6308 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6309 from inner REGs to paradoxical SUBREGs.
6311 2018-02-16 Richard Biener <rguenther@suse.de>
6313 PR tree-optimization/84399
6314 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6315 For operands we can analyze at their definition make sure we can
6316 analyze them at each use as well.
6318 2018-02-16 Richard Biener <rguenther@suse.de>
6320 PR tree-optimization/84190
6321 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6322 volatile accesses if the decl isn't volatile.
6324 2018-02-15 Jason Merrill <jason@redhat.com>
6326 PR c++/84314 - ICE with templates and fastcall attribute.
6327 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6328 TYPE_CANONICAL on an existing type.
6330 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6332 PR tree-optimization/84383
6333 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6334 dstoff nor call operand_equal_p if dstbase is NULL.
6336 PR tree-optimization/84334
6337 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6338 also a CONSTANT_CLASS_P, punt.
6340 2018-02-14 Jim Wilson <jimw@sifive.com>
6342 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6343 first SMALL_OPERAND check. New local min_second_step. Move assert
6344 to where locals are set. Add TARGET_RVC support.
6345 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6347 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6349 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6351 2018-02-14 Martin Sebor <msebor@redhat.com>
6353 PR tree-optimization/83698
6354 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6355 arrays constrain the offset range to their bounds.
6356 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6357 (builtin_access::overlap): Avoid setting the size of overlap if it's
6359 (maybe_diag_overlap): Also consider arrays when deciding what values
6360 of offsets to include in diagnostics.
6362 2018-02-14 Martin Sebor <msebor@redhat.com>
6365 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6366 that correspond to the kind of a declaration.
6368 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6371 * config/pa/pa.md: Load address of PIC label using the linkage table
6372 if the label is nonlocal.
6374 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6376 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6377 warning message if user requests -maltivec=be.
6378 * doc/invoke.texi: Document deprecation of -maltivec=be.
6380 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6383 * config/rs6000/rs6000-c.c: Update definitions for
6384 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6385 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6387 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6390 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6391 add _get_ssp intrinsics. Remove argument from
6392 __builtin_ia32_rdssp[d|q].
6393 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6394 * config/i386/i386-builtin.def: Remove argument from
6395 __builtin_ia32_rdssp[d|q].
6396 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6397 ix86_expand_special_args_builtin for _rdssp[d|q].
6398 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6399 Clear register before usage.
6400 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6401 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6403 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6405 PR tree-optimization/84357
6406 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6407 operand 1 of an ARRAY_REF too.
6409 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6412 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6413 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6415 (set_of_reg): New struct.
6416 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6417 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6418 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6420 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6421 Split into bitclr, bitset, bitinvert patterns if appropriate.
6422 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6423 use rx_fuse_in_memory_bitop.
6424 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6425 to named insn, correct maximum insn length.
6427 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6430 * machmode.def: Define a complex mode for PARTIAL_INT.
6431 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6433 * doc/rtl.texi: Document CSPImode.
6434 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6436 (msp430_hard_regno_nregs_with_padding): Likewise.
6438 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6441 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6443 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6445 PR rtl-optimization/84169
6446 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6447 we generated a parallel as new i3 and we split that to new i2 and i3
6448 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6449 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6450 those to i2, not i1. Partially rewrite this scan code.
6452 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6455 * stor-layout.c (place_field): For variable length fields, adjust
6456 offset_align afterwards not just based on the field's alignment,
6457 but also on the size.
6460 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6461 of exps and logs in the use_exp2 case.
6463 2018-02-13 Jeff Law <law@redhat.com>
6465 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6468 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6471 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6475 * common.opt (gas-loc-support, gas-locview-support): New.
6476 (ginline-points, ginternal-reset-location-views): New.
6477 * doc/invoke.texi: Document them. Use @itemx where intended.
6478 (gvariable-location-views): Adjust.
6479 * target.def (reset_location_view): New.
6480 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6481 (TARGET_RESET_LOCATION_VIEW): New.
6482 * doc/tm.texi: Rebuilt.
6483 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6484 (dwarf2out_default_as_locview_support): New.
6485 (output_asm_line_debug_info): Use option variables.
6486 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6487 (output_loc_list): Likewise.
6488 (add_high_low_attributes): Check option variables.
6489 Don't output entry view attribute in strict mode.
6490 (gen_inlined_subroutine_die): Check option variables.
6491 (dwarf2out_inline_entry): Likewise.
6492 (init_sections_and_labels): Likewise.
6493 (dwarf2out_early_finish): Likewise.
6494 (maybe_reset_location_view): New, from...
6495 (dwarf2out_var_location): ... here. Call it.
6496 * debug.h (dwarf2out_default_as_loc_support): Declare.
6497 (dwarf2out_default_as_locview_support): Declare.
6498 * hooks.c (hook_int_rtx_insn_0): New.
6499 * hooks.h (hook_int_rtx_insn_0): Declare.
6500 * toplev.c (process_options): Take -gas-loc-support and
6501 -gas-locview-support from dwarf2out. Enable
6502 -gvariable-location-views by default only with locview
6503 assembler support. Enable -ginternal-reset-location-views by
6504 default only if the target defines the corresponding hook.
6505 Enable -ginline-points by default if location views are
6506 enabled; force it disabled if statement frontiers are
6508 * tree-inline.c (expand_call_inline): Check option variables.
6509 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6511 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6513 PR tree-optimization/84321
6514 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6515 handling. Also check whether the anti-range contains any values
6516 that satisfy the mask; switch to a VR_RANGE if not.
6518 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6521 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6523 2018-02-13 Martin Jambor <mjambor@suse.cz>
6526 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6527 of call statements, also set location of a load to a temporary.
6529 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6531 * config/rl78/rl78.c (add_vector_labels): New function.
6532 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6533 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6534 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6535 which checks that no arguments are passed.
6536 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6537 * doc/extend.texi: Documentation for the new attribute.
6539 2018-02-13 Andreas Schwab <schwab@suse.de>
6541 * config/riscv/linux.h (CPP_SPEC): Define.
6543 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6546 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6547 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6548 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6549 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6550 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6551 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6552 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6553 temporarily for AES and PCLMUL builtins.
6555 PR tree-optimization/84339
6556 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6557 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6561 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6562 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6563 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6565 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6567 * omp-simd-clone.h: New file.
6568 * omp-simd-clone.c: Include omp-simd-clone.h.
6569 (expand_simd_clones): No longer static.
6570 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6571 cgraph.h and omp-simd-clone.h.
6572 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6573 (vect_recog_widen_shift_pattern): Formatting fix.
6574 (vect_pattern_recog_1): Don't check optab for calls.
6577 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6578 operands[2] into a REG before using gen_lowpart on it.
6580 2018-02-12 Jeff Law <law@redhat.com>
6583 * config/sh/sh.c (find_barrier): Consider a sibling call
6586 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6587 successfully back substituting a reg.
6589 2018-02-12 Richard Biener <rguenther@suse.de>
6591 PR tree-optimization/84037
6592 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6593 parameter, move visited init to caller.
6594 (vect_slp_analyze_operations): Separate cost from validity
6595 check, initialize visited once for all instances.
6596 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6598 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6600 (vect_model_promotion_demotion_cost): Likewise.
6601 (vectorizable_bswap): Guard cost modeling with !slp_node
6602 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6604 (vectorizable_call): Likewise.
6605 (vectorizable_conversion): Likewise.
6606 (vectorizable_assignment): Likewise.
6607 (vectorizable_shift): Likewise.
6608 (vectorizable_operation): Likewise.
6609 (vectorizable_store): Likewise.
6610 (vectorizable_load): Likewise.
6611 (vectorizable_condition): Likewise.
6612 (vectorizable_comparison): Likewise.
6614 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6617 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6618 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6619 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6621 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6624 * params.def (inline-min-speedup): Increase from 8 to 15.
6625 (max-inline-insns-auto): Decrease from 40 to 30.
6626 * ipa-split.c (consider_split): Add some buffer for function to
6627 be considered inlining candidate.
6628 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6631 2018-02-12 Richard Biener <rguenther@suse.de>
6633 PR tree-optimization/84037
6634 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6635 matched stmts if we cannot swap the non-matched ones.
6637 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6639 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6640 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6641 _mm_maskz_scalef_round_ss): New intrinsics.
6642 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6643 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6644 __builtin_ia32_scalefss_round): Remove.
6645 (__builtin_ia32_scalefsd_mask_round,
6646 __builtin_ia32_scalefss_mask_round): New intrinsics.
6647 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6648 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6649 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6650 "<round_constraint>")): Changed to ...
6651 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6652 "<round_scalar_constraint>")): ... this.
6653 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6654 %0, %1, %2<round_op3>}"): Changed to ...
6655 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6656 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6657 %2<round_scalar_mask_op3>}"): ... this.
6658 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6660 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6662 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6663 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6664 (_mm_maskz_sqrt_round_ss): New intrinsics.
6665 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6666 (__builtin_ia32_sqrtsd_mask_round)
6667 (__builtin_ia32_sqrtss_mask_round): New builtins.
6668 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6669 (__builtin_ia32_sqrtss_round): Remove.
6670 (__builtin_ia32_sqrtsd_mask_round)
6671 (__builtin_ia32_sqrtss_mask_round): New builtins.
6672 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6673 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6674 ((match_operand:VF_128 1 "vector_operand"
6675 "xBm,<round_constraint>")): Changed to ...
6676 ((match_operand:VF_128 1 "vector_operand"
6677 "xBm,<round_scalar_constraint>")): ... this.
6678 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6679 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6680 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6681 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6682 %<iptr>1<round_scalar_mask_op3>}): ... this.
6683 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6684 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6686 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6689 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6690 Cast vec_cmpeq result to correct type.
6691 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6692 Cast vec_cmpgt result to correct type.
6694 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6696 * final.c (final_scan_insn_1): Renamed from...
6697 (final_scan_insn): ... this. New wrapper, to recover
6698 seen from the outermost call in recursive ones.
6699 * config/sparc/sparc.c (output_return): Drop seen from call.
6700 (output_sibcall): Likewise.
6701 * config/visium/visium.c (output_branch): Likewise.
6703 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6705 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6708 2018-02-10 Alan Modra <amodra@gmail.com>
6711 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6712 Specify LR as an input.
6714 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6717 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6718 remove_member_access_dummy_vars): New functions.
6719 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6720 lower_omp_1, execute_lower_omp): Use them.
6722 PR rtl-optimization/84308
6723 * shrink-wrap.c (spread_components): Release todo vector.
6725 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6727 PR rtl-optimization/57193
6728 * ira-color.c (struct allocno_color_data): Add member
6729 conflict_allocno_hard_prefs.
6730 (update_conflict_allocno_hard_prefs): New.
6731 (bucket_allocno_compare_func): Add a preference based on
6732 conflict_allocno_hard_prefs.
6733 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6734 (color_allocnos): Remove a dead code. Initiate
6735 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6737 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6740 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6741 constraint from =wa to wa. Avoid a subreg on the output operand,
6742 instead use a pseudo and subreg it in a move.
6743 (p9_xxbrd_<mode>): Changed to ...
6744 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6745 (p9_xxbrd_v2df): New expander.
6746 (p9_xxbrw_<mode>): Changed to ...
6747 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6748 (p9_xxbrw_v4sf): New expander.
6750 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6752 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6754 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6757 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6758 multiply in 32-bit mode.
6759 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6760 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6763 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6765 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6766 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6767 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6768 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6770 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6773 * dwarf2out.c (is_trivial_indirect_ref): New function.
6774 (dwarf2out_late_global_decl): Do not generate a location
6775 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6776 and that are not defined in the current unit.
6778 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6780 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6781 instead of a libcall for UNORDERED.
6783 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6786 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6787 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6789 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6792 * config/s390/s390.c (s390_set_current_function): Invoke
6793 s390_indirect_branch_settings also if fndecl didn't change.
6795 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6797 * config/rs6000/rs6000.md (blockage): Set length to zero.
6799 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6801 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6803 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6806 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6807 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6811 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6812 PARALLEL incoming that failed vt_get_decl_and_offset check.
6815 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6816 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6818 (get_variable_section): For decls in named .bss* sections pass true as
6819 second argument to bss_initializer_p.
6821 2018-02-09 Marek Polacek <polacek@redhat.com>
6822 Jakub Jelinek <jakub@redhat.com>
6825 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6826 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6827 Sync some changes from cxx_fold_indirect_ref.
6829 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6831 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6833 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6834 (BLOCK_INLINE_ENTRY_LABEL): New.
6835 (dwarf2out_var_location): Disregard inline entry markers.
6836 (inline_entry_data): New struct.
6837 (inline_entry_data_hasher): New hashtable type.
6838 (inline_entry_data_hasher::hash): New.
6839 (inline_entry_data_hasher::equal): New.
6840 (inline_entry_data_table): New variable.
6841 (add_high_low_attributes): Add DW_AT_entry_pc and
6842 DW_AT_GNU_entry_view attributes if a pending entry is found
6843 in inline_entry_data_table. Add old entry_pc attribute only
6844 if debug nonbinding markers are disabled.
6845 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6846 markers are enabled.
6847 (block_within_block_p, dwarf2out_inline_entry): New.
6848 (dwarf2out_finish): Check that no entries remained in
6849 inline_entry_data_table.
6850 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6851 (final_scan_insn, notice_source_line): Likewise.
6852 (rest_of_clean_state): Skip inline entry markers.
6853 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6855 * gimple.c (gimple_build_debug_inline_entry): New.
6856 * gimple.h (enum gimple_debug_subcode): Add
6857 GIMPLE_DEBUG_INLINE_ENTRY.
6858 (gimple_build_debug_inline_entry): Declare.
6859 (gimple_debug_inline_entry_p): New.
6860 (gimple_debug_nonbind_marker_p): Adjust.
6861 * insn-notes.def (INLINE_ENTRY): New.
6862 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6863 inline entry marker notes.
6864 (print_insn): Likewise.
6865 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6866 (INSN_DEBUG_MARKER_KIND): Likewise.
6867 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6868 * tree-inline.c (expand_call_inline): Build and insert
6869 debug_inline_entry stmt.
6870 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6871 inline entry blocks early, if nonbind markers are enabled.
6872 (dump_scope_block): Dump fragment info.
6873 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6874 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6875 (gimple_build_debug_inline_entry): New.
6876 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6877 Enable/disable inline entry points too.
6878 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6879 (DEBUG_INSN): Describe inline entry markers.
6881 * common.opt (gvariable-location-views): New.
6882 (gvariable-location-views=incompat5): New.
6883 * config.in: Rebuilt.
6884 * configure: Rebuilt.
6885 * configure.ac: Test assembler for view support.
6886 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6887 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6888 * dwarf2out.c (var_loc_view): New typedef.
6889 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6890 (dwarf2out_locviews_in_attribute): New.
6891 (dwarf2out_locviews_in_loclist): New.
6892 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6893 (enum dw_line_info_opcode): Add LI_adv_address.
6894 (struct dw_line_info_table): Add view.
6895 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6896 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6897 (zero_view_p): New variable.
6898 (ZERO_VIEW_P): New macro.
6899 (output_asm_line_debug_info): New.
6900 (struct var_loc_node): Add view.
6901 (add_AT_view_list, AT_loc_list): New.
6902 (add_var_loc_to_decl): Add view param. Test it against last.
6903 (new_loc_list): Add view params. Record them.
6904 (AT_loc_list_ptr): Handle loc and view lists.
6905 (view_list_to_loc_list_val_node): New.
6906 (print_dw_val): Handle dw_val_class_view_list.
6907 (size_of_die): Likewise.
6908 (value_format): Likewise.
6909 (loc_list_has_views): New.
6910 (gen_llsym): Set vl_symbol too.
6911 (maybe_gen_llsym, skip_loc_list_entry): New.
6912 (dwarf2out_maybe_output_loclist_view_pair): New.
6913 (output_loc_list): Output view list or entries too.
6914 (output_view_list_offset): New.
6915 (output_die): Handle dw_val_class_view_list.
6916 (output_dwarf_version): New.
6917 (output_compilation_unit_header): Use it.
6918 (output_skeleton_debug_sections): Likewise.
6919 (output_rnglists, output_line_info): Likewise.
6920 (output_pubnames, output_aranges): Update version comments.
6921 (output_one_line_info_table): Output view numbers in asm comments.
6922 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6923 Call maybe_gen_llsym.
6924 (loc_list_from_tree_1): Adjust.
6925 (add_AT_location_description): Create view list attribute if
6926 needed, check it's absent otherwise.
6927 (convert_cfa_to_fb_loc_list): Adjust.
6928 (maybe_emit_file): Call output_asm_line_debug_info for test.
6929 (dwarf2out_var_location): Reset views as needed. Precompute
6930 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6931 attribute. Set view.
6932 (new_line_info_table): Reset next view.
6933 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6934 (dwarf2out_source_line): Likewise. Output view resets and labels to
6935 the assembler, or select appropriate line info opcodes.
6936 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6937 (optimize_string_length): Catch it. Adjust.
6938 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6939 dw_val_class_view_list, and remove it if no longer needed.
6940 (hash_loc_list): Hash view numbers.
6941 (loc_list_hasher::equal): Compare them.
6942 (optimize_location_lists): Check whether a view list symbol is
6943 needed, and whether the locview attribute is present, and
6944 whether they match. Remove the locview attribute if no longer
6946 (index_location_lists): Call skip_loc_list_entry for test.
6947 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6948 Use output_dwarf_version.
6949 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6950 (struct dw_val_node): Add val_view_list.
6951 * final.c (SEEN_NEXT_VIEW): New.
6952 (set_next_view_needed): New.
6953 (clear_next_view_needed): New.
6954 (maybe_output_next_view): New.
6955 (final_start_function): Rename to...
6956 (final_start_function_1): ... this. Take pointer to FIRST,
6957 add SEEN parameter. Emit param bindings in the initial view.
6958 (final_start_function): Reintroduce SEEN-less interface.
6959 (final): Rename to...
6960 (final_1): ... this. Take SEEN parameter. Output final pending
6961 next view at the end.
6962 (final): Reintroduce seen-less interface.
6963 (final_scan_insn): Output pending next view before switching
6964 sections or ending a block. Mark the next view as needed when
6965 outputting variable locations. Notify debug backend of section
6966 changes, and of location view changes.
6967 (rest_of_handle_final): Adjust.
6968 * toplev.c (process_options): Autodetect value for debug variable
6969 location views option. Warn on incompat5 without -gdwarf-5.
6970 * doc/invoke.texi (gvariable-location-views): New.
6971 (gvariable-location-views=incompat5): New.
6972 (gno-variable-location-views): New.
6974 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6976 PR tree-optimization/84136
6977 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6978 that the result of find_edge is non-NULL.
6980 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6983 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6984 storing integer register in SImode. Fix cost of 256 and 512
6985 byte aligned SSE register store.
6987 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6989 * config/i386/i386.c (ix86_multiplication_cost): Fix
6990 multiplication cost for TARGET_AVX512DQ.
6992 2018-02-08 Marek Polacek <polacek@redhat.com>
6994 PR tree-optimization/84238
6995 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6998 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7000 PR tree-optimization/84265
7001 * tree-vect-stmts.c (vectorizable_store): Don't treat
7002 VMAT_CONTIGUOUS accesses as grouped.
7003 (vectorizable_load): Likewise.
7005 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7007 PR tree-optimization/81635
7008 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7009 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7010 (test_round_for_mask): New functions.
7011 (wide_int_cc_tests): Call test_round_for_mask.
7012 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7013 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7014 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7015 range returned by get_range_info.
7017 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7020 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7021 * symtab.c: Include builtins.h
7022 (symtab_node::output_to_lto_symbol_table_p): Move here
7023 from lto-streamer-out.c:output_symbol_p.
7024 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7025 (output_symbol_p): Move all logic to symtab.c
7026 (produce_symtab): Update.
7028 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7030 * config/s390/s390-opts.h (enum indirect_branch): Define.
7031 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7032 (s390_indirect_branch_via_thunk)
7033 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7034 (enum s390_indirect_branch_type): Define.
7035 * config/s390/s390.c (struct s390_frame_layout, struct
7036 machine_function): Remove.
7037 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7038 (indirect_branch_table_label_no, indirect_branch_table_name):
7040 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7041 (enum s390_indirect_branch_option): Define.
7042 (s390_return_addr_from_memory): New function.
7043 (s390_handle_string_attribute): New function.
7044 (s390_attribute_table): Add new attribute handler.
7045 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7046 (s390_indirect_branch_via_thunk): New function.
7047 (s390_indirect_branch_via_inline_thunk): New function.
7048 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7049 sibling call optimization for non z10 compiles.
7050 (s390_emit_call): Force indirect branch target to be a single
7051 register. Add r1 clobber for non-z10 compiles.
7052 (s390_emit_epilogue): Emit return jump via return_use expander.
7053 (s390_reorg): Handle JUMP_INSNs as execute targets.
7054 (s390_option_override_internal): Perform validity checks for the
7055 new command line options.
7056 (s390_indirect_branch_attrvalue): New function.
7057 (s390_indirect_branch_settings): New function.
7058 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7059 (s390_output_indirect_thunk_function): New function.
7060 (s390_code_end): Implement target hook.
7061 (s390_case_values_threshold): Implement target hook.
7062 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7064 * config/s390/s390.h (struct s390_frame_layout)
7065 (struct machine_function): Move here from s390.c.
7066 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7067 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7068 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7069 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7070 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7071 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7072 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7073 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7074 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7075 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7076 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7077 (mnemonic attribute): Add values which aren't recognized
7079 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7080 pattern for branch conversion. Fix mnemonic attribute.
7081 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7082 indirect branch via thunk if requested.
7083 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7084 ("*indirect_jump"): Disable for branch conversion using out of
7086 ("indirect_jump_via_thunk<mode>_z10")
7087 ("indirect_jump_via_thunk<mode>")
7088 ("indirect_jump_via_inlinethunk<mode>_z10")
7089 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7090 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7091 ("casesi_jump_via_inlinethunk<mode>_z10")
7092 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7093 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7094 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7095 ("*indirect2_jump"): Disable for branch conversion.
7096 ("casesi_jump"): Turn into expander and expand patterns for branch
7098 ("return_use"): New expander.
7099 ("*return"): Emit return via thunk and rename it to ...
7100 ("*return<mode>"): ... this one.
7101 * config/s390/s390.opt: Add new options and and enum for the
7104 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7106 * lra-constraints.c (match_reload): Unconditionally use
7107 gen_lowpart_SUBREG, rather than selecting between that
7108 and equivalent gen_rtx_SUBREG code.
7110 2018-02-08 Richard Biener <rguenther@suse.de>
7112 PR tree-optimization/84233
7113 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7114 changed flag instead of boguously re-using phi_inserted.
7116 2018-02-08 Martin Jambor <mjambor@suse.cz>
7118 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7119 static local variables.
7121 2018-02-08 Richard Biener <rguenther@suse.de>
7123 PR tree-optimization/84278
7124 * tree-vect-stmts.c (vectorizable_store): When looking for
7125 smaller vector types to perform grouped strided loads/stores
7126 make sure the mode is supported by the target.
7127 (vectorizable_load): Likewise.
7129 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7131 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7132 Increase LDP/STP opportunities by adding adjacent callee-saves.
7134 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7136 PR rtl-optimization/84068
7137 PR rtl-optimization/83459
7138 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7140 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7142 PR tree-optimization/84224
7143 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7144 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7147 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7150 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7151 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7152 count, remove one USE.
7154 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7156 * doc/install.texi (Configuration): Document the
7157 --with-long-double-format={ibm,ieee} PowerPC configuration
7161 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7162 Convert from define_expand to be define_insn_and_split. Rework
7163 float/double/_Float128 conversions to QI/HI/SImode to work with
7164 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7165 conversions to QI/HImode types did a store and then a load to
7166 truncate the value. For conversions to VSX registers, don't split
7167 the insn, instead emit the code directly. Use the code iterator
7168 any_fix to combine signed and unsigned conversions.
7169 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7170 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7171 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7172 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7173 (fix_<mode>di2_hw): Likewise.
7174 (fixuns_<mode>di2_hw): Likewise.
7175 (fix_<mode>si2_hw): Likewise.
7176 (fixuns_<mode>si2_hw): Likewise.
7177 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7178 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7179 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7180 fix<uns>_trunc<SFDF:mode>si2_p8.
7181 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7183 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7184 (fix<uns>_<mode>_mem): Likewise.
7185 (fctiw<u>z_<mode>_mem): Likewise.
7186 (fix<uns>_<mode>_mem): Likewise.
7187 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7188 the register allocator from doing a direct move to the GPRs to do
7189 a store, and instead use the ISA 3.0 store byte/half-word from
7190 vector register instruction. For IEEE 128-bit floating point,
7191 also optimize stores of 32-bit ints.
7192 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7194 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7196 * genextract.c (push_pathstr_operand): New function to support
7198 (walk_rtx): Call push_pathstr_operand.
7199 (print_path): Support [a-zA-Z].
7201 2018-02-07 Richard Biener <rguenther@suse.de>
7203 PR tree-optimization/84037
7204 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7205 (cse_and_gimplify_to_preheader): Declare.
7206 (vect_get_place_in_interleaving_chain): Likewise.
7207 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7209 (_loop_vec_info::~_loop_vec_info): Delete it.
7210 (cse_and_gimplify_to_preheader): New function.
7211 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7212 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7213 (vectorizable_load): Likewise. For grouped stores always base
7214 the IV on the first element.
7215 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7216 condition before gimplifying.
7218 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7220 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7221 *DIV_EXPR and *MOD_EXPR.
7223 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7226 * config/i386/i386.c (ix86_option_override_internal): Mask out
7227 the CF_SET bit when checking -fcf-protection.
7229 2018-02-07 Tom de Vries <tom@codesourcery.com>
7232 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7235 2018-02-07 Richard Biener <rguenther@suse.de>
7237 PR tree-optimization/84204
7238 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7241 PR tree-optimization/84205
7242 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7243 special-case isl_ast_op_zdiv_r.
7245 PR tree-optimization/84223
7246 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7247 Only add conditions from within the region.
7248 (gather_bbs::after_dom_children): Adjust.
7250 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7253 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7254 * config/avr/avr.md: Only post-reload split REG-REG moves if
7255 either register is GENERAL_REG_P.
7257 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7259 PR tree-optimization/84235
7260 * tree-ssa-scopedtables.c
7261 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7262 if the subtraction is performed in floating point type where NaNs are
7263 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7264 build 1. Formatting fix.
7266 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7269 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7270 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7271 and skip it regardless of bb boundaries. Use CALL_P macro,
7272 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7273 unnecessarily, formatting fix.
7275 2018-02-06 Michael Collison <michael.collison@arm.com>
7277 * config/arm/thumb2.md:
7278 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7279 (*thumb_mov_notscc): Ditto.
7281 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7284 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7285 unsigned_fix, not "s".
7287 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7289 * configure.ac (gcc_fn_eh_frame_ro): New function.
7290 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7291 correct .eh_frame permissions.
7292 * configure: Regenerate.
7294 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7296 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7299 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7301 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7302 Display warning message for -mno-speculate-indirect-jumps.
7304 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7306 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7308 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7310 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7312 PR tree-optimization/84225
7313 * tree-eh.c (find_trapping_overflow): Only call
7314 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7316 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7319 * config/i386/i386.c: Reimplement the check of possible options
7320 -mibt/-mshstk conbination. Change error messages.
7321 * doc/invoke.texi: Fix a typo: remove extra '='.
7323 2018-02-06 Marek Polacek <polacek@redhat.com>
7325 PR tree-optimization/84228
7326 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7328 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7331 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7332 emitted arch directives.
7333 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7334 __ARM_FEATURE_COPROC before changing architectures.
7336 2018-02-06 Richard Biener <rguenther@suse.de>
7338 * config/i386/i386.c (print_reg): Fix typo.
7339 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7341 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7343 * configure: Regenerate.
7345 2018-02-05 Martin Sebor <msebor@redhat.com>
7347 PR tree-optimization/83369
7348 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7351 2018-02-05 Martin Liska <mliska@suse.cz>
7353 * doc/invoke.texi: Cherry-pick upstream r323995.
7355 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7357 * ira.c (ira_init_register_move_cost): Adjust comment.
7359 2018-02-05 Martin Liska <mliska@suse.cz>
7361 PR gcov-profile/84137
7362 * doc/gcov.texi: Fix typo in documentation.
7364 2018-02-05 Martin Liska <mliska@suse.cz>
7366 PR gcov-profile/83879
7367 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7368 using dlopen functionality.
7370 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7372 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7373 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7374 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7375 _mm_maskz_range_round_ss): New intrinsics.
7376 (__builtin_ia32_rangesd128_round)
7377 (__builtin_ia32_rangess128_round): Remove.
7378 (__builtin_ia32_rangesd128_mask_round,
7379 __builtin_ia32_rangess128_mask_round): New builtins.
7380 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7381 __builtin_ia32_rangess128_round): Remove.
7382 (__builtin_ia32_rangesd128_mask_round,
7383 __builtin_ia32_rangess128_mask_round): New builtins.
7384 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7385 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7386 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7387 "<round_saeonly_constraint>")): Changed to ...
7388 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7389 "<round_saeonly_scalar_constraint>")): ... this.
7390 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7391 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7392 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7393 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7394 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7396 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7398 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7400 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7401 Remove all values except native, 8540 and 8548.
7403 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7405 * config/i386/i386.c (ix86_output_function_return): Pass
7406 INVALID_REGNUM, instead of -1, as invalid register number to
7407 indirect_thunk_name and output_indirect_thunk.
7409 2018-02-02 Julia Koval <julia.koval@intel.com>
7411 * config.gcc: Add -march=icelake.
7412 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7413 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7414 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7415 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7416 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7417 (processor_target_table): Add icelake.
7418 (ix86_option_override_internal): Handle new PTAs.
7419 (get_builtin_code_for_version): Handle icelake.
7420 (M_INTEL_COREI7_ICELAKE): New.
7421 (fold_builtin_cpu): Handle icelake.
7422 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7423 * doc/invoke.texi: Add -march=icelake.
7425 2018-02-02 Julia Koval <julia.koval@intel.com>
7427 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7428 to wide_int_bitmask.
7429 * wide-int-bitmask.h: New.
7431 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7434 * config/i386/i386.md: Replace Pmode with word_mode in
7435 builtin_setjmp_setup and builtin_longjmp to support x32.
7437 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7441 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7443 (rs6000_supported_cpu_names): New static variable.
7444 (linux_cpu_translation_table): Likewise.
7445 (elf_platform) <cpu>: Define new static variable and use it.
7446 Translate kernel AT_PLATFORM name to canonical name if needed.
7447 Error if platform name is unknown.
7449 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7452 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7454 2018-02-01 Jeff Law <law@redhat.com>
7457 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7458 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7459 the scratch if RELEASE_VIA_POP is false.
7460 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7461 If we have to save a temporary register, decrement SIZE appropriately.
7462 Pass new arguments to release_scratch_register_on_entry.
7463 (ix86_adjust_stack_and_probe): Likewise.
7464 (ix86_emit_probe_stack_range): Pass new arguments to
7465 release_scratch_register_on_entry.
7467 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7469 PR rtl-optimization/84157
7470 * combine.c (change_zero_ext): Use REG_P predicate in
7471 front of HARD_REGISTER_P predicate.
7473 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7475 * config/avr/avr.c (avr_option_override): Move disabling of
7476 -fdelete-null-pointer-checks to...
7477 * common/config/avr/avr-common.c (avr_option_optimization_table):
7480 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7482 PR tree-optimization/81635
7483 * tree-data-ref.c (split_constant_offset_1): For types that
7484 wrap on overflow, try to use range info to prove that wrapping
7487 2018-02-01 Renlin Li <renlin.li@arm.com>
7490 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7492 (aarch64_register_move_cost): Likewise.
7493 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7495 (REG_CLASS_NAMES): Likewise.
7496 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7497 TAILCALL_ADDR_REGS. Remove IP registers.
7498 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7500 2018-02-01 Richard Biener <rguenther@suse.de>
7502 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7503 for specifying RPO order and allow NULL for that.
7504 * domwalk.c (dom_walker::dom_walker): Likewise.
7505 (dom_walker::walk): Handle NULL RPO order.
7506 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7508 (rewrite_update_dom_walker): Likewise.
7509 (mark_def_dom_walker): Likewise.
7511 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7513 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7514 (aarch64_maybe_expand_sve_subreg_move): Declare.
7515 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7516 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7518 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7519 that are semantically a reverse operation.
7520 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7521 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7522 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7524 (aarch64_can_change_mode_class): For big-endian, forbid changes
7525 between two SVE modes if they have different element sizes.
7527 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7529 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7530 the TImode handling for big-endian targets.
7532 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7534 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7535 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7537 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7538 Remove BSWAP handing for big-endian targets and use the form of
7539 LD1RQ appropariate for the mode.
7541 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7543 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7544 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7547 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7550 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7551 check for operands that need to go through aarch64_sve_reload_be.
7553 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7555 PR tree-optimization/81661
7556 PR tree-optimization/84117
7557 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7558 * tree-eh.c: Include gimplify.h.
7559 (find_trapping_overflow, replace_trapping_overflow,
7560 rewrite_to_non_trapping_overflow): New functions.
7561 * tree-vect-loop.c: Include tree-eh.h.
7562 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7563 * tree-data-ref.c: Include tree-eh.h.
7564 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7566 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7568 PR rtl-optimization/84123
7569 * combine.c (change_zero_ext): Check if hard register satisfies
7570 can_change_dest_mode before calling gen_lowpart_SUBREG.
7572 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7575 * ira.c (ira_init_register_move_cost): Remove assert.
7577 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7579 PR rtl-optimization/84071
7580 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7581 * doc/tm.texi: Regenerate.
7583 2018-01-31 Richard Biener <rguenther@suse.de>
7585 PR tree-optimization/84132
7586 * tree-data-ref.c (analyze_miv_subscript): Properly
7587 check whether evolution_function_is_affine_multivariate_p
7588 before calling gcd_of_steps_may_divide_p.
7590 2018-01-31 Julia Koval <julia.koval@intel.com>
7593 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7594 * config/i386/i386.md (rdpid_rex64) New.
7595 (rdpid): Make 32bit only.
7597 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7600 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7601 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7603 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7606 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7608 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7610 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7612 PR rtl-optimization/84071
7613 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7614 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7616 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7618 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7619 (arc_attribute_table): Add 'aux' attribute.
7620 (arc_in_small_data_p): Consider aux like variables.
7621 (arc_is_aux_reg_p): New function.
7622 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7623 (arc_get_aux_arg): New function.
7624 (prepare_move_operands): Handle aux-register access.
7625 (arc_handle_aux_attribute): New function.
7626 * doc/extend.texi (ARC Variable attributes): Add subsection.
7628 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7630 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7631 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7632 (arc_attribute_table): Add 'uncached' attribute.
7633 (arc_print_operand): Print '.di' flag for uncached memory
7635 (arc_in_small_data_p): Do not consider for small data the uncached
7637 (arc_is_uncached_mem_p): New function.
7638 * config/arc/predicates.md (compact_store_memory_operand): Check
7639 for uncached memory accesses.
7640 (nonvol_nonimm_operand): Likewise.
7641 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7643 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7646 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7647 falign-loops=): Add Optimization flag.
7649 2018-01-30 Jeff Law <law@redhat.com>
7652 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7653 INT_REGISTERS_SAVED. Check it prior to calling
7654 get_scratch_register_on_entry.
7655 (ix86_adjust_stack_and_probe): Similarly.
7656 (ix86_emit_probe_stack_range): Similarly.
7657 (ix86_expand_prologue): Corresponding changes.
7659 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7662 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7663 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7665 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7668 * lra-constraints.c (curr_insn_transform): Process AND in the
7671 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7673 PR rtl-optimization/83986
7674 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7675 dependence against last_pending_memory_flush in addition to
7678 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7680 PR tree-optimization/81611
7681 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7684 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7687 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7690 2018-01-30 Richard Biener <rguenther@suse.de>
7691 Jakub Jelinek <jakub@redhat.com>
7693 PR tree-optimization/84111
7694 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7695 inner loops added during recursion, as they don't have up-to-date
7698 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7701 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7702 (can_inline_edge_by_limits_p): ... here.
7703 (can_early_inline_edge_p, check_callers,
7704 update_caller_keys, update_callee_keys, recursive_inlining,
7705 add_new_edges_to_heap, speculation_useful_p,
7706 inline_small_functions,
7707 inline_small_functions, flatten_function,
7708 inline_to_all_callers_1): Update.
7710 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7712 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7715 2018-01-30 Richard Biener <rguenther@suse.de>
7717 PR tree-optimization/83008
7718 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7719 invariant and constant vector uses in stmts when they need
7722 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7725 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7726 * configure: Regenerate.
7728 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7730 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7732 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7733 Use gen_rtx_REG rather than gen_lowpart.
7735 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7737 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7738 rather than 0 when creating partial subregs.
7740 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7742 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7745 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7748 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7749 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7750 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7751 flags. This restores the settings used before the 2017-07-24.
7752 Turning off pre increment/decrement/modify allows IVOPTS to
7753 optimize DF/SF loops where the index is an int.
7755 2018-01-29 Richard Biener <rguenther@suse.de>
7756 Kelvin Nilsen <kelvin@gcc.gnu.org>
7759 * tree-vect-stmts.c (vectorizable_call): Don't call
7760 targetm.vectorize_builtin_md_vectorized_function if callee is
7763 2018-01-22 Carl Love <cel@us.ibm.com>
7765 * doc/extend.tex: Fix typo in second arg in
7766 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7768 2018-01-29 Richard Biener <rguenther@suse.de>
7770 PR tree-optimization/84086
7771 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7772 (flush_ssaname_freelist): When SSA names were released reset
7773 the SCEV hash table.
7775 2018-01-29 Richard Biener <rguenther@suse.de>
7777 PR tree-optimization/84057
7778 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7779 removed paths when removing edges.
7781 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7783 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7784 -mfunction-return=@var{choice}.
7786 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7789 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7791 (layout::print_source_line): Likewise.
7792 (test_get_line_width_without_trailing_whitespace): Add test cases.
7794 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7797 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7800 2018-01-26 Jim Wilson <jimw@sifive.com>
7802 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7804 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7807 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7809 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7810 and CMP + SUB-immediate -> SUBS.
7812 2018-01-26 Martin Sebor <msebor@redhat.com>
7814 PR tree-optimization/83896
7815 * tree-ssa-strlen.c (get_string_len): Rename...
7816 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7817 Avoid assuming length is constant.
7818 (handle_char_store): Use HOST_WIDE_INT for string length.
7820 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7823 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7824 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7826 2018-01-26 Richard Biener <rguenther@suse.de>
7828 PR rtl-optimization/84003
7829 * dse.c (record_store): Only record redundant stores when
7830 the earlier store aliases at least all accesses the later one does.
7832 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7834 PR rtl-optimization/83985
7835 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7836 REG_CFA_RESTORE insns.
7837 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7838 REG_CFA_RESTORE insns here.
7841 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7842 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7844 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7846 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7847 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7848 (arc_init): Likewise.
7849 (arc_override_options): Likewise.
7850 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7852 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7854 * config/arc/arc.h (TARGET_DBNZ): Define.
7855 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7856 properly set the tune attribute.
7857 (dbnz): Use TARGET_DBNZ guard.
7858 * config/arc/arc.opt (mtune): Add core3 option.
7860 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7862 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7863 recognize new pic like addresses.
7864 (arc_delegitimize_address): Clean up.
7866 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7868 * config/arc/arc-arches.def: Option mrf16 valid for all
7870 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7871 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7872 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7873 * config/arc/arc-tables.opt: Regenerate.
7874 * config/arc/arc.c (arc_conditional_register_usage): Handle
7875 reduced register file case.
7876 (arc_file_start): Set must have build attributes.
7877 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7879 * config/arc/arc.opt (mrf16): Add new option.
7880 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7881 * config/arc/genmultilib.awk: Handle new mrf16 option.
7882 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7883 * config/arc/t-multilib: Regenerate.
7884 * doc/invoke.texi (ARC Options): Document mrf16 option.
7886 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7888 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7889 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7890 (arc_attribute_table): Add 'secure_call' attribute.
7891 (arc_print_operand): Print secure call operand.
7892 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7894 (arc_is_secure_call_p): New function. * config/arc/arc.md
7895 (call_i): Add support for sjli instruction.
7896 (call_value_i): Likewise.
7897 * config/arc/constraints.md (Csc): New constraint.
7899 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7900 John Eric Martin <John.Martin@emmicro-us.com>
7902 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7903 * config/arc/arc.c (_arc_jli_section): New struct.
7904 (arc_jli_section): New type.
7905 (rc_jli_sections): New static variable.
7906 (arc_handle_jli_attribute): New function.
7907 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7908 (arc_file_end): New function.
7909 (TARGET_ASM_FILE_END): Define.
7910 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7911 (arc_add_jli_section): New function.
7912 (jli_call_scan): Likewise.
7913 (arc_reorg): Call jli_call_scan.
7914 (arc_output_addsi): Remove 'S' from printing asm operand.
7915 (arc_is_jli_call_p): New function.
7916 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7918 (movhi_insn): Likewise.
7919 (movsi_insn): Likewise.
7920 (movsi_set_cc_insn): Likewise.
7921 (loadqi_update): Likewise.
7922 (load_zeroextendqisi_update): Likewise.
7923 (load_signextendqisi_update): Likewise.
7924 (loadhi_update): Likewise.
7925 (load_zeroextendhisi_update): Likewise.
7926 (load_signextendhisi_update): Likewise.
7927 (loadsi_update): Likewise.
7928 (loadsf_update): Likewise.
7929 (movsicc_insn): Likewise.
7930 (bset_insn): Likewise.
7931 (bxor_insn): Likewise.
7932 (bclr_insn): Likewise.
7933 (bmsk_insn): Likewise.
7934 (bicsi3_insn): Likewise.
7935 (cmpsi_cc_c_insn): Likewise.
7936 (movsi_ne): Likewise.
7937 (movsi_cond_exec): Likewise.
7938 (clrsbsi2): Likewise.
7949 (call_i): Remove 'S' asm letter, add jli instruction.
7950 (call_value_i): Likewise.
7951 * config/arc/arc.op (mjli-always): New option.
7952 * config/arc/constraints.md (Cji): New constraint.
7953 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7955 (subsf3_fpx): Likewise.
7956 (mulsf3_fpx): Likewise.
7957 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7959 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7961 * doc/invoke.texi (ARC): Document mjli-always option.
7963 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7965 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7966 avoid addition with 0 and use incw and decw where possible.
7968 2018-01-26 Richard Biener <rguenther@suse.de>
7970 PR tree-optimization/81082
7971 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7972 association if it requires casting to unsigned.
7973 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7974 from fold_plusminus_mult_expr to catch important cases late when
7975 range info is available.
7977 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7979 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7980 * configure.ac (hidden_linkonce): New test.
7981 * configure: Regenerate.
7982 * config.in: Regenerate.
7984 2018-01-26 Julia Koval <julia.koval@intel.com>
7986 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7987 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7988 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7989 _mm_mask_bitshuffle_epi64_mask): Fix type.
7990 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7991 USI_FTYPE_V4DI_V4DI_USI): Remove.
7992 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7993 __builtin_ia32_vpshufbitqmb256_mask,
7994 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7995 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7996 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7998 2018-01-26 Alan Modra <amodra@gmail.com>
8001 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8002 UNSPEC_VBPERMQ. Sort other unspecs.
8004 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8006 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8008 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8011 * predict.c (drop_profile): Do not push/pop cfun; update also
8013 (handle_missing_profiles): Fix logic looking for zero profiles.
8015 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8018 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8019 on functions with #pragma omp declare simd or functions with simd
8021 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8022 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8023 Remove trailing \n from warning_at calls.
8025 2018-01-25 Tom de Vries <tom@codesourcery.com>
8028 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8029 for neutered workers.
8031 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8034 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8035 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8037 2018-01-24 Jeff Law <law@redhat.com>
8040 * i386.c (get_probe_interval): Move to earlier point.
8041 (ix86_compute_frame_layout): If -fstack-clash-protection and
8042 the frame is larger than the probe interval, then use pushes
8043 to save registers rather than reg->mem moves.
8044 (ix86_expand_prologue): Remove conditional for int_registers_saved
8047 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8050 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8051 min/max for never referenced object.
8053 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8056 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8058 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8059 attributes from DECL_ATTRIBUTES (decl) without affecting
8060 DECL_ATTRIBUTES (current_function_decl).
8061 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8062 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8064 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8066 PR tree-optimization/83979
8067 * fold-const.c (fold_comparison): Use constant_boolean_node
8068 instead of boolean_{true,false}_node.
8070 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8072 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8075 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8077 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8078 Simplify the clause that sets the length attribute.
8079 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8080 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8081 clause that sets the length attribute.
8082 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8084 2018-01-24 Tom de Vries <tom@codesourcery.com>
8087 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8088 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8089 Add strict parameter.
8090 (prevent_branch_around_nothing): Insert dummy insn between branch to
8091 label and label with no ptx insn inbetween.
8092 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8094 2018-01-24 Tom de Vries <tom@codesourcery.com>
8097 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8098 for neutered threads in warp.
8099 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8101 2018-01-24 Richard Biener <rguenther@suse.de>
8103 PR tree-optimization/83176
8104 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8107 2018-01-24 Richard Biener <rguenther@suse.de>
8109 PR tree-optimization/82819
8110 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8111 code generating pluses that are no-ops in the target precision.
8113 2018-01-24 Richard Biener <rguenther@suse.de>
8116 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8118 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8120 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8121 to merge probabilities.
8122 * predict.c (probably_never_executed): Also mark as cold functions
8123 with global 0 profile and guessed local profile.
8124 * profile-count.c (profile_probability::combine_with_count): New
8126 * profile-count.h (profile_probability::operator*,
8127 profile_probability::operator*=, profile_probability::operator/,
8128 profile_probability::operator/=): Reduce precision to adjusted
8129 and set value to guessed on contradictory divisions.
8130 (profile_probability::combine_with_freq): Remove.
8131 (profile_probability::combine_wiht_count): Declare.
8132 (profile_count::force_nonzero):: Set to adjusted.
8133 (profile_count::probability_in):: Set quality to adjusted.
8134 * tree-ssa-tail-merge.c (replace_block_by): Use
8137 2018-01-23 Andrew Waterman <andrew@sifive.com>
8138 Jim Wilson <jimw@sifive.com>
8140 * config/riscv/riscv.c (riscv_stack_boundary): New.
8141 (riscv_option_override): Set riscv_stack_boundary. Handle
8142 riscv_preferred_stack_boundary_arg.
8143 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8144 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8145 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8146 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8147 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8148 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8150 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8153 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8154 of struct ix86_frame.
8155 (ix86_expand_epilogue): Likewise. Add a local variable for
8156 the reg_save_offset field in struct ix86_frame.
8158 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8160 PR tree-optimization/82604
8161 * tree-loop-distribution.c (enum partition_kind): New enum item
8162 PKIND_PARTIAL_MEMSET.
8163 (partition_builtin_p): Support above new enum item.
8164 (generate_code_for_partition): Ditto.
8165 (compute_access_range): Differentiate cases that equality can be
8166 proven at all loops, the innermost loops or no loops.
8167 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8168 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8169 (finalize_partitions, distribute_loop): Don't fuse partition of
8170 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8171 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8174 2018-01-23 Martin Liska <mliska@suse.cz>
8176 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8177 order to ignore the predictor.
8178 (PRED_POLYMORPHIC_CALL): Likewise.
8179 (PRED_RECURSIVE_CALL): Likewise.
8181 2018-01-23 Martin Liska <mliska@suse.cz>
8183 * tree-profile.c (tree_profiling): Print function header to
8184 aware reader which function we are working on.
8185 * value-prof.c (gimple_find_values_to_profile): Do not print
8186 not interesting value histograms.
8188 2018-01-23 Martin Liska <mliska@suse.cz>
8190 * profile-count.h (enum profile_quality): Add
8191 profile_uninitialized as the first value. Do not number values
8192 as they are zero based.
8193 (profile_count::verify): Update sanity check.
8194 (profile_probability::verify): Likewise.
8196 2018-01-23 Nathan Sidwell <nathan@acm.org>
8198 * doc/invoke.texi (ffor-scope): Deprecate.
8200 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8202 PR tree-optimization/83510
8203 * domwalk.c (set_all_edges_as_executable): New function.
8204 (dom_walker::dom_walker): Convert bool param
8205 "skip_unreachable_blocks" to enum reachability. Move setup of
8206 edge flags to set_all_edges_as_executable and only do it when
8207 reachability is REACHABLE_BLOCKS.
8208 * domwalk.h (enum dom_walker::reachability): New enum.
8209 (dom_walker::dom_walker): Convert bool param
8210 "skip_unreachable_blocks" to enum reachability.
8211 (set_all_edges_as_executable): New decl.
8212 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8213 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8215 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8216 but converting true to REACHABLE_BLOCKS.
8217 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8219 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8220 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8221 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8223 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8224 if check_all_array_refs will be called.
8226 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8228 * tree.c (selftest::test_location_wrappers): Add more test
8231 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8233 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8234 (selftest::test_bit_in_range): Likewise.
8236 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8239 * doc/sourcebuild.texi (vect_float): Say that the selector
8240 only describes the situation when -funsafe-math-optimizations is on.
8241 (vect_float_strict): Document.
8243 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8245 PR tree-optimization/83965
8246 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8247 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8248 instead of checking only for a reduction.
8249 (vect_recog_widen_sum_pattern): Likewise.
8251 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8253 * predict.c (probably_never_executed): Only use precise profile info.
8254 (compute_function_frequency): Skip after inlining hack since we now
8255 have quality checking.
8257 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8259 * profile-count.h (profile_probability::very_unlikely,
8260 profile_probability::unlikely, profile_probability::even): Set
8261 precision to guessed.
8263 2018-01-23 Richard Biener <rguenther@suse.de>
8265 PR tree-optimization/83963
8266 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8267 Properly terminate dominator walk when crossing the exit edge not
8268 when visiting its source block.
8270 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8273 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8274 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8276 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8278 PR tree-optimization/83957
8279 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8280 semicolon after for body surrounded by braces.
8282 PR tree-optimization/83081
8283 * profile-count.h (profile_probability::split): New method.
8284 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8285 Use profile_probability::split.
8286 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8287 when splitting a single conditional jump into 2.
8289 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8291 PR tree-optimization/69452
8292 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8295 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8297 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8298 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8299 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8301 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8303 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8304 * config/rl78/rl78.md (movdi): New define_expand.
8305 * config/rl78/rl78.c (rl78_split_movdi): New function.
8307 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8310 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8312 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8313 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8314 128-bit to produce an UNSPEC move to get the double word with the
8315 signbit and then a shift directly to do signbit.
8316 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8317 implementation with a new version that just does either a direct
8318 move or a regular move. Move memory interface to separate insns.
8319 Move insns so they are next to the expander.
8320 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8321 with signbit move. Split big and little endian case.
8322 (signbit<mode>2_dm_mem_le): Likewise.
8323 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8324 (signbit<mode>2_dm2): Likewise.
8326 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8328 * config/rl78/rl78.md (anddi3): New define_expand.
8330 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8332 * config/rl78/rl78.md (umindi3): New define_expand.
8334 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8336 * config/rl78/rl78.md (smindi3): New define_expand.
8338 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8340 * config/rl78/rl78.md (smaxdi3): New define_expand.
8342 2018-01-22 Carl Love <cel@us.ibm.com>
8344 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8345 LVX_V1TI): Add macro expansion.
8346 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8347 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8348 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8349 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8350 Change check to determine if the instruction is a byte reversing
8351 entry. Fix typo in comment.
8352 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8353 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8354 Add def_builtin calls for new builtins.
8355 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8356 Add define_insn expansion.
8358 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8360 * config/rl78/rl78.md (umaxdi3): New define_expand.
8362 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8364 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8365 for non-QImode registers.
8367 2018-01-22 Richard Biener <rguenther@suse.de>
8369 PR tree-optimization/83963
8370 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8371 including the loop exit block.
8372 (scop_detection::merge_sese): Likewise.
8373 (scop_detection::add_scop): Do it here instead.
8375 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8377 * doc/sourcebuild.texi (arm_softfloat): Document.
8379 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8382 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8383 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8384 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8386 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8387 David Edelsohn <dje.gcc@gmail.com>
8390 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8391 Change "crset eq" to "crset 2".
8392 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8393 (*call_indirect_aix<mode>_nospec): Likewise.
8394 (*call_value_indirect_aix<mode>_nospec): Likewise.
8395 (*call_indirect_elfv2<mode>_nospec): Likewise.
8396 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8397 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8398 change assembly output from . to $.
8399 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8400 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8401 (*tablejump<mode>_internal1_nospec): Likewise.
8403 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8406 * config/sh/sh_optimize_sett_clrt.cc:
8407 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8409 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8411 PR tree-optimization/83940
8412 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8413 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8414 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8415 use it to pass back the definition type.
8416 (vect_check_store_rhs): Likewise rhs_dt_out.
8417 (vect_build_gather_load_calls): Add a mask_dt argument and use
8418 it instead of a call to vect_is_simple_use.
8419 (vectorizable_store): Update calls to vect_check_load_store_mask
8420 and vect_check_store_rhs. Use the dt returned by the latter instead
8421 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8422 instead of calls to vect_is_simple_use. Pass the scalar rather
8423 than the vector operand to vect_is_simple_use when handling
8424 second and subsequent copies of an rhs value.
8425 (vectorizable_load): Update calls to vect_check_load_store_mask
8426 and vect_build_gather_load_calls. Use the cached mask_dt and
8427 gs_info.offset_dt instead of calls to vect_is_simple_use.
8429 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8432 * tree-emutls.c: Include gimplify.h.
8433 (lower_emutls_2): New function.
8434 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8435 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8436 it before further processing.
8439 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8440 UINTVAL (trueop1) instead of INTVAL (op1).
8442 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8446 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8447 INCOMING_FRAME_SP_OFFSET if not defined.
8448 (scan_trace): Add ENTRY argument. If true and
8449 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8450 emit a note to adjust the CFA offset.
8451 (create_cfi_notes): Adjust scan_trace callers.
8452 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8453 INCOMING_FRAME_SP_OFFSET in the CIE.
8454 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8455 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8457 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8458 * doc/tm.texi: Regenerated.
8460 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8462 PR rtl-optimization/83147
8463 * lra-constraints.c (remove_inheritance_pseudos): Use
8464 lra_substitute_pseudo_within_insn.
8466 2018-01-19 Tom de Vries <tom@codesourcery.com>
8467 Cesar Philippidis <cesar@codesourcery.com>
8470 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8472 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8475 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8476 spaces for function labels.
8478 2018-01-19 Martin Liska <mliska@suse.cz>
8480 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8481 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8482 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8483 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8484 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8485 (PRED_CONST_RETURN): Change from 69 to 65.
8486 (PRED_NULL_RETURN): Change from 91 to 71.
8487 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8488 (PRED_LOOP_GUARD): Change from 66 to 73.
8490 2018-01-19 Martin Liska <mliska@suse.cz>
8492 * predict.c (predict_insn_def): Add new assert.
8493 (struct branch_predictor): Change type to signed integer.
8494 (test_prediction_value_range): Amend test to cover
8496 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8497 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8498 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8499 (PRED_LOOP_IV_COMPARE): Likewise.
8500 * predict.h (PROB_UNINITIALIZED): Define new constant.
8502 2018-01-19 Martin Liska <mliska@suse.cz>
8504 * predict.c (dump_prediction): Add new format for
8505 analyze_brprob.py script which is enabled with -details
8507 * profile-count.h (precise_p): New function.
8509 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8511 PR tree-optimization/83922
8512 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8513 there are no statements that need masking.
8514 (vect_active_double_reduction_p): New function.
8515 (vect_analyze_loop_operations): Use it when handling phis that
8516 are not in the loop header.
8518 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8520 PR tree-optimization/83914
8521 * tree-vect-loop.c (vectorizable_induction): Don't convert
8522 init_expr or apply the peeling adjustment for inductions
8523 that are nested within the vectorized loop.
8525 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8527 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8530 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8534 * function.h (gimplify_parameters): Add gimple_seq * argument.
8535 * function.c: Include gimple.h and options.h.
8536 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8537 for the added local temporaries if needed.
8538 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8539 if there are any parameter cleanups, wrap whole body into a
8540 try/finally with the cleanups.
8542 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8545 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8546 Use GET_MODE_CLASS for scalar floating point.
8548 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8552 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8553 Fix call of call_cgraph_insertion_hooks.
8555 2018-01-18 Martin Sebor <msebor@redhat.com>
8557 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8559 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8562 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8565 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8568 * common.opt: (-ffile-prefix-map): New option.
8569 * opts.c (common_handle_option): Defer it.
8570 * opts-global.c (handle_common_deferred_options): Handle it.
8571 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8572 * file-prefix-map.h: New file.
8573 (remap_debug_filename, add_debug_prefix_map): ...here.
8574 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8575 * final.c (debug_prefix_map, add_debug_prefix_map
8576 remap_debug_filename): Move to...
8577 * file-prefix-map.c: New file.
8578 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8579 generalize, get rid of alloca(), use strrchr() instead of strchr().
8580 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8581 Implement in terms of add_prefix_map().
8582 (remap_macro_filename, remap_debug_filename): Implement in term of
8584 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8585 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8586 * dbxout.c: Include file-prefix-map.h.
8587 * varasm.c: Likewise.
8588 * vmsdbgout.c: Likewise.
8589 * xcoffout.c: Likewise.
8590 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8591 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8592 * doc/invoke.texi (-ffile-prefix-map): Document.
8593 (-fdebug-prefix-map): Update description.
8595 2018-01-18 Martin Liska <mliska@suse.cz>
8597 * config/i386/i386.c (indirect_thunk_name): Document that also
8599 (output_indirect_thunk): Document why both instructions
8600 (pause and lfence) are generated.
8602 2018-01-18 Richard Biener <rguenther@suse.de>
8604 PR tree-optimization/83887
8605 * graphite-scop-detection.c
8606 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8607 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8608 (scop_detection::merge_sese): Re-implement with a flood-fill
8609 algorithm that properly finds a SESE region if it exists.
8611 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8614 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8615 pointer_diff optimizations use view_convert instead of convert.
8617 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8619 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8620 Generate different code for -mno-speculate-indirect-jumps.
8621 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8622 (*call_indirect_aix<mode>): Disable for
8623 -mno-speculate-indirect-jumps.
8624 (*call_indirect_aix<mode>_nospec): New define_insn.
8625 (*call_value_indirect_aix<mode>): Disable for
8626 -mno-speculate-indirect-jumps.
8627 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8628 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8629 -mno-speculate-indirect-jumps.
8630 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8632 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8634 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8635 long double type, set the flags for noting the default long double
8636 type, even if we don't pass or return a long double type.
8638 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8641 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8644 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8646 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8647 support for merge[hl].
8648 (fold_mergehl_helper): New helper function.
8649 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8650 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8651 (altivec_vmrglw_direct): Add xxmrglw insn.
8653 2018-01-17 Andrew Waterman <andrew@sifive.com>
8655 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8656 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8658 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8661 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8662 call the lto_location_cache before reading the
8663 DECL_SOURCE_LOCATION of the types.
8665 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8666 Richard Sandiford <richard.sandiford@linaro.org>
8668 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8669 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8670 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8671 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8672 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8674 * config/aarch64/constraints.md (aarch64_movti_operand):
8676 * config/aarch64/predicates.md (Uti): Add new constraint.
8678 2018-01-17 Carl Love <cel@us.ibm.com>
8680 * config/rs6000/vsx.md (define_expand xl_len_r,
8681 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8682 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8684 (define_expand, define_insn): Move the shift left from the
8685 define_insn to the define_expand for lxvl and stxvl instructions.
8686 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8687 and XL_LEN_R definitions to PURE.
8689 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8691 * config/i386/i386.c (indirect_thunk_name): Declare regno
8692 as unsigned int. Compare regno with INVALID_REGNUM.
8693 (output_indirect_thunk): Ditto.
8694 (output_indirect_thunk_function): Ditto.
8695 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8696 in the call to output_indirect_thunk_function.
8698 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8701 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8702 rather than the size of inner_type to determine the stack slot size
8703 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8705 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8708 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8711 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8713 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8714 endian Linux systems to optionally enable multilibs for selecting
8715 the long double type if the user configured an explicit type.
8716 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8717 have no long double multilibs if not defined.
8718 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8719 warn if the user used -mabi={ieee,ibm}longdouble and we built
8720 multilibs for long double.
8721 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8722 appropriate multilib option.
8723 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8725 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8726 for building long double multilibs.
8727 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8729 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8731 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8734 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8736 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8739 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8742 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8745 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8747 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8748 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8750 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8752 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8753 different rtl trees depending on TARGET_64BIT.
8754 (rs6000_gen_lvx): Likewise.
8756 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8758 * config/visium/visium.md (nop): Tweak comment.
8759 (hazard_nop): Likewise.
8761 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8763 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8764 -mspeculate-indirect-jumps.
8765 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8766 for -mno-speculate-indirect-jumps.
8767 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8768 (*call_value_indirect_elfv2<mode>): Disable for
8769 -mno-speculate-indirect-jumps.
8770 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8771 (indirect_jump): Emit different RTL for
8772 -mno-speculate-indirect-jumps.
8773 (*indirect_jump<mode>): Disable for
8774 -mno-speculate-indirect-jumps.
8775 (*indirect_jump<mode>_nospec): New define_insn.
8776 (tablejump): Emit different RTL for
8777 -mno-speculate-indirect-jumps.
8778 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8779 (tablejumpsi_nospec): New define_expand.
8780 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8781 (tablejumpdi_nospec): New define_expand.
8782 (*tablejump<mode>_internal1): Disable for
8783 -mno-speculate-indirect-jumps.
8784 (*tablejump<mode>_internal1_nospec): New define_insn.
8785 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8788 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8790 * caller-save.c (insert_save): Drop unnecessary parameter. All
8793 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8794 Richard Biener <rguenth@suse.de>
8797 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8798 return early, inline manually is_gimple_sizepos. Make sure if we
8799 call gimplify_expr we don't end up with a gimple constant.
8800 * tree.c (variably_modified_type_p): Don't return true for
8801 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8802 * gimplify.h (is_gimple_sizepos): Remove.
8804 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8806 PR tree-optimization/83857
8807 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8808 vectorizable_live_operation for pure SLP statements.
8809 (vectorizable_live_operation): Handle PHIs.
8811 2018-01-16 Richard Biener <rguenther@suse.de>
8813 PR tree-optimization/83867
8814 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8815 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8817 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8820 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8821 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8822 If off is not INTEGER_CST, issue a may not be aligned warning
8823 rather than isn't aligned. Use isn%'t rather than isn't.
8824 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8826 <case MULT_EXPR>: Improve the case when bottom and one of the
8827 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8828 operand, in that case check if the other operand is multiple of
8829 bottom divided by the INTEGER_CST operand.
8831 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8834 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8835 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8836 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8837 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8838 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8839 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8840 (pa_function_arg_size): New function.
8842 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8844 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8845 in a separate statement.
8847 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8849 PR tree-optimization/83847
8850 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8851 group gathers and scatters.
8853 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8855 PR rtl-optimization/86620
8856 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8858 PR rtl-optimization/83213
8859 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8860 to last if both are JUMP_INSNs.
8862 PR tree-optimization/83843
8863 * gimple-ssa-store-merging.c
8864 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8865 store_immediate_info for bswap/nop orig_stores.
8867 2018-01-15 Andrew Waterman <andrew@sifive.com>
8869 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8871 <UDIV>: Increase cost if !TARGET_DIV.
8873 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8875 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8876 (define_attr "cr_logical_3op"): New.
8877 (cceq_ior_compare): Adjust.
8878 (cceq_ior_compare_complement): Adjust.
8879 (*cceq_rev_compare): Adjust.
8880 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8881 (is_cracked_insn): Adjust.
8882 (insn_must_be_first_in_group): Adjust.
8883 * config/rs6000/40x.md: Adjust.
8884 * config/rs6000/440.md: Adjust.
8885 * config/rs6000/476.md: Adjust.
8886 * config/rs6000/601.md: Adjust.
8887 * config/rs6000/603.md: Adjust.
8888 * config/rs6000/6xx.md: Adjust.
8889 * config/rs6000/7450.md: Adjust.
8890 * config/rs6000/7xx.md: Adjust.
8891 * config/rs6000/8540.md: Adjust.
8892 * config/rs6000/cell.md: Adjust.
8893 * config/rs6000/e300c2c3.md: Adjust.
8894 * config/rs6000/e500mc.md: Adjust.
8895 * config/rs6000/e500mc64.md: Adjust.
8896 * config/rs6000/e5500.md: Adjust.
8897 * config/rs6000/e6500.md: Adjust.
8898 * config/rs6000/mpc.md: Adjust.
8899 * config/rs6000/power4.md: Adjust.
8900 * config/rs6000/power5.md: Adjust.
8901 * config/rs6000/power6.md: Adjust.
8902 * config/rs6000/power7.md: Adjust.
8903 * config/rs6000/power8.md: Adjust.
8904 * config/rs6000/power9.md: Adjust.
8905 * config/rs6000/rs64.md: Adjust.
8906 * config/rs6000/titan.md: Adjust.
8908 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8910 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8911 ix86_indirect_branch_register logic.
8913 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8915 * config/i386/constraints.md (Bs): Update
8916 ix86_indirect_branch_register check. Don't check
8917 ix86_indirect_branch_register with GOT_memory_operand.
8919 * config/i386/predicates.md (GOT_memory_operand): Don't check
8920 ix86_indirect_branch_register here.
8921 (GOT32_symbol_operand): Likewise.
8923 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8925 * config/i386/predicates.md (constant_call_address_operand):
8926 Rewrite ix86_indirect_branch_register logic.
8927 (sibcall_insn_operand): Likewise.
8929 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8931 * config/i386/constraints.md (Bs): Replace
8932 ix86_indirect_branch_thunk_register with
8933 ix86_indirect_branch_register.
8935 * config/i386/i386.md (indirect_jump): Likewise.
8936 (tablejump): Likewise.
8937 (*sibcall_memory): Likewise.
8938 (*sibcall_value_memory): Likewise.
8939 Peepholes of indirect call and jump via memory: Likewise.
8940 * config/i386/i386.opt: Likewise.
8941 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8942 (GOT_memory_operand): Likewise.
8943 (call_insn_operand): Likewise.
8944 (sibcall_insn_operand): Likewise.
8945 (GOT32_symbol_operand): Likewise.
8947 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8950 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8951 type rather than type addr's type points to.
8952 (expand_omp_atomic_mutex): Likewise.
8953 (expand_omp_atomic): Likewise.
8955 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8958 * config/i386/i386.c (output_indirect_thunk_function): Use
8959 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8960 for __x86_return_thunk.
8962 2018-01-15 Richard Biener <rguenther@suse.de>
8965 * expmed.c (extract_bit_field_1): Fix typo.
8967 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8970 * config/arm/iterators.md (VF): New mode iterator.
8971 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8972 Remove integer-related logic from pattern.
8973 (neon_vabd<mode>_3): Likewise.
8975 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8978 * common.opt (fstrict-overflow): No longer an alias.
8979 (fwrapv-pointer): New option.
8980 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8981 also for pointer types based on flag_wrapv_pointer.
8982 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8983 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8984 opts->x_flag_wrapv got set.
8985 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8986 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8987 POINTER_TYPE_OVERFLOW_UNDEFINED.
8988 * match.pd: Likewise in address comparison pattern.
8989 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8991 2018-01-15 Richard Biener <rguenther@suse.de>
8994 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8995 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8996 Reset type names to their identifier if their TYPE_DECL doesn't
8997 have linkage (and thus is used for ODR and devirt).
8998 (save_debug_info_for_decl): Remove.
8999 (save_debug_info_for_type): Likewise.
9000 (add_tree_to_fld_list): Adjust.
9001 * tree-pretty-print.c (dump_generic_node): Make dumping of
9002 type names more robust.
9004 2018-01-15 Richard Biener <rguenther@suse.de>
9006 * BASE-VER: Bump to 8.0.1.
9008 2018-01-14 Martin Sebor <msebor@redhat.com>
9011 * builtins.c (check_access): Avoid warning when the no-warning bit
9014 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9016 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9017 * ira-color (allocno_hard_regs_compare): Likewise.
9019 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9022 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9023 Use .pushsection/.popsection.
9025 2018-01-14 Martin Sebor <msebor@redhat.com>
9028 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9030 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9032 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9033 entry from extra_headers.
9034 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9035 extra_headers, make the list bitwise identical to the i?86-*-* one.
9037 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9039 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9040 -mcmodel=large with -mindirect-branch=thunk,
9041 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9042 -mfunction-return=thunk-extern.
9043 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9044 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9045 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9047 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9049 * config/i386/i386.c (print_reg): Print the name of the full
9050 integer register without '%'.
9051 (ix86_print_operand): Handle 'V'.
9052 * doc/extend.texi: Document 'V' modifier.
9054 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9056 * config/i386/constraints.md (Bs): Disallow memory operand for
9057 -mindirect-branch-register.
9059 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9060 (GOT_memory_operand): Likewise.
9061 (call_insn_operand): Likewise.
9062 (sibcall_insn_operand): Likewise.
9063 (GOT32_symbol_operand): Likewise.
9064 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9065 for -mindirect-branch-register.
9066 (tablejump): Likewise.
9067 (*sibcall_memory): Likewise.
9068 (*sibcall_value_memory): Likewise.
9069 Disallow peepholes of indirect call and jump via memory for
9070 -mindirect-branch-register.
9071 (*call_pop): Replace m with Bw.
9072 (*call_value_pop): Likewise.
9073 (*sibcall_pop_memory): Replace m with Bs.
9074 * config/i386/i386.opt (mindirect-branch-register): New option.
9075 * doc/invoke.texi: Document -mindirect-branch-register option.
9077 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9079 * config/i386/i386-protos.h (ix86_output_function_return): New.
9080 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9081 set function_return_type.
9082 (indirect_thunk_name): Add ret_p to indicate thunk for function
9084 (output_indirect_thunk_function): Pass false to
9085 indirect_thunk_name.
9086 (ix86_output_indirect_branch_via_reg): Likewise.
9087 (ix86_output_indirect_branch_via_push): Likewise.
9088 (output_indirect_thunk_function): Create alias for function
9089 return thunk if regno < 0.
9090 (ix86_output_function_return): New function.
9091 (ix86_handle_fndecl_attribute): Handle function_return.
9092 (ix86_attribute_table): Add function_return.
9093 * config/i386/i386.h (machine_function): Add
9094 function_return_type.
9095 * config/i386/i386.md (simple_return_internal): Use
9096 ix86_output_function_return.
9097 (simple_return_internal_long): Likewise.
9098 * config/i386/i386.opt (mfunction-return=): New option.
9099 (indirect_branch): Mention -mfunction-return=.
9100 * doc/extend.texi: Document function_return function attribute.
9101 * doc/invoke.texi: Document -mfunction-return= option.
9103 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9105 * config/i386/i386-opts.h (indirect_branch): New.
9106 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9107 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9108 with local indirect jump when converting indirect call and jump.
9109 (ix86_set_indirect_branch_type): New.
9110 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9111 (indirectlabelno): New.
9112 (indirect_thunk_needed): Likewise.
9113 (indirect_thunk_bnd_needed): Likewise.
9114 (indirect_thunks_used): Likewise.
9115 (indirect_thunks_bnd_used): Likewise.
9116 (INDIRECT_LABEL): Likewise.
9117 (indirect_thunk_name): Likewise.
9118 (output_indirect_thunk): Likewise.
9119 (output_indirect_thunk_function): Likewise.
9120 (ix86_output_indirect_branch_via_reg): Likewise.
9121 (ix86_output_indirect_branch_via_push): Likewise.
9122 (ix86_output_indirect_branch): Likewise.
9123 (ix86_output_indirect_jmp): Likewise.
9124 (ix86_code_end): Call output_indirect_thunk_function if needed.
9125 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9127 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9128 (ix86_attribute_table): Add indirect_branch.
9129 * config/i386/i386.h (machine_function): Add indirect_branch_type
9130 and has_local_indirect_jump.
9131 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9133 (tablejump): Likewise.
9134 (*indirect_jump): Use ix86_output_indirect_jmp.
9135 (*tablejump_1): Likewise.
9136 (simple_return_indirect_internal): Likewise.
9137 * config/i386/i386.opt (mindirect-branch=): New option.
9138 (indirect_branch): New.
9141 (thunk-inline): Likewise.
9142 (thunk-extern): Likewise.
9143 * doc/extend.texi: Document indirect_branch function attribute.
9144 * doc/invoke.texi: Document -mindirect-branch= option.
9146 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9149 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9151 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9153 * ipa-inline.c (want_inline_small_function_p): Return false if
9154 inlining has already failed with CIF_FINAL_ERROR.
9155 (update_caller_keys): Call want_inline_small_function_p before
9157 (update_callee_keys): Likewise.
9159 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9161 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9163 (rs6000_quadword_masked_address_p): Likewise.
9164 (quad_aligned_load_p): Likewise.
9165 (quad_aligned_store_p): Likewise.
9166 (const_load_sequence_p): Add comment to describe the outer-most loop.
9167 (mimic_memory_attributes_and_flags): New function.
9168 (rs6000_gen_stvx): Likewise.
9169 (replace_swapped_aligned_store): Likewise.
9170 (rs6000_gen_lvx): Likewise.
9171 (replace_swapped_aligned_load): Likewise.
9172 (replace_swapped_load_constant): Capitalize argument name in
9173 comment describing this function.
9174 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9175 and stores that access quad-word aligned addresses and replace
9176 with stvx or lvx instructions when appropriate.
9177 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9178 New function prototype.
9179 (rs6000_quadword_masked_address_p): Likewise.
9180 (rs6000_gen_lvx): Likewise.
9181 (rs6000_gen_stvx): Likewise.
9182 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9183 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9184 when memory address is aligned.
9185 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9186 this split to select lvx instruction when memory address is aligned.
9187 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9188 instruction when memory address is aligned.
9189 (*vsx_le_perm_load_v16qi): Likewise.
9190 (four unnamed splitters): Modify to select the stvx instruction
9191 when memory is aligned.
9193 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9195 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9196 which appears in the queue multiple times.
9198 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9199 Alan Hayward <alan.hayward@arm.com>
9200 David Sherwood <david.sherwood@arm.com>
9202 * tree-vectorizer.h (vec_lower_bound): New structure.
9203 (_loop_vec_info): Add check_nonzero and lower_bounds.
9204 (LOOP_VINFO_CHECK_NONZERO): New macro.
9205 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9206 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9207 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9208 fields. Make seg_len the distance travelled, not including the
9210 (dr_direction_indicator): Declare.
9211 (dr_zero_step_indicator): Likewise.
9212 (dr_known_forward_stride_p): Likewise.
9213 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9215 (runtime_alias_check_p): Allow runtime alias checks with
9217 (operator ==): Compare access_size and align.
9218 (prune_runtime_alias_test_list): Rework for new distinction between
9219 the access_size and seg_len.
9220 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9222 (get_segment_min_max): New function.
9223 (create_intersect_range_checks): Use it.
9224 (dr_step_indicator): New function.
9225 (dr_direction_indicator): Likewise.
9226 (dr_zero_step_indicator): Likewise.
9227 (dr_known_forward_stride_p): Likewise.
9228 * tree-loop-distribution.c (data_ref_segment_size): Return
9229 DR_STEP * (niters - 1).
9230 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9232 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9233 (vect_preserves_scalar_order_p): New function, split out from...
9234 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9235 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9236 (vect_vfa_access_size): New function.
9237 (vect_vfa_align): Likewise.
9238 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9239 (dump_lower_bound): New function.
9240 (vect_check_lower_bound): Likewise.
9241 (vect_small_gap_p): Likewise.
9242 (vectorizable_with_step_bound_p): Likewise.
9243 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9244 depencies if the vectorization factor is 1. Convert the checks
9245 for nonzero steps into checks on the bounds of DR_STEP. Try using
9246 a bunds check for variable steps if the minimum required step is
9247 relatively small. Update calls to the dr_with_seg_len
9248 constructor and to vect_compile_time_alias.
9249 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9251 (vect_loop_versioning): Call it.
9252 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9254 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9256 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9257 Alan Hayward <alan.hayward@arm.com>
9258 David Sherwood <david.sherwood@arm.com>
9260 * doc/sourcebuild.texi (vect_scatter_store): Document.
9261 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9263 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9265 * genopinit.c (main): Add supports_vec_scatter_store and
9266 supports_vec_scatter_store_cached to target_optabs.
9267 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9268 IFN_MASK_SCATTER_STORE.
9269 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9271 * internal-fn.h (internal_store_fn_p): Declare.
9272 (internal_fn_stored_value_index): Likewise.
9273 * internal-fn.c (scatter_store_direct): New macro.
9274 (expand_scatter_store_optab_fn): New function.
9275 (direct_scatter_store_optab_supported_p): New macro.
9276 (internal_store_fn_p): New function.
9277 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9278 IFN_MASK_SCATTER_STORE.
9279 (internal_fn_mask_index): Likewise.
9280 (internal_fn_stored_value_index): New function.
9281 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9283 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9284 * optabs-query.c (supports_vec_scatter_store_p): New function.
9285 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9286 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9287 true for scatter stores.
9288 (vect_gather_scatter_fn_p): Handle scatter stores too.
9289 (vect_check_gather_scatter): Consider using scatter stores if
9290 supports_vec_scatter_store_p.
9291 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9293 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9294 internal_fn_stored_value_index.
9295 (check_load_store_masking): Handle scatter stores too.
9296 (vect_get_store_rhs): Make public.
9297 (vectorizable_call): Use internal_store_fn_p.
9298 (vectorizable_store): Handle scatter store internal functions.
9299 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9300 when deciding whether the end of the group has been reached.
9301 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9302 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9303 (mask_scatter_store<mode>): New insns.
9305 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9306 Alan Hayward <alan.hayward@arm.com>
9307 David Sherwood <david.sherwood@arm.com>
9309 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9310 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9311 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9313 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9314 Use vect_truncate_gather_scatter_offset if we can't treat the
9315 operation as a normal gather load or scatter store.
9316 (get_group_load_store_type): Take the gather_scatter_info
9317 as argument. Try using a gather load or scatter store for
9318 single-element groups.
9319 (get_load_store_type): Update calls to get_group_load_store_type
9320 and vect_use_strided_gather_scatters_p.
9322 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9323 Alan Hayward <alan.hayward@arm.com>
9324 David Sherwood <david.sherwood@arm.com>
9326 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9327 optional tree argument.
9328 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9330 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9331 but continue to use the current value as a fallback.
9332 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9333 to compare the updates.
9334 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9335 (get_load_store_type): Use it when handling a strided access.
9336 (vect_get_strided_load_store_ops): New function.
9337 (vect_get_data_ptr_increment): Likewise.
9338 (vectorizable_load): Handle strided gather loads. Always pass
9339 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9341 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9342 Alan Hayward <alan.hayward@arm.com>
9343 David Sherwood <david.sherwood@arm.com>
9345 * doc/md.texi (gather_load@var{m}): Document.
9346 (mask_gather_load@var{m}): Likewise.
9347 * genopinit.c (main): Add supports_vec_gather_load and
9348 supports_vec_gather_load_cached to target_optabs.
9349 * optabs-tree.c (init_tree_optimization_optabs): Use
9350 ggc_cleared_alloc to allocate target_optabs.
9351 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9352 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9354 * internal-fn.h (internal_load_fn_p): Declare.
9355 (internal_gather_scatter_fn_p): Likewise.
9356 (internal_fn_mask_index): Likewise.
9357 (internal_gather_scatter_fn_supported_p): Likewise.
9358 * internal-fn.c (gather_load_direct): New macro.
9359 (expand_gather_load_optab_fn): New function.
9360 (direct_gather_load_optab_supported_p): New macro.
9361 (direct_internal_fn_optab): New function.
9362 (internal_load_fn_p): Likewise.
9363 (internal_gather_scatter_fn_p): Likewise.
9364 (internal_fn_mask_index): Likewise.
9365 (internal_gather_scatter_fn_supported_p): Likewise.
9366 * optabs-query.c (supports_at_least_one_mode_p): New function.
9367 (supports_vec_gather_load_p): Likewise.
9368 * optabs-query.h (supports_vec_gather_load_p): Declare.
9369 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9370 and memory_type field.
9371 (NUM_PATTERNS): Bump to 15.
9372 * tree-vect-data-refs.c: Include internal-fn.h.
9373 (vect_gather_scatter_fn_p): New function.
9374 (vect_describe_gather_scatter_call): Likewise.
9375 (vect_check_gather_scatter): Try using internal functions for
9376 gather loads. Recognize existing calls to a gather load function.
9377 (vect_analyze_data_refs): Consider using gather loads if
9378 supports_vec_gather_load_p.
9379 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9380 (vect_get_gather_scatter_offset_type): Likewise.
9381 (vect_convert_mask_for_vectype): Likewise.
9382 (vect_add_conversion_to_patterm): Likewise.
9383 (vect_try_gather_scatter_pattern): Likewise.
9384 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9385 (vect_vect_recog_func_ptrs): Add it.
9386 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9387 internal_fn_mask_index and internal_gather_scatter_fn_p.
9388 (check_load_store_masking): Take the gather_scatter_info as an
9389 argument and handle gather loads.
9390 (vect_get_gather_scatter_ops): New function.
9391 (vectorizable_call): Check internal_load_fn_p.
9392 (vectorizable_load): Likewise. Handle gather load internal
9394 (vectorizable_store): Update call to check_load_store_masking.
9395 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9396 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9397 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9398 (aarch64_gather_scale_operand_d): New predicates.
9399 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9400 (mask_gather_load<mode>): New insns.
9402 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9403 Alan Hayward <alan.hayward@arm.com>
9404 David Sherwood <david.sherwood@arm.com>
9406 * optabs.def (fold_left_plus_optab): New optab.
9407 * doc/md.texi (fold_left_plus_@var{m}): Document.
9408 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9409 * internal-fn.c (fold_left_direct): Define.
9410 (expand_fold_left_optab_fn): Likewise.
9411 (direct_fold_left_optab_supported_p): Likewise.
9412 * fold-const-call.c (fold_const_fold_left): New function.
9413 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9414 * tree-parloops.c (valid_reduction_p): New function.
9415 (gather_scalar_reductions): Use it.
9416 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9417 (vect_finish_replace_stmt): Declare.
9418 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9419 (needs_fold_left_reduction_p): New function, split out from...
9420 (vect_is_simple_reduction): ...here. Accept reductions that
9421 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9422 (vect_force_simple_reduction): Also store the reduction type in
9423 the assignment's STMT_VINFO_REDUC_TYPE.
9424 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9425 (merge_with_identity): New function.
9426 (vect_expand_fold_left): Likewise.
9427 (vectorize_fold_left_reduction): Likewise.
9428 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9429 scalar phi in place for it. Check for target support and reject
9430 cases that would reassociate the operation. Defer the transform
9431 phase to vectorize_fold_left_reduction.
9432 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9433 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9434 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9436 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9438 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9439 call to ifc_temp_var.
9441 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9442 Alan Hayward <alan.hayward@arm.com>
9443 David Sherwood <david.sherwood@arm.com>
9445 * target.def (legitimize_address_displacement): Take the original
9446 offset as a poly_int.
9447 * targhooks.h (default_legitimize_address_displacement): Update
9449 * targhooks.c (default_legitimize_address_displacement): Likewise.
9450 * doc/tm.texi: Regenerate.
9451 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9452 as an argument, moving assert of ad->disp == ad->disp_term to...
9453 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9454 Try calling targetm.legitimize_address_displacement before expanding
9455 the address rather than afterwards, and adjust for the new interface.
9456 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9457 Match the new hook interface. Handle SVE addresses.
9458 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9461 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9463 * Makefile.in (OBJS): Add early-remat.o.
9464 * target.def (select_early_remat_modes): New hook.
9465 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9466 * doc/tm.texi: Regenerate.
9467 * targhooks.h (default_select_early_remat_modes): Declare.
9468 * targhooks.c (default_select_early_remat_modes): New function.
9469 * timevar.def (TV_EARLY_REMAT): New timevar.
9470 * passes.def (pass_early_remat): New pass.
9471 * tree-pass.h (make_pass_early_remat): Declare.
9472 * early-remat.c: New file.
9473 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9475 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9477 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9478 Alan Hayward <alan.hayward@arm.com>
9479 David Sherwood <david.sherwood@arm.com>
9481 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9482 vfm1 with a bound_epilog parameter.
9483 (vect_do_peeling): Update calls accordingly, and move the prologue
9484 call earlier in the function. Treat the base bound_epilog as 0 for
9485 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9486 this base when peeling for gaps.
9487 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9488 with fully-masked loops.
9489 (vect_estimate_min_profitable_iters): Handle the single peeled
9490 iteration in that case.
9492 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9493 Alan Hayward <alan.hayward@arm.com>
9494 David Sherwood <david.sherwood@arm.com>
9496 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9497 single-element interleaving even if the size is not a power of 2.
9498 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9499 accesses for single-element interleaving if the group size is
9502 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9503 Alan Hayward <alan.hayward@arm.com>
9504 David Sherwood <david.sherwood@arm.com>
9506 * doc/md.texi (fold_extract_last_@var{m}): Document.
9507 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9508 * optabs.def (fold_extract_last_optab): New optab.
9509 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9510 * internal-fn.c (fold_extract_direct): New macro.
9511 (expand_fold_extract_optab_fn): Likewise.
9512 (direct_fold_extract_optab_supported_p): Likewise.
9513 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9514 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9515 EXTRACT_LAST_REDUCTION.
9516 (get_initial_def_for_reduction): Do not create an initial vector
9517 for EXTRACT_LAST_REDUCTION reductions.
9518 (vectorizable_reduction): Leave the scalar phi in place for
9519 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9520 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9521 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9522 transform phase to vectorizable_condition.
9523 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9525 (vect_finish_stmt_generation): ...here.
9526 (vect_finish_replace_stmt): New function.
9527 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9528 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9530 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9532 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9533 Alan Hayward <alan.hayward@arm.com>
9534 David Sherwood <david.sherwood@arm.com>
9536 * doc/md.texi (extract_last_@var{m}): Document.
9537 * optabs.def (extract_last_optab): New optab.
9538 * internal-fn.def (EXTRACT_LAST): New internal function.
9539 * internal-fn.c (cond_unary_direct): New macro.
9540 (expand_cond_unary_optab_fn): Likewise.
9541 (direct_cond_unary_optab_supported_p): Likewise.
9542 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9543 loops using EXTRACT_LAST.
9544 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9545 (extract_last_<mode>): ...this optab.
9546 (vec_extract<mode><Vel>): Update accordingly.
9548 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9549 Alan Hayward <alan.hayward@arm.com>
9550 David Sherwood <david.sherwood@arm.com>
9552 * target.def (empty_mask_is_expensive): New hook.
9553 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9554 * doc/tm.texi: Regenerate.
9555 * targhooks.h (default_empty_mask_is_expensive): Declare.
9556 * targhooks.c (default_empty_mask_is_expensive): New function.
9557 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9558 if the target says that empty masks are expensive.
9559 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9561 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9563 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9564 Alan Hayward <alan.hayward@arm.com>
9565 David Sherwood <david.sherwood@arm.com>
9567 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9568 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9569 (vect_use_loop_mask_for_alignment_p): New function.
9570 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9571 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9572 niters_skip argument. Make sure that the first niters_skip elements
9573 of the first iteration are inactive.
9574 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9575 Update call to vect_set_loop_masks_directly.
9576 (get_misalign_in_elems): New function, split out from...
9577 (vect_gen_prolog_loop_niters): ...here.
9578 (vect_update_init_of_dr): Take a code argument that specifies whether
9579 the adjustment should be added or subtracted.
9580 (vect_update_init_of_drs): Likewise.
9581 (vect_prepare_for_masked_peels): New function.
9582 (vect_do_peeling): Skip prologue peeling if we're using a mask
9583 instead. Update call to vect_update_inits_of_drs.
9584 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9586 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9587 alignment. Do not include the number of peeled iterations in
9588 the minimum threshold in that case.
9589 (vectorizable_induction): Adjust the start value down by
9590 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9591 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9592 Take the number of skipped iterations into account when calculating
9594 * tree-vect-stmts.c (vect_gen_while_not): New function.
9596 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9597 Alan Hayward <alan.hayward@arm.com>
9598 David Sherwood <david.sherwood@arm.com>
9600 * doc/sourcebuild.texi (vect_fully_masked): Document.
9601 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9603 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9605 (vect_analyze_loop_2): ...here. Don't check the vectorization
9606 factor against the number of loop iterations if the loop is
9609 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9610 Alan Hayward <alan.hayward@arm.com>
9611 David Sherwood <david.sherwood@arm.com>
9613 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9614 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9615 (dump_groups): Update accordingly.
9616 (iv_use::mem_type): New member variable.
9617 (address_p): New function.
9618 (record_use): Add a mem_type argument and initialize the new
9620 (record_group_use): Add a mem_type argument. Use address_p.
9621 Remove obsolete null checks of base_object. Update call to record_use.
9622 (find_interesting_uses_op): Update call to record_group_use.
9623 (find_interesting_uses_cond): Likewise.
9624 (find_interesting_uses_address): Likewise.
9625 (get_mem_type_for_internal_fn): New function.
9626 (find_address_like_use): Likewise.
9627 (find_interesting_uses_stmt): Try find_address_like_use before
9628 calling find_interesting_uses_op.
9629 (addr_offset_valid_p): Use the iv mem_type field as the type
9630 of the addressed memory.
9631 (add_autoinc_candidates): Likewise.
9632 (get_address_cost): Likewise.
9633 (split_small_address_groups_p): Use address_p.
9634 (split_address_groups): Likewise.
9635 (add_iv_candidate_for_use): Likewise.
9636 (autoinc_possible_for_pair): Likewise.
9637 (rewrite_groups): Likewise.
9638 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9639 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9640 (get_alias_ptr_type_for_ptr_address): New function.
9641 (rewrite_use_address): Rewrite address uses in calls that were
9642 identified by find_address_like_use.
9644 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9645 Alan Hayward <alan.hayward@arm.com>
9646 David Sherwood <david.sherwood@arm.com>
9648 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9650 * gimple-expr.h (is_gimple_addressable: Likewise.
9651 * gimple-expr.c (is_gimple_address): Likewise.
9652 * internal-fn.c (expand_call_mem_ref): New function.
9653 (expand_mask_load_optab_fn): Use it.
9654 (expand_mask_store_optab_fn): Likewise.
9656 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9657 Alan Hayward <alan.hayward@arm.com>
9658 David Sherwood <david.sherwood@arm.com>
9660 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9661 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9662 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9663 (cond_umax@var{mode}): Document.
9664 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9665 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9666 (cond_umin_optab, cond_umax_optab): New optabs.
9667 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9668 (COND_IOR, COND_XOR): New internal functions.
9669 * internal-fn.h (get_conditional_internal_fn): Declare.
9670 * internal-fn.c (cond_binary_direct): New macro.
9671 (expand_cond_binary_optab_fn): Likewise.
9672 (direct_cond_binary_optab_supported_p): Likewise.
9673 (get_conditional_internal_fn): New function.
9674 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9675 Cope with reduction statements that are vectorized as calls rather
9677 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9678 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9679 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9680 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9681 (UNSPEC_COND_EOR): New unspecs.
9682 (optab): Add mappings for them.
9683 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9684 (sve_int_op, sve_fp_op): New int attributes.
9686 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9687 Alan Hayward <alan.hayward@arm.com>
9688 David Sherwood <david.sherwood@arm.com>
9690 * optabs.def (while_ult_optab): New optab.
9691 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9692 * internal-fn.def (WHILE_ULT): New internal function.
9693 * internal-fn.h (direct_internal_fn_supported_p): New override
9694 that takes two types as argument.
9695 * internal-fn.c (while_direct): New macro.
9696 (expand_while_optab_fn): New function.
9697 (convert_optab_supported_p): Likewise.
9698 (direct_while_optab_supported_p): New macro.
9699 * wide-int.h (wi::udiv_ceil): New function.
9700 * tree-vectorizer.h (rgroup_masks): New structure.
9701 (vec_loop_masks): New typedef.
9702 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9704 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9705 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9706 (vect_max_vf): New function.
9707 (slpeel_make_loop_iterate_ntimes): Delete.
9708 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9709 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9710 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9711 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9712 internal-fn.h, stor-layout.h and optabs-query.h.
9713 (vect_set_loop_mask): New function.
9714 (add_preheader_seq): Likewise.
9715 (add_header_seq): Likewise.
9716 (interleave_supported_p): Likewise.
9717 (vect_maybe_permute_loop_masks): Likewise.
9718 (vect_set_loop_masks_directly): Likewise.
9719 (vect_set_loop_condition_masked): Likewise.
9720 (vect_set_loop_condition_unmasked): New function, split out from
9721 slpeel_make_loop_iterate_ntimes.
9722 (slpeel_make_loop_iterate_ntimes): Rename to..
9723 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9724 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9725 (vect_do_peeling): Update call accordingly.
9726 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9728 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9729 mask_compare_type, can_fully_mask_p and fully_masked_p.
9730 (release_vec_loop_masks): New function.
9731 (_loop_vec_info): Use it to free the loop masks.
9732 (can_produce_all_loop_masks_p): New function.
9733 (vect_get_max_nscalars_per_iter): Likewise.
9734 (vect_verify_full_masking): Likewise.
9735 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9736 retries, and free the mask rgroups before retrying. Check loop-wide
9737 reasons for disallowing fully-masked loops. Make the final decision
9738 about whether use a fully-masked loop or not.
9739 (vect_estimate_min_profitable_iters): Do not assume that peeling
9740 for the number of iterations will be needed for fully-masked loops.
9741 (vectorizable_reduction): Disable fully-masked loops.
9742 (vectorizable_live_operation): Likewise.
9743 (vect_halve_mask_nunits): New function.
9744 (vect_double_mask_nunits): Likewise.
9745 (vect_record_loop_mask): Likewise.
9746 (vect_get_loop_mask): Likewise.
9747 (vect_transform_loop): Handle the case in which the final loop
9748 iteration might handle a partial vector. Call vect_set_loop_condition
9749 instead of slpeel_make_loop_iterate_ntimes.
9750 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9751 (check_load_store_masking): New function.
9752 (prepare_load_store_mask): Likewise.
9753 (vectorizable_store): Handle fully-masked loops.
9754 (vectorizable_load): Likewise.
9755 (supportable_widening_operation): Use vect_halve_mask_nunits for
9757 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9758 (vect_gen_while): New function.
9759 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9760 (aarch64_uqdec<mode>): New insn.
9762 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9763 Alan Hayward <alan.hayward@arm.com>
9764 David Sherwood <david.sherwood@arm.com>
9766 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9767 (reduc_xor_scal_optab): New optabs.
9768 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9769 (reduc_xor_scal_@var{m}): Document.
9770 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9771 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9773 * fold-const-call.c (fold_const_call): Handle them.
9774 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9775 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9776 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9777 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9778 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9779 (UNSPEC_XORV): New unspecs.
9780 (optab): Add entries for them.
9781 (BITWISEV): New int iterator.
9782 (bit_reduc_op): New int attributes.
9784 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9785 Alan Hayward <alan.hayward@arm.com>
9786 David Sherwood <david.sherwood@arm.com>
9788 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9789 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9790 * optabs.def (vec_shl_insert_optab): New optab.
9791 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9792 (duplicate_and_interleave): Likewise.
9793 * tree-vect-loop.c: Include internal-fn.h.
9794 (neutral_op_for_slp_reduction): New function, split out from
9795 get_initial_defs_for_reduction.
9796 (get_initial_def_for_reduction): Handle option 2 for variable-length
9797 vectors by loading the neutral value into a vector and then shifting
9798 the initial value into element 0.
9799 (get_initial_defs_for_reduction): Replace the code argument with
9800 the neutral value calculated by neutral_op_for_slp_reduction.
9801 Use gimple_build_vector for constant-length vectors.
9802 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9803 but the first group_size elements have a neutral value.
9804 Use duplicate_and_interleave otherwise.
9805 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9806 Update call to get_initial_defs_for_reduction. Handle SLP
9807 reductions for variable-length vectors by creating one vector
9808 result for each scalar result, with the elements associated
9809 with other scalar results stubbed out with the neutral value.
9810 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9811 Require IFN_VEC_SHL_INSERT for double reductions on
9812 variable-length vectors, or SLP reductions that have
9813 a neutral value. Require can_duplicate_and_interleave_p
9814 support for variable-length unchained SLP reductions if there
9815 is no neutral value, such as for MIN/MAX reductions. Also require
9816 the number of vector elements to be a multiple of the number of
9817 SLP statements when doing variable-length unchained SLP reductions.
9818 Update call to vect_create_epilog_for_reduction.
9819 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9820 and remove initial values.
9821 (duplicate_and_interleave): Make public.
9822 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9823 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9825 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9826 Alan Hayward <alan.hayward@arm.com>
9827 David Sherwood <david.sherwood@arm.com>
9829 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9830 (can_duplicate_and_interleave_p): New function.
9831 (vect_get_and_check_slp_defs): Take the vector of statements
9832 rather than just the current one. Remove excess parentheses.
9833 Restriction rejectinon of vect_constant_def and vect_external_def
9834 for variable-length vectors to boolean types, or types for which
9835 can_duplicate_and_interleave_p is false.
9836 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9837 (duplicate_and_interleave): New function.
9838 (vect_get_constant_vectors): Use gimple_build_vector for
9839 constant-length vectors and suitable variable-length constant
9840 vectors. Use duplicate_and_interleave for other variable-length
9841 vectors. Don't defer the update when inserting new statements.
9843 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9844 Alan Hayward <alan.hayward@arm.com>
9845 David Sherwood <david.sherwood@arm.com>
9847 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9848 min_profitable_iters doesn't go negative.
9850 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9851 Alan Hayward <alan.hayward@arm.com>
9852 David Sherwood <david.sherwood@arm.com>
9854 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9855 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9856 * optabs.def (vec_mask_load_lanes_optab): New optab.
9857 (vec_mask_store_lanes_optab): Likewise.
9858 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9859 (MASK_STORE_LANES): Likewise.
9860 * internal-fn.c (mask_load_lanes_direct): New macro.
9861 (mask_store_lanes_direct): Likewise.
9862 (expand_mask_load_optab_fn): Handle masked operations.
9863 (expand_mask_load_lanes_optab_fn): New macro.
9864 (expand_mask_store_optab_fn): Handle masked operations.
9865 (expand_mask_store_lanes_optab_fn): New macro.
9866 (direct_mask_load_lanes_optab_supported_p): Likewise.
9867 (direct_mask_store_lanes_optab_supported_p): Likewise.
9868 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9870 (vect_load_lanes_supported): Likewise.
9871 * tree-vect-data-refs.c (strip_conversion): New function.
9872 (can_group_stmts_p): Likewise.
9873 (vect_analyze_data_ref_accesses): Use it instead of checking
9874 for a pair of assignments.
9875 (vect_store_lanes_supported): Take a masked_p parameter.
9876 (vect_load_lanes_supported): Likewise.
9877 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9878 vect_store_lanes_supported and vect_load_lanes_supported.
9879 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9880 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9881 parameter. Don't allow gaps for masked accesses.
9882 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9883 and vect_load_lanes_supported.
9884 (get_load_store_type): Take a masked_p parameter and update
9885 call to get_group_load_store_type.
9886 (vectorizable_store): Update call to get_load_store_type.
9887 Handle IFN_MASK_STORE_LANES.
9888 (vectorizable_load): Update call to get_load_store_type.
9889 Handle IFN_MASK_LOAD_LANES.
9891 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9892 Alan Hayward <alan.hayward@arm.com>
9893 David Sherwood <david.sherwood@arm.com>
9895 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9897 * config/aarch64/aarch64-protos.h
9898 (aarch64_sve_struct_memory_operand_p): Declare.
9899 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9900 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9901 (VPRED, vpred): Handle SVE structure modes.
9902 * config/aarch64/constraints.md (Utx): New constraint.
9903 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9904 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9905 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9906 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9907 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9908 structure modes. Split into pieces after RA.
9909 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9910 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9912 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9913 SVE structure modes.
9914 (aarch64_classify_address): Likewise.
9915 (sizetochar): Move earlier in file.
9916 (aarch64_print_operand): Handle SVE register lists.
9917 (aarch64_array_mode): New function.
9918 (aarch64_sve_struct_memory_operand_p): Likewise.
9919 (TARGET_ARRAY_MODE): Redefine.
9921 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9922 Alan Hayward <alan.hayward@arm.com>
9923 David Sherwood <david.sherwood@arm.com>
9925 * target.def (array_mode): New target hook.
9926 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9927 * doc/tm.texi: Regenerate.
9928 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9929 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9930 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9932 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9935 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9936 Alan Hayward <alan.hayward@arm.com>
9937 David Sherwood <david.sherwood@arm.com>
9939 * fold-const.c (fold_binary_loc): Check the argument types
9940 rather than the result type when testing for a vector operation.
9942 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9944 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9945 * doc/tm.texi: Regenerate.
9947 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9948 Alan Hayward <alan.hayward@arm.com>
9949 David Sherwood <david.sherwood@arm.com>
9951 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9952 (sve): Document new AArch64 extension.
9953 * doc/md.texi (w): Extend the description of the AArch64
9954 constraint to include SVE vectors.
9955 (Upl, Upa): Document new AArch64 predicate constraints.
9956 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9958 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9959 (msve-vector-bits=): New option.
9960 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9961 SVE when these are disabled.
9962 (sve): New extension.
9963 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9964 modes. Adjust their number of units based on aarch64_sve_vg.
9965 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9966 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9967 aarch64_addr_query_type.
9968 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9969 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9970 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9971 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9972 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9973 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9974 (aarch64_simd_imm_zero_p): Delete.
9975 (aarch64_check_zero_based_sve_index_immediate): Declare.
9976 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9977 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9978 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9979 (aarch64_sve_float_mul_immediate_p): Likewise.
9980 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9982 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9983 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9984 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9985 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9986 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9987 (aarch64_regmode_natural_size): Likewise.
9988 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9989 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9991 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9992 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9993 for VG and the SVE predicate registers.
9994 (V_ALIASES): Add a "z"-prefixed alias.
9995 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9996 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9997 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9998 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9999 (REG_CLASS_NAMES): Add entries for them.
10000 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10001 and the predicate registers.
10002 (aarch64_sve_vg): Declare.
10003 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10004 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10005 (REGMODE_NATURAL_SIZE): Define.
10006 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10008 * config/aarch64/aarch64.c: Include cfgrtl.h.
10009 (simd_immediate_info): Add a constructor for series vectors,
10010 and an associated step field.
10011 (aarch64_sve_vg): New variable.
10012 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10013 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10014 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10015 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10016 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10017 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10018 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10019 (aarch64_get_mask_mode): New functions.
10020 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10021 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10022 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10023 predicate modes and predicate registers. Explicitly restrict
10024 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10025 to store a vector mode if it is recognized by
10026 aarch64_classify_vector_mode.
10027 (aarch64_regmode_natural_size): New function.
10028 (aarch64_hard_regno_caller_save_mode): Return the original mode
10030 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10031 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10032 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10033 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10035 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10036 does not overlap dest if the function is frame-related. Handle
10038 (aarch64_split_add_offset): New function.
10039 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10040 them aarch64_add_offset.
10041 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10042 and update call to aarch64_sub_sp.
10043 (aarch64_add_cfa_expression): New function.
10044 (aarch64_expand_prologue): Pass extra temporary registers to the
10045 functions above. Handle the case in which we need to emit new
10046 DW_CFA_expressions for registers that were originally saved
10047 relative to the stack pointer, but now have to be expressed
10048 relative to the frame pointer.
10049 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10051 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10052 IP0 and IP1 values for SVE frames.
10053 (aarch64_expand_vec_series): New function.
10054 (aarch64_expand_sve_widened_duplicate): Likewise.
10055 (aarch64_expand_sve_const_vector): Likewise.
10056 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10057 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10058 into the register, rather than emitting a SET directly.
10059 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10060 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10061 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10062 (offset_9bit_signed_scaled_p): New functions.
10063 (aarch64_replicate_bitmask_imm): New function.
10064 (aarch64_bitmask_imm): Use it.
10065 (aarch64_cannot_force_const_mem): Reject expressions involving
10066 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10067 (aarch64_classify_index): Handle SVE indices, by requiring
10068 a plain register index with a scale that matches the element size.
10069 (aarch64_classify_address): Handle SVE addresses. Assert that
10070 the mode of the address is VOIDmode or an integer mode.
10071 Update call to aarch64_classify_symbol.
10072 (aarch64_classify_symbolic_expression): Update call to
10073 aarch64_classify_symbol.
10074 (aarch64_const_vec_all_in_range_p): New function.
10075 (aarch64_print_vector_float_operand): Likewise.
10076 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10077 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10078 and the FP immediates 1.0 and 0.5.
10079 (aarch64_print_address_internal): Handle SVE addresses.
10080 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10081 (aarch64_regno_regclass): Handle predicate registers.
10082 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10084 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10085 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10086 (aarch64_convert_sve_vector_bits): New function.
10087 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10088 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10089 rather than an rtx.
10090 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10091 Handle SVE vector and predicate modes. Accept VL-based constants
10092 that need only one temporary register, and VL offsets that require
10093 no temporary registers.
10094 (aarch64_conditional_register_usage): Mark the predicate registers
10095 as fixed if SVE isn't available.
10096 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10097 Return true for SVE vector and predicate modes.
10098 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10099 rather than an unsigned int. Handle SVE modes.
10100 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10102 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10104 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10105 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10106 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10107 (aarch64_sve_float_mul_immediate_p): New functions.
10108 (aarch64_sve_valid_immediate): New function.
10109 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10110 Explicitly reject structure modes. Check for INDEX constants.
10111 Handle PTRUE and PFALSE constants.
10112 (aarch64_check_zero_based_sve_index_immediate): New function.
10113 (aarch64_simd_imm_zero_p): Delete.
10114 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10115 vector modes. Accept constants in the range of CNT[BHWD].
10116 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10117 ask for an Advanced SIMD mode.
10118 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10119 (aarch64_simd_vector_alignment): Handle SVE predicates.
10120 (aarch64_vectorize_preferred_vector_alignment): New function.
10121 (aarch64_simd_vector_alignment_reachable): Use it instead of
10123 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10124 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10126 (MAX_VECT_LEN): Delete.
10127 (expand_vec_perm_d): Add a vec_flags field.
10128 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10129 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10130 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10132 (aarch64_evpc_rev): Rename to...
10133 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10134 (aarch64_evpc_rev_global): New function.
10135 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10136 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10138 (aarch64_evpc_sve_tbl): New function.
10139 (aarch64_expand_vec_perm_const_1): Update after rename of
10140 aarch64_evpc_rev. Handle SVE permutes too, trying
10141 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10142 than aarch64_evpc_tbl.
10143 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10144 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10145 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10146 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10147 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10148 (aarch64_expand_sve_vcond): New functions.
10149 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10150 of aarch64_vector_mode_p.
10151 (aarch64_dwarf_poly_indeterminate_value): New function.
10152 (aarch64_compute_pressure_classes): Likewise.
10153 (aarch64_can_change_mode_class): Likewise.
10154 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10155 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10156 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10157 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10158 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10159 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10160 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10161 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10163 (Dn, Dl, Dr): Accept const as well as const_vector.
10164 (Dz): Likewise. Compare against CONST0_RTX.
10165 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10166 of "vector" where appropriate.
10167 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10168 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10169 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10170 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10171 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10172 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10173 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10174 (v_int_equiv): Extend to SVE modes.
10175 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10177 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10178 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10179 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10180 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10181 (SVE_COND_FP_CMP): New int iterators.
10182 (perm_hilo): Handle the new unpack unspecs.
10183 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10185 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10186 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10187 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10188 (aarch64_equality_operator, aarch64_constant_vector_operand)
10189 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10190 (aarch64_sve_nonimmediate_operand): Likewise.
10191 (aarch64_sve_general_operand): Likewise.
10192 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10193 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10194 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10195 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10196 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10197 (aarch64_sve_float_arith_immediate): Likewise.
10198 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10199 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10200 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10201 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10202 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10203 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10204 (aarch64_sve_float_arith_operand): Likewise.
10205 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10206 (aarch64_sve_float_mul_operand): Likewise.
10207 (aarch64_sve_vec_perm_operand): Likewise.
10208 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10209 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10210 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10211 as well as const_vector.
10212 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10213 in file. Use CONST0_RTX and CONSTM1_RTX.
10214 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10215 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10216 Use aarch64_simd_imm_zero.
10217 * config/aarch64/aarch64-sve.md: New file.
10218 * config/aarch64/aarch64.md: Include it.
10219 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10220 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10221 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10222 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10223 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10224 (sve): New attribute.
10225 (enabled): Disable instructions with the sve attribute unless
10227 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10228 aarch64_expand_mov_immediate.
10229 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10230 CNT[BHSD] immediates.
10231 (movti): Split CONST_POLY_INT moves into two halves.
10232 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10233 Split additions that need a temporary here if the destination
10234 is the stack pointer.
10235 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10236 (*add<mode>3_poly_1): New instruction.
10237 (set_clobber_cc): New expander.
10239 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10241 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10242 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10243 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10244 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10245 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10246 Change innermode from fixed_mode_size to machine_mode.
10247 (simplify_subreg): Update call accordingly. Handle a constant-sized
10248 subreg of a variable-length CONST_VECTOR.
10250 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10251 Alan Hayward <alan.hayward@arm.com>
10252 David Sherwood <david.sherwood@arm.com>
10254 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10255 (add_offset_to_base): New function, split out from...
10256 (create_mem_ref): ...here. When handling a scale other than 1,
10257 check first whether the address is valid without the offset.
10258 Add it into the base if so, leaving the index and scale as-is.
10260 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10263 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10264 fold_for_warn before checking if arg2 is INTEGER_CST.
10266 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10268 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10269 (store_multiple_operation): Delete.
10270 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10271 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10272 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10273 guarded by TARGET_STRING.
10274 (rs6000_output_load_multiple): Delete.
10275 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10276 OPTION_MASK_STRING / TARGET_STRING handling.
10277 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10278 (const rs6000_opt_masks) <"string">: Change mask to 0.
10279 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10280 (MASK_STRING): Delete.
10281 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10283 (load_multiple): Delete.
10290 (store_multiple): Delete.
10297 (movmemsi_8reg): Delete.
10298 (corresponding unnamed define_insn): Delete.
10299 (movmemsi_6reg): Delete.
10300 (corresponding unnamed define_insn): Delete.
10301 (movmemsi_4reg): Delete.
10302 (corresponding unnamed define_insn): Delete.
10303 (movmemsi_2reg): Delete.
10304 (corresponding unnamed define_insn): Delete.
10305 (movmemsi_1reg): Delete.
10306 (corresponding unnamed define_insn): Delete.
10307 * config/rs6000/rs6000.opt (mno-string): New.
10308 (mstring): Replace by deprecation warning stub.
10309 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10311 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10313 * regrename.c (regrename_do_replace): If replacing the same
10314 reg multiple times, try to reuse last created gen_raw_REG.
10317 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10318 main to workaround a bug in GDB.
10320 2018-01-12 Tom de Vries <tom@codesourcery.com>
10323 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10325 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10327 PR rtl-optimization/80481
10328 * ira-color.c (get_cap_member): New function.
10329 (allocnos_conflict_by_live_ranges_p): Use it.
10330 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10331 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10333 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10336 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10337 (*saddl_se_1): Ditto.
10338 (*ssubsi_1): Ditto.
10339 (*ssubl_se_1): Ditto.
10341 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10343 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10344 rather than wi::to_widest for DR_INITs.
10345 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10346 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10347 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10349 (vect_analyze_group_access_1): Note that here.
10351 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10353 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10354 polynomial type sizes.
10356 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10358 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10359 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10360 (gimple_add_tmp_var): Likewise.
10362 2018-01-12 Martin Liska <mliska@suse.cz>
10364 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10365 (gimple_alloc_sizes): Likewise.
10366 (dump_gimple_statistics): Use PRIu64 in printf format.
10367 * gimple.h: Change uint64_t to int.
10369 2018-01-12 Martin Liska <mliska@suse.cz>
10371 * tree-core.h: Use uint64_t instead of int.
10372 * tree.c (tree_node_counts): Likewise.
10373 (tree_node_sizes): Likewise.
10374 (dump_tree_statistics): Use PRIu64 in printf format.
10376 2018-01-12 Martin Liska <mliska@suse.cz>
10378 * Makefile.in: As qsort_chk is implemented in vec.c, add
10379 vec.o to linkage of gencfn-macros.
10380 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10381 passing the info to record_node_allocation_statistics.
10382 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10384 * ggc-common.c (struct ggc_usage): Add operator== and use
10385 it in operator< and compare function.
10386 * mem-stats.h (struct mem_usage): Likewise.
10387 * vec.c (struct vec_usage): Remove operator< and compare
10388 function. Can be simply inherited.
10390 2018-01-12 Martin Jambor <mjambor@suse.cz>
10393 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10394 * tree-ssa-math-opts.c: Include domwalk.h.
10395 (convert_mult_to_fma_1): New function.
10396 (fma_transformation_info): New type.
10397 (fma_deferring_state): Likewise.
10398 (cancel_fma_deferring): New function.
10399 (result_of_phi): Likewise.
10400 (last_fma_candidate_feeds_initial_phi): Likewise.
10401 (convert_mult_to_fma): Added deferring logic, split actual
10402 transformation to convert_mult_to_fma_1.
10403 (math_opts_dom_walker): New type.
10404 (math_opts_dom_walker::after_dom_children): New method, body moved
10405 here from pass_optimize_widening_mul::execute, added deferring logic
10407 (pass_optimize_widening_mul::execute): Moved most of code to
10408 math_opts_dom_walker::after_dom_children.
10409 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10410 * config/i386/i386.c (ix86_option_override_internal): Added
10411 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10413 2018-01-12 Richard Biener <rguenther@suse.de>
10416 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10417 inline instance vars.
10419 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10422 * config/rx/rx.c (rx_is_restricted_memory_address):
10423 Handle SUBREG case.
10425 2018-01-12 Richard Biener <rguenther@suse.de>
10427 PR tree-optimization/80846
10428 * target.def (split_reduction): New target hook.
10429 * targhooks.c (default_split_reduction): New function.
10430 * targhooks.h (default_split_reduction): Declare.
10431 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10432 target requests first reduce vectors by combining low and high
10434 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10435 (get_vectype_for_scalar_type_and_size): Export.
10436 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10437 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10438 * doc/tm.texi: Regenerate.
10439 * config/i386/i386.c (ix86_split_reduction): Implement
10440 TARGET_VECTORIZE_SPLIT_REDUCTION.
10442 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10445 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10446 in PIC mode except for TARGET_VXWORKS_RTP.
10447 * config/sparc/sparc.c: Include cfgrtl.h.
10448 (TARGET_INIT_PIC_REG): Define.
10449 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10450 (sparc_pic_register_p): New predicate.
10451 (sparc_legitimate_address_p): Use it.
10452 (sparc_legitimize_pic_address): Likewise.
10453 (sparc_delegitimize_address): Likewise.
10454 (sparc_mode_dependent_address_p): Likewise.
10455 (gen_load_pcrel_sym): Remove 4th parameter.
10456 (load_got_register): Adjust call to above. Remove obsolete stuff.
10457 (sparc_expand_prologue): Do not call load_got_register here.
10458 (sparc_flat_expand_prologue): Likewise.
10459 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10460 (sparc_use_pseudo_pic_reg): New function.
10461 (sparc_init_pic_reg): Likewise.
10462 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10463 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10465 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10467 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10468 Add item for branch_cost.
10470 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10472 PR rtl-optimization/83565
10473 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10474 not extend the result to a larger mode for rotate operations.
10475 (num_sign_bit_copies1): Likewise.
10477 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10480 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10482 Use values-Xc.o for -pedantic.
10483 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10485 2018-01-12 Martin Liska <mliska@suse.cz>
10488 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10490 (possible_polymorphic_call_targets): Use it.
10491 (ipa_devirt): Likewise.
10493 2018-01-12 Martin Liska <mliska@suse.cz>
10495 * profile-count.h (enum profile_quality): Use 0 as invalid
10496 enum value of profile_quality.
10498 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10500 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10501 -mext-string options.
10503 2018-01-12 Richard Biener <rguenther@suse.de>
10505 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10506 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10507 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10509 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10511 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10513 * configure.ac (--with-long-double-format): Add support for the
10514 configuration option to change the default long double format on
10516 * config.gcc (powerpc*-linux*-*): Likewise.
10517 * configure: Regenerate.
10518 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10519 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10520 used without modification.
10522 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10524 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10525 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10526 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10527 MISC_BUILTIN_SPEC_BARRIER.
10528 (rs6000_init_builtins): Likewise.
10529 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10531 (speculation_barrier): New define_insn.
10532 * doc/extend.texi: Document __builtin_speculation_barrier.
10534 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10537 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10538 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10539 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10541 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10542 integral modes instead of "ss" and "sd".
10543 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10544 vectors with 32-bit and 64-bit elements.
10545 (vecdupssescalarmodesuffix): New mode attribute.
10546 (vec_dup<mode>): Use it.
10548 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10551 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10552 frame if argument is passed on stack.
10554 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10557 * ree.c (combine_reaching_defs): Optimize also
10558 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10559 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10561 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10563 PR middle-end/83189
10564 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10566 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10568 PR middle-end/83718
10569 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10570 after they are computed.
10572 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10574 PR tree-optimization/83695
10575 * gimple-loop-linterchange.cc
10576 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10577 reset cached scev information after interchange.
10578 (pass_linterchange::execute): Remove call to scev_reset_htab.
10580 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10582 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10583 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10584 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10585 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10586 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10587 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10588 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10589 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10590 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10591 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10592 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10593 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10594 (V_lane_reg): Likewise.
10595 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10597 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10598 (vfmal_lane_low<mode>_intrinsic,
10599 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10600 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10601 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10602 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10603 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10604 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10606 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10608 * config/arm/arm-cpus.in (fp16fml): New feature.
10609 (ALL_SIMD): Add fp16fml.
10610 (armv8.2-a): Add fp16fml as an option.
10611 (armv8.3-a): Likewise.
10612 (armv8.4-a): Add fp16fml as part of fp16.
10613 * config/arm/arm.h (TARGET_FP16FML): Define.
10614 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10616 * config/arm/arm-modes.def (V2HF): Define.
10617 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10618 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10619 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10620 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10621 vfmsl_low, vfmsl_high): New set of builtins.
10622 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10623 (vfml_op): New code attribute.
10624 (VFMLHALVES): New int iterator.
10625 (VFML, VFMLSEL): New mode attributes.
10626 (V_reg): Define mapping for V2HF.
10627 (V_hi, V_lo): New mode attributes.
10628 (VF_constraint): Likewise.
10629 (vfml_half, vfml_half_selector): New int attributes.
10630 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10632 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10633 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10635 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10636 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10637 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10638 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10640 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10641 Document new effective target and option set.
10643 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10645 * config/arm/arm-cpus.in (armv8_4): New feature.
10646 (ARMv8_4a): New fgroup.
10647 (armv8.4-a): New arch.
10648 * config/arm/arm-tables.opt: Regenerate.
10649 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10650 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10651 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10652 Add matching rules for -march=armv8.4-a and extensions.
10653 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10655 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10658 * config/rx/rx.md (BW): New mode attribute.
10659 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10661 2018-01-11 Richard Biener <rguenther@suse.de>
10663 PR tree-optimization/83435
10664 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10665 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10666 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10668 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10669 Alan Hayward <alan.hayward@arm.com>
10670 David Sherwood <david.sherwood@arm.com>
10672 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10674 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10675 (aarch64_print_address_internal): Use it to check for a zero offset.
10677 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10678 Alan Hayward <alan.hayward@arm.com>
10679 David Sherwood <david.sherwood@arm.com>
10681 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10682 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10683 Return a poly_int64 rather than a HOST_WIDE_INT.
10684 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10685 rather than a HOST_WIDE_INT.
10686 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10687 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10688 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10689 final_offset from HOST_WIDE_INT to poly_int64.
10690 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10691 to_constant when getting the number of units in an Advanced SIMD
10693 (aarch64_builtin_vectorized_function): Check for a constant number
10695 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10697 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10698 attribute instead of GET_MODE_NUNITS.
10699 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10700 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10701 GET_MODE_SIZE for fixed-size registers.
10702 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10703 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10704 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10705 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10706 (aarch64_print_operand, aarch64_print_address_internal)
10707 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10708 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10709 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10710 Handle polynomial GET_MODE_SIZE.
10711 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10712 wider than SImode without modification.
10713 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10714 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10715 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10716 passing and returning SVE modes.
10717 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10718 rather than GEN_INT.
10719 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10720 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10721 (aarch64_allocate_and_probe_stack_space): Likewise.
10722 (aarch64_layout_frame): Cope with polynomial offsets.
10723 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10724 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10725 polynomial offsets.
10726 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10727 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10728 poly_int64 rather than a HOST_WIDE_INT.
10729 (aarch64_get_separate_components, aarch64_process_components)
10730 (aarch64_expand_prologue, aarch64_expand_epilogue)
10731 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10732 (aarch64_anchor_offset): New function, split out from...
10733 (aarch64_legitimize_address): ...here.
10734 (aarch64_builtin_vectorization_cost): Handle polynomial
10735 TYPE_VECTOR_SUBPARTS.
10736 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10738 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10739 number of elements from the PARALLEL rather than the mode.
10740 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10741 rather than GET_MODE_BITSIZE.
10742 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10743 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10744 (aarch64_expand_vec_perm_const_1): Handle polynomial
10745 d->perm.length () and d->perm elements.
10746 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10747 Apply to_constant to d->perm elements.
10748 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10749 polynomial CONST_VECTOR_NUNITS.
10750 (aarch64_move_pointer): Take amount as a poly_int64 rather
10752 (aarch64_progress_pointer): Avoid temporary variable.
10753 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10754 the mode attribute instead of GET_MODE.
10756 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10757 Alan Hayward <alan.hayward@arm.com>
10758 David Sherwood <david.sherwood@arm.com>
10760 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10761 x exists before using it.
10762 (aarch64_add_constant_internal): Rename to...
10763 (aarch64_add_offset_1): ...this. Replace regnum with separate
10764 src and dest rtxes. Handle the case in which they're different,
10765 including when the offset is zero. Replace scratchreg with an rtx.
10766 Use 2 additions if there is no spare register into which we can
10767 move a 16-bit constant.
10768 (aarch64_add_constant): Delete.
10769 (aarch64_add_offset): Replace reg with separate src and dest
10770 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10771 Use aarch64_add_offset_1.
10772 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10773 an rtx rather than an int. Take the delta as a poly_int64
10774 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10775 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10776 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10777 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10778 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10779 and aarch64_add_sp.
10780 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10781 aarch64_add_constant.
10783 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10785 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10786 Use scalar_float_mode.
10788 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10790 * config/aarch64/aarch64-simd.md
10791 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10792 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10793 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10794 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10795 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10796 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10797 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10798 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10799 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10800 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10802 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10805 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10806 targ_options->x_arm_arch_string is non NULL.
10808 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10810 * config/aarch64/aarch64.h
10811 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10813 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10816 * expmed.c (emit_store_flag_force): Swap if const op0
10817 and change VOIDmode to mode of op0.
10819 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10821 PR rtl-optimization/83761
10822 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10823 than bytes to mode_for_size.
10825 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10827 PR middle-end/83189
10828 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10829 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10832 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10834 PR middle-end/83575
10835 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10836 when in layout mode.
10837 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10838 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10841 2018-01-10 Michael Collison <michael.collison@arm.com>
10843 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10844 * config/aarch64/aarch64-option-extension.def: Add
10845 AARCH64_OPT_EXTENSION of 'fp16fml'.
10846 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10847 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10848 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10849 * config/aarch64/constraints.md (Ui7): New constraint.
10850 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10851 (VFMLA_SEL_W): Ditto.
10854 (VFMLA16_LOW): New int iterator.
10855 (VFMLA16_HIGH): Ditto.
10856 (UNSPEC_FMLAL): New unspec.
10857 (UNSPEC_FMLSL): Ditto.
10858 (UNSPEC_FMLAL2): Ditto.
10859 (UNSPEC_FMLSL2): Ditto.
10860 (f16mac): New code attribute.
10861 * config/aarch64/aarch64-simd-builtins.def
10862 (aarch64_fmlal_lowv2sf): Ditto.
10863 (aarch64_fmlsl_lowv2sf): Ditto.
10864 (aarch64_fmlalq_lowv4sf): Ditto.
10865 (aarch64_fmlslq_lowv4sf): Ditto.
10866 (aarch64_fmlal_highv2sf): Ditto.
10867 (aarch64_fmlsl_highv2sf): Ditto.
10868 (aarch64_fmlalq_highv4sf): Ditto.
10869 (aarch64_fmlslq_highv4sf): Ditto.
10870 (aarch64_fmlal_lane_lowv2sf): Ditto.
10871 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10872 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10873 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10874 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10875 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10876 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10877 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10878 (aarch64_fmlal_lane_highv2sf): Ditto.
10879 (aarch64_fmlsl_lane_highv2sf): Ditto.
10880 (aarch64_fmlal_laneq_highv2sf): Ditto.
10881 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10882 (aarch64_fmlalq_lane_highv4sf): Ditto.
10883 (aarch64_fmlsl_lane_highv4sf): Ditto.
10884 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10885 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10886 * config/aarch64/aarch64-simd.md:
10887 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10888 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10889 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10890 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10891 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10892 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10893 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10894 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10895 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10896 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10897 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10898 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10899 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10900 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10901 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10902 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10903 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10904 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10905 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10906 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10907 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10908 (vfmlsl_low_u32): Ditto.
10909 (vfmlalq_low_u32): Ditto.
10910 (vfmlslq_low_u32): Ditto.
10911 (vfmlal_high_u32): Ditto.
10912 (vfmlsl_high_u32): Ditto.
10913 (vfmlalq_high_u32): Ditto.
10914 (vfmlslq_high_u32): Ditto.
10915 (vfmlal_lane_low_u32): Ditto.
10916 (vfmlsl_lane_low_u32): Ditto.
10917 (vfmlal_laneq_low_u32): Ditto.
10918 (vfmlsl_laneq_low_u32): Ditto.
10919 (vfmlalq_lane_low_u32): Ditto.
10920 (vfmlslq_lane_low_u32): Ditto.
10921 (vfmlalq_laneq_low_u32): Ditto.
10922 (vfmlslq_laneq_low_u32): Ditto.
10923 (vfmlal_lane_high_u32): Ditto.
10924 (vfmlsl_lane_high_u32): Ditto.
10925 (vfmlal_laneq_high_u32): Ditto.
10926 (vfmlsl_laneq_high_u32): Ditto.
10927 (vfmlalq_lane_high_u32): Ditto.
10928 (vfmlslq_lane_high_u32): Ditto.
10929 (vfmlalq_laneq_high_u32): Ditto.
10930 (vfmlslq_laneq_high_u32): Ditto.
10931 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10932 (AARCH64_FL_FOR_ARCH8_4): New.
10933 (AARCH64_ISA_F16FML): New ISA flag.
10934 (TARGET_F16FML): New feature flag for fp16fml.
10935 (doc/invoke.texi): Document new fp16fml option.
10937 2018-01-10 Michael Collison <michael.collison@arm.com>
10939 * config/aarch64/aarch64-builtins.c:
10940 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10941 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10942 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10943 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10944 (AARCH64_ISA_SHA3): New ISA flag.
10945 (TARGET_SHA3): New feature flag for sha3.
10946 * config/aarch64/iterators.md (sha512_op): New int attribute.
10947 (CRYPTO_SHA512): New int iterator.
10948 (UNSPEC_SHA512H): New unspec.
10949 (UNSPEC_SHA512H2): Ditto.
10950 (UNSPEC_SHA512SU0): Ditto.
10951 (UNSPEC_SHA512SU1): Ditto.
10952 * config/aarch64/aarch64-simd-builtins.def
10953 (aarch64_crypto_sha512hqv2di): New builtin.
10954 (aarch64_crypto_sha512h2qv2di): Ditto.
10955 (aarch64_crypto_sha512su0qv2di): Ditto.
10956 (aarch64_crypto_sha512su1qv2di): Ditto.
10957 (aarch64_eor3qv8hi): Ditto.
10958 (aarch64_rax1qv2di): Ditto.
10959 (aarch64_xarqv2di): Ditto.
10960 (aarch64_bcaxqv8hi): Ditto.
10961 * config/aarch64/aarch64-simd.md:
10962 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10963 (aarch64_crypto_sha512su0qv2di): Ditto.
10964 (aarch64_crypto_sha512su1qv2di): Ditto.
10965 (aarch64_eor3qv8hi): Ditto.
10966 (aarch64_rax1qv2di): Ditto.
10967 (aarch64_xarqv2di): Ditto.
10968 (aarch64_bcaxqv8hi): Ditto.
10969 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10970 (vsha512h2q_u64): Ditto.
10971 (vsha512su0q_u64): Ditto.
10972 (vsha512su1q_u64): Ditto.
10973 (veor3q_u16): Ditto.
10974 (vrax1q_u64): Ditto.
10975 (vxarq_u64): Ditto.
10976 (vbcaxq_u16): Ditto.
10977 * config/arm/types.md (crypto_sha512): New type attribute.
10978 (crypto_sha3): Ditto.
10979 (doc/invoke.texi): Document new sha3 option.
10981 2018-01-10 Michael Collison <michael.collison@arm.com>
10983 * config/aarch64/aarch64-builtins.c:
10984 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10985 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10986 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10987 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10988 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10989 (AARCH64_ISA_SM4): New ISA flag.
10990 (TARGET_SM4): New feature flag for sm4.
10991 * config/aarch64/aarch64-simd-builtins.def
10992 (aarch64_sm3ss1qv4si): Ditto.
10993 (aarch64_sm3tt1aq4si): Ditto.
10994 (aarch64_sm3tt1bq4si): Ditto.
10995 (aarch64_sm3tt2aq4si): Ditto.
10996 (aarch64_sm3tt2bq4si): Ditto.
10997 (aarch64_sm3partw1qv4si): Ditto.
10998 (aarch64_sm3partw2qv4si): Ditto.
10999 (aarch64_sm4eqv4si): Ditto.
11000 (aarch64_sm4ekeyqv4si): Ditto.
11001 * config/aarch64/aarch64-simd.md:
11002 (aarch64_sm3ss1qv4si): Ditto.
11003 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11004 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11005 (aarch64_sm4eqv4si): Ditto.
11006 (aarch64_sm4ekeyqv4si): Ditto.
11007 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11008 (sm3part_op): Ditto.
11009 (CRYPTO_SM3TT): Ditto.
11010 (CRYPTO_SM3PART): Ditto.
11011 (UNSPEC_SM3SS1): New unspec.
11012 (UNSPEC_SM3TT1A): Ditto.
11013 (UNSPEC_SM3TT1B): Ditto.
11014 (UNSPEC_SM3TT2A): Ditto.
11015 (UNSPEC_SM3TT2B): Ditto.
11016 (UNSPEC_SM3PARTW1): Ditto.
11017 (UNSPEC_SM3PARTW2): Ditto.
11018 (UNSPEC_SM4E): Ditto.
11019 (UNSPEC_SM4EKEY): Ditto.
11020 * config/aarch64/constraints.md (Ui2): New constraint.
11021 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11022 * config/arm/types.md (crypto_sm3): New type attribute.
11023 (crypto_sm4): Ditto.
11024 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11025 (vsm3tt1aq_u32): Ditto.
11026 (vsm3tt1bq_u32): Ditto.
11027 (vsm3tt2aq_u32): Ditto.
11028 (vsm3tt2bq_u32): Ditto.
11029 (vsm3partw1q_u32): Ditto.
11030 (vsm3partw2q_u32): Ditto.
11031 (vsm4eq_u32): Ditto.
11032 (vsm4ekeyq_u32): Ditto.
11033 (doc/invoke.texi): Document new sm4 option.
11035 2018-01-10 Michael Collison <michael.collison@arm.com>
11037 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11038 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11039 (AARCH64_FL_FOR_ARCH8_4): New.
11040 (AARCH64_FL_V8_4): New flag.
11041 (doc/invoke.texi): Document new armv8.4-a option.
11043 2018-01-10 Michael Collison <michael.collison@arm.com>
11045 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11046 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11047 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11048 * config/aarch64/aarch64-option-extension.def: Add
11049 AARCH64_OPT_EXTENSION of 'sha2'.
11050 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11051 (crypto): Disable sha2 and aes if crypto disabled.
11052 (crypto): Enable aes and sha2 if enabled.
11053 (simd): Disable sha2 and aes if simd disabled.
11054 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11056 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11057 (TARGET_SHA2): New feature flag for sha2.
11058 (TARGET_AES): New feature flag for aes.
11059 * config/aarch64/aarch64-simd.md:
11060 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11061 conditional on TARGET_AES.
11062 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11063 (aarch64_crypto_sha1hsi): Make pattern conditional
11065 (aarch64_crypto_sha1hv4si): Ditto.
11066 (aarch64_be_crypto_sha1hv4si): Ditto.
11067 (aarch64_crypto_sha1su1v4si): Ditto.
11068 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11069 (aarch64_crypto_sha1su0v4si): Ditto.
11070 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11071 (aarch64_crypto_sha256su0v4si): Ditto.
11072 (aarch64_crypto_sha256su1v4si): Ditto.
11073 (doc/invoke.texi): Document new aes and sha2 options.
11075 2018-01-10 Martin Sebor <msebor@redhat.com>
11077 PR tree-optimization/83781
11078 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11081 2018-01-11 Martin Sebor <msebor@gmail.com>
11082 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11084 PR tree-optimization/83501
11085 PR tree-optimization/81703
11087 * tree-ssa-strlen.c (get_string_cst): Rename...
11088 (get_string_len): ...to this. Handle global constants.
11089 (handle_char_store): Adjust.
11091 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11092 Jim Wilson <jimw@sifive.com>
11094 * config/riscv/riscv-protos.h (riscv_output_return): New.
11095 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11096 (riscv_attribute_table, riscv_output_return),
11097 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11098 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11099 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11101 (riscv_expand_prologue): Add early return for naked function.
11102 (riscv_expand_epilogue): Likewise.
11103 (riscv_function_ok_for_sibcall): Return false for naked function.
11104 (riscv_set_current_function): New.
11105 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11106 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11107 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11108 * doc/extend.texi (RISC-V Function Attributes): New.
11110 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11112 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11113 check for 128-bit long double before checking TCmode.
11114 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11115 128-bit long doubles before checking TFmode or TCmode.
11116 (FLOAT128_IBM_P): Likewise.
11118 2018-01-10 Martin Sebor <msebor@redhat.com>
11120 PR tree-optimization/83671
11121 * builtins.c (c_strlen): Unconditionally return zero for the empty
11123 Use -Warray-bounds for warnings.
11124 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11125 for non-constant array indices with COMPONENT_REF, arrays of
11126 arrays, and pointers to arrays.
11127 (gimple_fold_builtin_strlen): Determine and set length range for
11128 non-constant character arrays.
11130 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11132 PR middle-end/81897
11133 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11136 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11138 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11140 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11143 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11144 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11145 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11146 indexed_or_indirect_operand predicate.
11147 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11148 (*vsx_le_perm_load_v8hi): Likewise.
11149 (*vsx_le_perm_load_v16qi): Likewise.
11150 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11151 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11152 (*vsx_le_perm_store_v8hi): Likewise.
11153 (*vsx_le_perm_store_v16qi): Likewise.
11154 (eight unnamed splitters): Likewise.
11156 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11158 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11159 * config/rs6000/emmintrin.h: Likewise.
11160 * config/rs6000/mmintrin.h: Likewise.
11161 * config/rs6000/xmmintrin.h: Likewise.
11163 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11166 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11168 * tree.c (tree_nop_conversion): Return true for location wrapper
11170 (maybe_wrap_with_location): New function.
11171 (selftest::check_strip_nops): New function.
11172 (selftest::test_location_wrappers): New function.
11173 (selftest::tree_c_tests): Call it.
11174 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11175 (maybe_wrap_with_location): New decl.
11176 (EXPR_LOCATION_WRAPPER_P): New macro.
11177 (location_wrapper_p): New inline function.
11178 (tree_strip_any_location_wrapper): New inline function.
11180 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11183 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11184 stack_realign_offset for the largest alignment of stack slot
11186 (ix86_find_max_used_stack_alignment): New function.
11187 (ix86_finalize_stack_frame_flags): Use it. Set
11188 max_used_stack_alignment if we don't realign stack.
11189 * config/i386/i386.h (machine_function): Add
11190 max_used_stack_alignment.
11192 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11194 * config/arm/arm.opt (-mbranch-cost): New option.
11195 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11198 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11201 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11202 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11204 2018-01-10 Richard Biener <rguenther@suse.de>
11207 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11208 early out so it also covers the case where we have a non-NULL
11211 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11213 PR tree-optimization/83753
11214 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11215 for non-strided grouped accesses if the number of elements is 1.
11217 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11220 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11221 * i386.h (TARGET_USE_GATHER): Define.
11222 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11224 2018-01-10 Martin Liska <mliska@suse.cz>
11227 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11228 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11230 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11231 CLEANUP_NO_PARTITIONING is not set.
11233 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11235 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11236 for vectors, as a partial revert of r254296.
11237 * rtl.h (const_vec_p): Delete.
11238 (const_vec_duplicate_p): Don't test for vector CONSTs.
11239 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11240 * expmed.c (make_tree): Likewise.
11243 * common.md (E, F): Use CONSTANT_P instead of checking for
11245 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11246 checking for CONST_VECTOR.
11248 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11250 PR middle-end/83575
11251 * predict.c (force_edge_cold): Handle in more sane way edges
11252 with no prediction.
11254 2018-01-09 Carl Love <cel@us.ibm.com>
11256 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11258 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11259 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11260 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11261 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11262 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11263 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11264 * config/rs6000/rs6000-protos.h: Add extern defition for
11265 rs6000_generate_float2_double_code.
11266 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11268 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11269 (float2_v2df): Add define_expand.
11271 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11274 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11275 op_mode in the force_to_mode call.
11277 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11279 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11280 instead of checking each element individually.
11281 (aarch64_evpc_uzp): Likewise.
11282 (aarch64_evpc_zip): Likewise.
11283 (aarch64_evpc_ext): Likewise.
11284 (aarch64_evpc_rev): Likewise.
11285 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11286 instead of checking each element individually. Return true without
11288 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11289 whether all selected elements come from the same input, instead of
11290 checking each element individually. Remove calls to gen_rtx_REG,
11291 start_sequence and end_sequence and instead assert that no rtl is
11294 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11296 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11297 order of HIGH and CONST checks.
11299 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11301 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11302 if the destination isn't an SSA_NAME.
11304 2018-01-09 Richard Biener <rguenther@suse.de>
11306 PR tree-optimization/83668
11307 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11309 (canonicalize_loop_form): ... here, renamed from ...
11310 (canonicalize_loop_closed_ssa_form): ... this and amended to
11311 swap successor edges for loop exit blocks to make us use
11312 the RPO order we need for initial schedule generation.
11314 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11316 PR tree-optimization/64811
11317 * match.pd: When optimizing comparisons with Inf, avoid
11318 introducing or losing exceptions from comparisons with NaN.
11320 2018-01-09 Martin Liska <mliska@suse.cz>
11323 * asan.c (shadow_mem_size): Add gcc_assert.
11325 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11327 Don't save registers in main().
11330 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11331 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11332 * config/avr/avr.c (avr_set_current_function): Don't error if
11333 naked, OS_task or OS_main are specified at the same time.
11334 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11336 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11338 * common/config/avr/avr-common.c (avr_option_optimization_table):
11339 Switch on -mmain-is-OS_task for optimizing compilations.
11341 2018-01-09 Richard Biener <rguenther@suse.de>
11343 PR tree-optimization/83572
11344 * graphite.c: Include cfganal.h.
11345 (graphite_transform_loops): Connect infinite loops to exit
11346 and remove fake edges at the end.
11348 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11350 * ipa-inline.c (edge_badness): Revert accidental checkin.
11352 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11355 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11356 symbols; not inline clones.
11358 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11361 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11362 hard registers. Formatting fixes.
11364 PR preprocessor/83722
11365 * gcc.c (try_generate_repro): Pass
11366 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11367 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11370 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11371 Kito Cheng <kito.cheng@gmail.com>
11373 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11374 (riscv_leaf_function_p): Delete.
11375 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11377 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11379 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11381 (do_ifelse): New function.
11382 (do_isel): New function.
11383 (do_sub3): New function.
11384 (do_add3): New function.
11385 (do_load_mask_compare): New function.
11386 (do_overlap_load_compare): New function.
11387 (expand_compare_loop): New function.
11388 (expand_block_compare): Call expand_compare_loop() when appropriate.
11389 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11390 option description.
11391 (-mblock-compare-inline-loop-limit): New option.
11393 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11396 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11397 Reverse order of second and third operands in first alternative.
11398 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11399 of first and second elements in UNSPEC_VPERMR vector.
11400 (altivec_expand_vec_perm_le): Likewise.
11402 2018-01-08 Jeff Law <law@redhat.com>
11404 PR rtl-optimizatin/81308
11405 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11406 (process_switch): If group_case_labels makes a change, then set
11408 (pass_convert_switch::execute): If a switch is converted, then
11409 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11411 PR rtl-optimization/81308
11412 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11415 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11417 PR target/83663 - Revert r255946
11418 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11419 generation for cases where splatting a value is not useful.
11420 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11421 across a vec_duplicate and a paradoxical subreg forming a vector
11422 mode to a vec_concat.
11424 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11426 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11427 -march=armv8.3-a variants.
11428 * config/arm/t-multilib: Likewise.
11429 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11431 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11433 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11435 (cceq_ior_compare_complement): Give it a name so I can use it, and
11436 change boolean_or_operator predicate to boolean_operator so it can
11437 be used to generate a crand.
11438 (eqne): New code iterator.
11439 (bd/bd_neg): New code_attrs.
11440 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11441 a single define_insn.
11442 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11443 decrement (bdnzt/bdnzf/bdzt/bdzf).
11444 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11445 with the new names of the branch decrement patterns, and added the
11446 names of the branch decrement conditional patterns.
11448 2018-01-08 Richard Biener <rguenther@suse.de>
11450 PR tree-optimization/83563
11451 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11454 2018-01-08 Richard Biener <rguenther@suse.de>
11456 PR middle-end/83713
11457 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11459 2018-01-08 Richard Biener <rguenther@suse.de>
11461 PR tree-optimization/83685
11462 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11463 references to abnormals.
11465 2018-01-08 Richard Biener <rguenther@suse.de>
11468 * dwarf2out.c (output_indirect_strings): Handle empty
11469 skeleton_debug_str_hash.
11470 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11472 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11474 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11475 (emit_store_direct): Likewise.
11476 (arc_trampoline_adjust_address): Likewise.
11477 (arc_asm_trampoline_template): New function.
11478 (arc_initialize_trampoline): Use asm_trampoline_template.
11479 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11480 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11481 * config/arc/arc.md (flush_icache): Delete pattern.
11483 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11485 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11486 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11489 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11492 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11493 by not USED_FOR_TARGET.
11494 (make_pass_resolve_sw_modes): Likewise.
11496 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11498 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11501 2018-01-08 Richard Biener <rguenther@suse.de>
11503 PR middle-end/83580
11504 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11506 2018-01-08 Richard Biener <rguenther@suse.de>
11508 PR middle-end/83517
11509 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11511 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11513 PR middle-end/81897
11514 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11515 basic blocks with a small number of successors.
11516 (convert_control_dep_chain_into_preds): Improve handling of
11518 (dump_predicates): Split apart into...
11519 (dump_pred_chain): ...here...
11520 (dump_pred_info): ...and here.
11521 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11522 (can_chain_union_be_invalidated_p): Improve check for invalidation
11524 (uninit_uses_cannot_happen): Avoid unnecessary if
11525 convert_control_dep_chain_into_preds yielded nothing.
11527 2018-01-06 Martin Sebor <msebor@redhat.com>
11529 PR tree-optimization/83640
11530 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11531 subtracting negative offset from size.
11532 (builtin_access::overlap): Adjust offset bounds of the access to fall
11533 within the size of the object if possible.
11535 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11537 PR rtl-optimization/83699
11538 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11539 extract_bit_field_as_subreg to cases in which the extracted
11540 value is also a vector.
11542 * lra-constraints.c (process_alt_operands): Test for the equivalence
11543 substitutions when detecting a possible reload cycle.
11545 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11548 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11549 by default if flag_selective_schedling{,2}. Formatting fixes.
11551 PR rtl-optimization/83682
11552 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11553 if it has non-VECTOR_MODE element mode.
11554 (vec_duplicate_p): Likewise.
11556 PR middle-end/83694
11557 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11558 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11560 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11563 * config/i386/i386-builtin.def
11564 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11565 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11566 Require also OPTION_MASK_ISA_AVX512F in addition to
11567 OPTION_MASK_ISA_GFNI.
11568 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11569 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11570 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11571 to OPTION_MASK_ISA_GFNI.
11572 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11573 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11574 OPTION_MASK_ISA_AVX512BW.
11575 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11576 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11577 addition to OPTION_MASK_ISA_GFNI.
11578 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11579 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11580 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11581 to OPTION_MASK_ISA_GFNI.
11582 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11583 a requirement for all ISAs rather than any of them with a few
11585 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11587 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11588 bitmasks to be enabled with 3 exceptions, instead of requiring any
11589 enabled ISA with lots of exceptions.
11590 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11591 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11592 Change avx512bw in isa attribute to avx512f.
11593 * config/i386/sgxintrin.h: Add license boilerplate.
11594 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11595 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11596 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11597 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11599 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11600 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11601 temporarily sse2 rather than sse if not enabled already.
11604 * config/i386/sse.md (VI248_VLBW): Rename to ...
11605 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11606 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11607 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11608 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11609 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11610 mode iterator instead of VI248_VLBW.
11612 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11614 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11615 (record_modified): Skip clobbers; add debug output.
11616 (param_change_prob): Use sreal frequencies.
11618 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11620 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11621 punt for user-aligned variables.
11623 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11625 * tree-chrec.c (chrec_contains_symbols): Return true for
11628 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11631 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11632 of (x|y) == x for BICS pattern.
11634 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11636 PR tree-optimization/83605
11637 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11638 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11641 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11643 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11644 * config/epiphany/rtems.h: New file.
11646 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11647 Uros Bizjak <ubizjak@gmail.com>
11650 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11651 QIreg_operand instead of register_operand predicate.
11652 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11653 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11654 comments instead of -fmitigate[-_]rop.
11656 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11659 * cgraphunit.c (symbol_table::compile): Switch to text_section
11660 before calling assembly_start debug hook.
11661 * run-rtl-passes.c (run_rtl_passes): Likewise.
11664 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11666 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11667 range_int_cst_p rather than !symbolic_range_p before calling
11668 extract_range_from_multiplicative_op_1.
11670 2018-01-04 Jeff Law <law@redhat.com>
11672 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11673 redundant test in assertion.
11675 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11677 * doc/rtl.texi: Document machine_mode wrapper classes.
11679 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11681 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11682 using tree_to_uhwi.
11684 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11686 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11687 the VEC_PERM_EXPR fold to fail.
11689 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11692 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11693 to switched_sections.
11695 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11698 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11699 test for d.testing.
11701 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11704 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11705 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11707 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11710 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11711 is BLKmode and bitpos not zero or mode change is needed.
11713 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11716 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11719 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11722 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11723 instead of MULT rtx. Update all corresponding splitters.
11724 (*saddl_se): Ditto.
11725 (*ssub<modesuffix>): Ditto.
11726 (*ssubl_se): Ditto.
11727 (*cmp_sadd_di): Update split patterns.
11728 (*cmp_sadd_si): Ditto.
11729 (*cmp_sadd_sidi): Ditto.
11730 (*cmp_ssub_di): Ditto.
11731 (*cmp_ssub_si): Ditto.
11732 (*cmp_ssub_sidi): Ditto.
11733 * config/alpha/predicates.md (const23_operand): New predicate.
11734 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11735 Look for ASHIFT, not MULT inner operand.
11736 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11738 2018-01-04 Martin Liska <mliska@suse.cz>
11740 PR gcov-profile/83669
11741 * gcov.c (output_intermediate_file): Add version to intermediate
11743 * doc/gcov.texi: Document new field 'version' in intermediate
11744 file format. Fix location of '-k' option of gcov command.
11746 2018-01-04 Martin Liska <mliska@suse.cz>
11749 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11751 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11753 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11755 2018-01-03 Martin Sebor <msebor@redhat.com>
11757 PR tree-optimization/83655
11758 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11759 checking calls with invalid arguments.
11761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11763 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11764 (vectorizable_mask_load_store): Delete.
11765 (vectorizable_call): Return false for masked loads and stores.
11766 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11767 instead of gimple_assign_rhs1.
11768 (vectorizable_load): Handle IFN_MASK_LOAD.
11769 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11773 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11775 (vectorizable_mask_load_store): ...here.
11776 (vectorizable_load): ...and here.
11778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11780 * tree-vect-stmts.c (vect_build_all_ones_mask)
11781 (vect_build_zero_merge_argument): New functions, split out from...
11782 (vectorizable_load): ...here.
11784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11786 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11788 (vectorizable_mask_load_store): ...here.
11789 (vectorizable_store): ...and here.
11791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11793 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11795 (vectorizable_mask_load_store): ...here.
11797 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11799 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11800 (vect_model_store_cost): Take a vec_load_store_type instead of a
11802 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11803 (vect_model_store_cost): Take a vec_load_store_type instead of a
11805 (vectorizable_mask_load_store): Update accordingly.
11806 (vectorizable_store): Likewise.
11807 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11811 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11812 IFN_MASK_LOAD calls here rather than...
11813 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11815 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11816 Alan Hayward <alan.hayward@arm.com>
11817 David Sherwood <david.sherwood@arm.com>
11819 * expmed.c (extract_bit_field_1): For vector extracts,
11820 fall back to extract_bit_field_as_subreg if vec_extract
11823 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11824 Alan Hayward <alan.hayward@arm.com>
11825 David Sherwood <david.sherwood@arm.com>
11827 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11828 they are variable or constant sized.
11829 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11830 slots for constant-sized data.
11832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11833 Alan Hayward <alan.hayward@arm.com>
11834 David Sherwood <david.sherwood@arm.com>
11836 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11837 handling COND_EXPRs with boolean comparisons, try to find a better
11838 basis for the mask type than the boolean itself.
11840 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11842 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11843 is calculated and how it can be overridden.
11844 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11845 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11847 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11851 Alan Hayward <alan.hayward@arm.com>
11852 David Sherwood <david.sherwood@arm.com>
11854 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11855 Remove the mode argument.
11856 (aarch64_simd_valid_immediate): Remove the mode and inverse
11858 * config/aarch64/iterators.md (bitsize): New iterator.
11859 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11860 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11861 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11862 aarch64_simd_valid_immediate.
11863 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11864 (aarch64_reg_or_bic_imm): Likewise.
11865 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11866 with an insn_type enum and msl with a modifier_type enum.
11867 Replace element_width with a scalar_mode. Change the shift
11868 to unsigned int. Add constructors for scalar_float_mode and
11869 scalar_int_mode elements.
11870 (aarch64_vect_float_const_representable_p): Delete.
11871 (aarch64_can_const_movi_rtx_p)
11872 (aarch64_simd_scalar_immediate_valid_for_move)
11873 (aarch64_simd_make_constant): Update call to
11874 aarch64_simd_valid_immediate.
11875 (aarch64_advsimd_valid_immediate_hs): New function.
11876 (aarch64_advsimd_valid_immediate): Likewise.
11877 (aarch64_simd_valid_immediate): Remove mode and inverse
11878 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11879 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11880 and aarch64_float_const_representable_p on the result.
11881 (aarch64_output_simd_mov_immediate): Remove mode argument.
11882 Update call to aarch64_simd_valid_immediate and use of
11883 simd_immediate_info.
11884 (aarch64_output_scalar_simd_mov_immediate): Update call
11887 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11888 Alan Hayward <alan.hayward@arm.com>
11889 David Sherwood <david.sherwood@arm.com>
11891 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11892 (mode_nunits): Likewise CONST_MODE_NUNITS.
11893 * machmode.def (ADJUST_NUNITS): Document.
11894 * genmodes.c (mode_data::need_nunits_adj): New field.
11895 (blank_mode): Update accordingly.
11896 (adj_nunits): New variable.
11897 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11899 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11900 listed in adj_nunits.
11901 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11902 listed in adj_nunits. Don't emit case statements for such modes.
11903 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11904 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11905 nothing if adj_nunits is nonnull.
11906 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11907 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11908 (emit_mode_fbit): Update use of print_maybe_const_decl.
11909 (emit_move_size): Likewise. Treat the array as non-const
11911 (emit_mode_adjustments): Handle adj_nunits.
11913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11915 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11916 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11917 (VECTOR_MODES): Use it.
11918 (make_vector_modes): Take the prefix as an argument.
11920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11921 Alan Hayward <alan.hayward@arm.com>
11922 David Sherwood <david.sherwood@arm.com>
11924 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11925 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11926 for MODE_VECTOR_BOOL.
11927 * machmode.def (VECTOR_BOOL_MODE): Document.
11928 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11929 (make_vector_bool_mode): New function.
11930 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11932 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11933 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11935 * stor-layout.c (int_mode_for_mode): Likewise.
11936 * tree.c (build_vector_type_for_mode): Likewise.
11937 * varasm.c (output_constant_pool_2): Likewise.
11938 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11939 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11940 for MODE_VECTOR_BOOL.
11941 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11942 of mode class checks.
11943 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11944 instead of a list of mode class checks.
11945 (expand_vector_scalar_condition): Likewise.
11946 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11949 Alan Hayward <alan.hayward@arm.com>
11950 David Sherwood <david.sherwood@arm.com>
11952 * machmode.h (mode_size): Change from unsigned short to
11954 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11955 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11956 or if measurement_type is not polynomial.
11957 (fixed_size_mode::includes_p): Check for constant-sized modes.
11958 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11959 return a poly_uint16 rather than an unsigned short.
11960 (emit_mode_size): Change the type of mode_size from unsigned short
11961 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11962 (emit_mode_adjustments): Cope with polynomial vector sizes.
11963 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11965 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11967 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11968 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11969 * caller-save.c (setup_save_areas): Likewise.
11970 (replace_reg_with_saved_mem): Likewise.
11971 * calls.c (emit_library_call_value_1): Likewise.
11972 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11973 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11974 (gen_lowpart_for_combine): Likewise.
11975 * convert.c (convert_to_integer_1): Likewise.
11976 * cse.c (equiv_constant, cse_insn): Likewise.
11977 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11978 (cselib_subst_to_values): Likewise.
11979 * dce.c (word_dce_process_block): Likewise.
11980 * df-problems.c (df_word_lr_mark_ref): Likewise.
11981 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11982 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11983 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11984 (rtl_for_decl_location): Likewise.
11985 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11986 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11987 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11988 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11989 (expand_expr_real_1): Likewise.
11990 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11991 (pad_below): Likewise.
11992 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11993 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11994 * ira.c (get_subreg_tracking_sizes): Likewise.
11995 * ira-build.c (ira_create_allocno_objects): Likewise.
11996 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11997 (ira_sort_regnos_for_alter_reg): Likewise.
11998 * ira-costs.c (record_operand_costs): Likewise.
11999 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12000 (resolve_simple_move): Likewise.
12001 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12002 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12003 (lra_constraints): Likewise.
12004 (CONST_POOL_OK_P): Reject variable-sized modes.
12005 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12006 (add_pseudo_to_slot, lra_spill): Likewise.
12007 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12008 * optabs-query.c (get_best_extraction_insn): Likewise.
12009 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12010 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12011 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12012 * recog.c (offsettable_address_addr_space_p): Likewise.
12013 * regcprop.c (maybe_mode_change): Likewise.
12014 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12015 * regrename.c (build_def_use): Likewise.
12016 * regstat.c (dump_reg_info): Likewise.
12017 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12018 (find_reloads, find_reloads_subreg_address): Likewise.
12019 * reload1.c (eliminate_regs_1): Likewise.
12020 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12021 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12022 (simplify_binary_operation_1, simplify_subreg): Likewise.
12023 * targhooks.c (default_function_arg_padding): Likewise.
12024 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12025 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12026 (verify_gimple_assign_ternary): Likewise.
12027 * tree-inline.c (estimate_move_cost): Likewise.
12028 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12029 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12030 (get_address_cost_ainc): Likewise.
12031 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12032 (vect_supportable_dr_alignment): Likewise.
12033 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12034 (vectorizable_reduction): Likewise.
12035 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12036 (vectorizable_operation, vectorizable_load): Likewise.
12037 * tree.c (build_same_sized_truth_vector_type): Likewise.
12038 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12039 * var-tracking.c (emit_note_insn_var_location): Likewise.
12040 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12041 (ADDR_VEC_ALIGN): Likewise.
12043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12044 Alan Hayward <alan.hayward@arm.com>
12045 David Sherwood <david.sherwood@arm.com>
12047 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12049 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12050 or if measurement_type is polynomial.
12051 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12052 * combine.c (make_extraction): Likewise.
12053 * dse.c (find_shift_sequence): Likewise.
12054 * dwarf2out.c (mem_loc_descriptor): Likewise.
12055 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12056 (extract_bit_field, extract_low_bits): Likewise.
12057 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12058 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12059 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12060 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12061 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12062 * reload.c (find_reloads): Likewise.
12063 * reload1.c (alter_reg): Likewise.
12064 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12065 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12066 * tree-if-conv.c (predicate_mem_writes): Likewise.
12067 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12068 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12069 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12070 * valtrack.c (dead_debug_insert_temp): Likewise.
12071 * varasm.c (mergeable_constant_section): Likewise.
12072 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12074 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12075 Alan Hayward <alan.hayward@arm.com>
12076 David Sherwood <david.sherwood@arm.com>
12078 * expr.c (expand_assignment): Cope with polynomial mode sizes
12079 when assigning to a CONCAT.
12081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12082 Alan Hayward <alan.hayward@arm.com>
12083 David Sherwood <david.sherwood@arm.com>
12085 * machmode.h (mode_precision): Change from unsigned short to
12087 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12089 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12090 or if measurement_type is not polynomial.
12091 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12092 in which the mode is already known to be a scalar_int_mode.
12093 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12094 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12096 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12097 for GET_MODE_PRECISION.
12098 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12099 for GET_MODE_PRECISION.
12100 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12102 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12103 (expand_field_assignment, make_extraction): Likewise.
12104 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12105 (get_last_value): Likewise.
12106 * convert.c (convert_to_integer_1): Likewise.
12107 * cse.c (cse_insn): Likewise.
12108 * expr.c (expand_expr_real_1): Likewise.
12109 * lra-constraints.c (simplify_operand_subreg): Likewise.
12110 * optabs-query.c (can_atomic_load_p): Likewise.
12111 * optabs.c (expand_atomic_load): Likewise.
12112 (expand_atomic_store): Likewise.
12113 * ree.c (combine_reaching_defs): Likewise.
12114 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12115 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12116 * tree.h (type_has_mode_precision_p): Likewise.
12117 * ubsan.c (instrument_si_overflow): Likewise.
12119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12120 Alan Hayward <alan.hayward@arm.com>
12121 David Sherwood <david.sherwood@arm.com>
12123 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12124 polynomial numbers of units.
12125 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12126 (valid_vector_subparts_p): New function.
12127 (build_vector_type): Remove temporary shim and take the number
12128 of units as a poly_uint64 rather than an int.
12129 (build_opaque_vector_type): Take the number of units as a
12130 poly_uint64 rather than an int.
12131 * tree.c (build_vector_from_ctor): Handle polynomial
12132 TYPE_VECTOR_SUBPARTS.
12133 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12134 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12135 (build_vector_from_val): If the number of units is variable,
12136 use build_vec_duplicate_cst for constant operands and
12137 VEC_DUPLICATE_EXPR otherwise.
12138 (make_vector_type): Remove temporary is_constant ().
12139 (build_vector_type, build_opaque_vector_type): Take the number of
12140 units as a poly_uint64 rather than an int.
12141 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12143 * cfgexpand.c (expand_debug_expr): Likewise.
12144 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12145 (store_constructor, expand_expr_real_1): Likewise.
12146 (const_scalar_mask_from_tree): Likewise.
12147 * fold-const-call.c (fold_const_reduction): Likewise.
12148 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12149 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12150 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12151 (fold_relational_const): Likewise.
12152 (native_interpret_vector): Likewise. Change the size from an
12153 int to an unsigned int.
12154 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12155 TYPE_VECTOR_SUBPARTS.
12156 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12157 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12158 duplicating a non-constant operand into a variable-length vector.
12159 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12160 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12161 * ipa-icf.c (sem_variable::equals): Likewise.
12162 * match.pd: Likewise.
12163 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12164 * print-tree.c (print_node): Likewise.
12165 * stor-layout.c (layout_type): Likewise.
12166 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12167 * tree-cfg.c (verify_gimple_comparison): Likewise.
12168 (verify_gimple_assign_binary): Likewise.
12169 (verify_gimple_assign_ternary): Likewise.
12170 (verify_gimple_assign_single): Likewise.
12171 * tree-pretty-print.c (dump_generic_node): Likewise.
12172 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12173 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12174 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12175 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12176 (vect_shift_permute_load_chain): Likewise.
12177 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12178 (expand_vector_condition, optimize_vector_constructor): Likewise.
12179 (lower_vec_perm, get_compute_type): Likewise.
12180 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12181 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12182 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12183 (vect_recog_mask_conversion_pattern): Likewise.
12184 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12185 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12186 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12187 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12188 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12189 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12190 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12191 (supportable_widening_operation): Likewise.
12192 (supportable_narrowing_operation): Likewise.
12193 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12195 * varasm.c (output_constant): Likewise.
12197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12198 Alan Hayward <alan.hayward@arm.com>
12199 David Sherwood <david.sherwood@arm.com>
12201 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12202 so that both the length == 3 and length != 3 cases set up their
12203 own permute vectors. Add comments explaining why we know the
12204 number of elements is constant.
12205 (vect_permute_load_chain): Likewise.
12207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12208 Alan Hayward <alan.hayward@arm.com>
12209 David Sherwood <david.sherwood@arm.com>
12211 * machmode.h (mode_nunits): Change from unsigned char to
12213 (ONLY_FIXED_SIZE_MODES): New macro.
12214 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12215 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12216 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12218 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12219 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12220 or if measurement_type is not polynomial.
12221 * genmodes.c (ZERO_COEFFS): New macro.
12222 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12224 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12225 Use ZERO_COEFFS when emitting initializers.
12226 * data-streamer.h (bp_pack_poly_value): New function.
12227 (bp_unpack_poly_value): Likewise.
12228 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12229 for GET_MODE_NUNITS.
12230 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12231 for GET_MODE_NUNITS.
12232 * tree.c (make_vector_type): Remove temporary shim and make
12233 the real function take the number of units as a poly_uint64
12234 rather than an int.
12235 (build_vector_type_for_mode): Handle polynomial nunits.
12236 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12237 * emit-rtl.c (const_vec_series_p_1): Likewise.
12238 (gen_rtx_CONST_VECTOR): Likewise.
12239 * fold-const.c (test_vec_duplicate_folding): Likewise.
12240 * genrecog.c (validate_pattern): Likewise.
12241 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12242 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12243 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12244 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12245 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12246 * rtlanal.c (subreg_get_info): Likewise.
12247 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12248 (vect_grouped_load_supported): Likewise.
12249 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12250 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12251 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12252 (simplify_const_unary_operation, simplify_binary_operation_1)
12253 (simplify_const_binary_operation, simplify_ternary_operation)
12254 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12255 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12256 instead of CONST_VECTOR_NUNITS.
12257 * varasm.c (output_constant_pool_2): Likewise.
12258 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12259 explicit-encoded elements in the XVEC for variable-length vectors.
12261 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12263 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12266 Alan Hayward <alan.hayward@arm.com>
12267 David Sherwood <david.sherwood@arm.com>
12269 * coretypes.h (fixed_size_mode): Declare.
12270 (fixed_size_mode_pod): New typedef.
12271 * builtins.h (target_builtins::x_apply_args_mode)
12272 (target_builtins::x_apply_result_mode): Change type to
12273 fixed_size_mode_pod.
12274 * builtins.c (apply_args_size, apply_result_size, result_vector)
12275 (expand_builtin_apply_args_1, expand_builtin_apply)
12276 (expand_builtin_return): Update accordingly.
12278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12280 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12281 * cselib.c (cselib_hash_rtx): Likewise.
12282 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12283 CONST_VECTOR encoding.
12285 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12286 Jeff Law <law@redhat.com>
12289 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12290 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12291 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12292 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12295 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12296 explicitly probe *sp in a noreturn function if there were any callee
12297 register saves or frame pointer is needed.
12299 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12302 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12303 BLKmode for ternary, binary or unary expressions.
12306 * var-tracking.c (delete_vta_debug_insn): New inline function.
12307 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12308 insns from get_insns () to NULL instead of each bb separately.
12309 Use delete_vta_debug_insn. No longer static.
12310 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12311 delete_vta_debug_insns callers.
12312 * rtl.h (delete_vta_debug_insns): Declare.
12313 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12314 instead of variable_tracking_main.
12316 2018-01-03 Martin Sebor <msebor@redhat.com>
12318 PR tree-optimization/83603
12319 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12320 arguments past the endof the argument list in functions declared
12321 without a prototype.
12322 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12323 Avoid checking when arguments are null.
12325 2018-01-03 Martin Sebor <msebor@redhat.com>
12328 * doc/extend.texi (attribute const): Fix a typo.
12329 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12330 issuing -Wsuggest-attribute for void functions.
12332 2018-01-03 Martin Sebor <msebor@redhat.com>
12334 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12335 offset_int::from instead of wide_int::to_shwi.
12336 (maybe_diag_overlap): Remove assertion.
12337 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12338 * gimple-ssa-sprintf.c (format_directive): Same.
12339 (parse_directive): Same.
12340 (sprintf_dom_walker::compute_format_length): Same.
12341 (try_substitute_return_value): Same.
12343 2018-01-03 Jeff Law <law@redhat.com>
12345 PR middle-end/83654
12346 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12347 non-constant residual for zero at runtime and avoid probing in
12348 that case. Reorganize code for trailing problem to mirror handling
12351 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12353 PR tree-optimization/83501
12354 * tree-ssa-strlen.c (get_string_cst): New.
12355 (handle_char_store): Call get_string_cst.
12357 2018-01-03 Martin Liska <mliska@suse.cz>
12359 PR tree-optimization/83593
12360 * tree-ssa-strlen.c: Include tree-cfg.h.
12361 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12362 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12363 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12365 (strlen_dom_walker::before_dom_children): Call
12366 gimple_purge_dead_eh_edges. Dump tranformation with details
12368 (strlen_dom_walker::before_dom_children): Update call by adding
12369 new argument cleanup_eh.
12370 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12372 2018-01-03 Martin Liska <mliska@suse.cz>
12375 * cif-code.def (VARIADIC_THUNK): New enum value.
12376 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12379 2018-01-03 Jan Beulich <jbeulich@suse.com>
12381 * sse.md (mov<mode>_internal): Tighten condition for when to use
12382 vmovdqu<ssescalarsize> for TI and OI modes.
12384 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12386 Update copyright years.
12388 2018-01-03 Martin Liska <mliska@suse.cz>
12391 * ipa-visibility.c (function_and_variable_visibility): Skip
12392 functions with noipa attribure.
12394 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12396 * gcc.c (process_command): Update copyright notice dates.
12397 * gcov-dump.c (print_version): Ditto.
12398 * gcov.c (print_version): Ditto.
12399 * gcov-tool.c (print_version): Ditto.
12400 * gengtype.c (create_file): Ditto.
12401 * doc/cpp.texi: Bump @copying's copyright year.
12402 * doc/cppinternals.texi: Ditto.
12403 * doc/gcc.texi: Ditto.
12404 * doc/gccint.texi: Ditto.
12405 * doc/gcov.texi: Ditto.
12406 * doc/install.texi: Ditto.
12407 * doc/invoke.texi: Ditto.
12409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12411 * vector-builder.h (vector_builder::m_full_nelts): Change from
12412 unsigned int to poly_uint64.
12413 (vector_builder::full_nelts): Update prototype accordingly.
12414 (vector_builder::new_vector): Likewise.
12415 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12416 (vector_builder::operator ==): Likewise.
12417 (vector_builder::finalize): Likewise.
12418 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12419 Take the number of elements as a poly_uint64 rather than an
12421 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12422 from unsigned int to poly_uint64.
12423 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12424 (vec_perm_indices::new_vector): Likewise.
12425 (vec_perm_indices::length): Likewise.
12426 (vec_perm_indices::nelts_per_input): Likewise.
12427 (vec_perm_indices::input_nelts): Likewise.
12428 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12429 number of elements per input as a poly_uint64 rather than an
12430 unsigned int. Use the original encoding for variable-length
12431 vectors, rather than clamping each individual element.
12432 For the second and subsequent elements in each pattern,
12433 clamp the step and base before clamping their sum.
12434 (vec_perm_indices::series_p): Handle polynomial element counts.
12435 (vec_perm_indices::all_in_range_p): Likewise.
12436 (vec_perm_indices_to_tree): Likewise.
12437 (vec_perm_indices_to_rtx): Likewise.
12438 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12439 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12440 (tree_vector_builder::new_binary_operation): Handle polynomial
12441 element counts. Return false if we need to know the number
12442 of elements at compile time.
12443 * fold-const.c (fold_vec_perm): Punt if the number of elements
12444 isn't known at compile time.
12446 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12448 * vec-perm-indices.h (vec_perm_builder): Change element type
12449 from HOST_WIDE_INT to poly_int64.
12450 (vec_perm_indices::element_type): Update accordingly.
12451 (vec_perm_indices::clamp): Handle polynomial element_types.
12452 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12453 (vec_perm_indices::all_in_range_p): Likewise.
12454 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12456 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12457 polynomial vec_perm_indices element types.
12458 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12459 * fold-const.c (fold_vec_perm): Likewise.
12460 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12461 * tree-vect-generic.c (lower_vec_perm): Likewise.
12462 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12463 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12464 element type to HOST_WIDE_INT.
12466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12467 Alan Hayward <alan.hayward@arm.com>
12468 David Sherwood <david.sherwood@arm.com>
12470 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12471 rather than an int. Use plus_constant.
12472 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12473 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12476 Alan Hayward <alan.hayward@arm.com>
12477 David Sherwood <david.sherwood@arm.com>
12479 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12480 a HOST_WIDE_INT to a poly_int64.
12482 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12483 Alan Hayward <alan.hayward@arm.com>
12484 David Sherwood <david.sherwood@arm.com>
12486 * calls.c (load_register_parameters): Cope with polynomial
12487 mode sizes. Require a constant size for BLKmode parameters
12488 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12489 forces a parameter to be padded at the lsb end in order to
12490 fill a complete number of words, require the parameter size
12491 to be ordered wrt UNITS_PER_WORD.
12493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12494 Alan Hayward <alan.hayward@arm.com>
12495 David Sherwood <david.sherwood@arm.com>
12497 * reload1.c (spill_stack_slot_width): Change element type
12498 from unsigned int to poly_uint64_pod.
12499 (alter_reg): Treat mode sizes as polynomial.
12501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12502 Alan Hayward <alan.hayward@arm.com>
12503 David Sherwood <david.sherwood@arm.com>
12505 * reload.c (complex_word_subreg_p): New function.
12506 (reload_inner_reg_of_subreg, push_reload): Use it.
12508 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12509 Alan Hayward <alan.hayward@arm.com>
12510 David Sherwood <david.sherwood@arm.com>
12512 * lra-constraints.c (process_alt_operands): Reject matched
12513 operands whose sizes aren't ordered.
12514 (match_reload): Refer to this check here.
12516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12517 Alan Hayward <alan.hayward@arm.com>
12518 David Sherwood <david.sherwood@arm.com>
12520 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12521 that the mode size is in the set {1, 2, 4, 8, 16}.
12523 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12524 Alan Hayward <alan.hayward@arm.com>
12525 David Sherwood <david.sherwood@arm.com>
12527 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12528 Use plus_constant instead of gen_rtx_PLUS.
12530 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12531 Alan Hayward <alan.hayward@arm.com>
12532 David Sherwood <david.sherwood@arm.com>
12534 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12535 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12536 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12537 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12538 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12539 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12540 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12541 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12542 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12543 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12545 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12546 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12547 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12548 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12549 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12550 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12551 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12552 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12553 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12554 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12556 * expr.c (emit_move_resolve_push): Treat the input and result
12557 of PUSH_ROUNDING as a poly_int64.
12558 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12559 (emit_push_insn): Likewise.
12560 * lra-eliminations.c (mark_not_eliminable): Likewise.
12561 * recog.c (push_operand): Likewise.
12562 * reload1.c (elimination_effects): Likewise.
12563 * rtlanal.c (nonzero_bits1): Likewise.
12564 * calls.c (store_one_arg): Likewise. Require the padding to be
12565 known at compile time.
12567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12568 Alan Hayward <alan.hayward@arm.com>
12569 David Sherwood <david.sherwood@arm.com>
12571 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12572 Use plus_constant instead of gen_rtx_PLUS.
12574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12575 Alan Hayward <alan.hayward@arm.com>
12576 David Sherwood <david.sherwood@arm.com>
12578 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12579 rather than an int.
12581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12582 Alan Hayward <alan.hayward@arm.com>
12583 David Sherwood <david.sherwood@arm.com>
12585 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12586 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12587 via stack temporaries. Treat the mode size as polynomial too.
12589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12590 Alan Hayward <alan.hayward@arm.com>
12591 David Sherwood <david.sherwood@arm.com>
12593 * expr.c (expand_expr_real_2): When handling conversions involving
12594 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12595 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12596 as a poly_uint64 too.
12598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12599 Alan Hayward <alan.hayward@arm.com>
12600 David Sherwood <david.sherwood@arm.com>
12602 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12604 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12605 Alan Hayward <alan.hayward@arm.com>
12606 David Sherwood <david.sherwood@arm.com>
12608 * combine.c (can_change_dest_mode): Handle polynomial
12609 REGMODE_NATURAL_SIZE.
12610 * expmed.c (store_bit_field_1): Likewise.
12611 * expr.c (store_constructor): Likewise.
12612 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12613 and polynomial REGMODE_NATURAL_SIZE.
12614 (gen_lowpart_common): Likewise.
12615 * reginfo.c (record_subregs_of_mode): Likewise.
12616 * rtlanal.c (read_modify_subreg_p): Likewise.
12618 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12619 Alan Hayward <alan.hayward@arm.com>
12620 David Sherwood <david.sherwood@arm.com>
12622 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12623 numbers of elements.
12625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12626 Alan Hayward <alan.hayward@arm.com>
12627 David Sherwood <david.sherwood@arm.com>
12629 * match.pd: Cope with polynomial numbers of vector elements.
12631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12632 Alan Hayward <alan.hayward@arm.com>
12633 David Sherwood <david.sherwood@arm.com>
12635 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12636 in a POINTER_PLUS_EXPR.
12638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12639 Alan Hayward <alan.hayward@arm.com>
12640 David Sherwood <david.sherwood@arm.com>
12642 * omp-simd-clone.c (simd_clone_subparts): New function.
12643 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12644 (ipa_simd_modify_function_body): Likewise.
12646 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12647 Alan Hayward <alan.hayward@arm.com>
12648 David Sherwood <david.sherwood@arm.com>
12650 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12651 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12652 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12653 (expand_vector_condition, vector_element): Likewise.
12654 (subparts_gt): New function.
12655 (get_compute_type): Use subparts_gt.
12656 (count_type_subparts): Delete.
12657 (expand_vector_operations_1): Use subparts_gt instead of
12658 count_type_subparts.
12660 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12661 Alan Hayward <alan.hayward@arm.com>
12662 David Sherwood <david.sherwood@arm.com>
12664 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12665 (vect_compile_time_alias): ...this new function. Do the calculation
12666 on poly_ints rather than trees.
12667 (vect_prune_runtime_alias_test_list): Update call accordingly.
12669 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12670 Alan Hayward <alan.hayward@arm.com>
12671 David Sherwood <david.sherwood@arm.com>
12673 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12675 (vect_schedule_slp_instance): Likewise.
12677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12678 Alan Hayward <alan.hayward@arm.com>
12679 David Sherwood <david.sherwood@arm.com>
12681 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12682 constant and extern definitions for variable-length vectors.
12683 (vect_get_constant_vectors): Note that the number of units
12684 is known to be constant.
12686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12687 Alan Hayward <alan.hayward@arm.com>
12688 David Sherwood <david.sherwood@arm.com>
12690 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12691 of units as polynomial. Choose between WIDE and NARROW based
12694 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12695 Alan Hayward <alan.hayward@arm.com>
12696 David Sherwood <david.sherwood@arm.com>
12698 * tree-vect-stmts.c (simd_clone_subparts): New function.
12699 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12706 vectors as polynomial. Use build_index_vector for
12707 IFN_GOMP_SIMD_LANE.
12709 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12710 Alan Hayward <alan.hayward@arm.com>
12711 David Sherwood <david.sherwood@arm.com>
12713 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12714 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12715 for variable-length vectors.
12716 (vectorizable_mask_load_store): Treat the number of units as
12717 polynomial, asserting that it is constant if the condition has
12718 already been enforced.
12719 (vectorizable_store, vectorizable_load): Likewise.
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 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12726 of units as polynomial. Punt if we can't tell at compile time
12727 which vector contains the final result.
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-loop.c (vectorizable_induction): Treat the number
12734 of units as polynomial. Punt on SLP inductions. Use an integer
12735 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12736 cast of such a series for variable-length floating-point
12739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12740 Alan Hayward <alan.hayward@arm.com>
12741 David Sherwood <david.sherwood@arm.com>
12743 * tree.h (build_index_vector): Declare.
12744 * tree.c (build_index_vector): New function.
12745 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12746 of units as polynomial, forcibly converting it to a constant if
12747 vectorizable_reduction has already enforced the condition.
12748 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12749 to create a {1,2,3,...} vector.
12750 (vectorizable_reduction): Treat the number of units as polynomial.
12751 Choose vectype_in based on the largest scalar element size rather
12752 than the smallest number of units. Enforce the restrictions
12755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12756 Alan Hayward <alan.hayward@arm.com>
12757 David Sherwood <david.sherwood@arm.com>
12759 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12760 number of units as polynomial.
12762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12763 Alan Hayward <alan.hayward@arm.com>
12764 David Sherwood <david.sherwood@arm.com>
12766 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12767 * target.def (autovectorize_vector_sizes): Return the vector sizes
12768 by pointer, using vector_sizes rather than a bitmask.
12769 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12770 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12771 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12773 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12774 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12775 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12776 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12777 * omp-general.c (omp_max_vf): Likewise.
12778 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12779 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12780 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12781 * tree-vect-slp.c (vect_slp_bb): Likewise.
12782 * doc/tm.texi: Regenerate.
12783 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12785 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12786 the vector size as a poly_uint64 rather than an unsigned int.
12787 (current_vector_size): Change from an unsigned int to a poly_uint64.
12788 (get_vectype_for_scalar_type): Update accordingly.
12789 * tree.h (build_truth_vector_type): Take the size and number of
12790 units as a poly_uint64 rather than an unsigned int.
12791 (build_vector_type): Add a temporary overload that takes
12792 the number of units as a poly_uint64 rather than an unsigned int.
12793 * tree.c (make_vector_type): Likewise.
12794 (build_truth_vector_type): Take the number of units as a poly_uint64
12795 rather than an unsigned int.
12797 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12798 Alan Hayward <alan.hayward@arm.com>
12799 David Sherwood <david.sherwood@arm.com>
12801 * target.def (get_mask_mode): Take the number of units and length
12802 as poly_uint64s rather than unsigned ints.
12803 * targhooks.h (default_get_mask_mode): Update accordingly.
12804 * targhooks.c (default_get_mask_mode): Likewise.
12805 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12806 * doc/tm.texi: Regenerate.
12808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12809 Alan Hayward <alan.hayward@arm.com>
12810 David Sherwood <david.sherwood@arm.com>
12812 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12813 * omp-general.c (omp_max_vf): Likewise.
12814 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12815 (expand_omp_simd): Handle polynomial safelen.
12816 * omp-low.c (omplow_simd_context): Add a default constructor.
12817 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12818 (lower_rec_simd_input_clauses): Update accordingly.
12819 (lower_rec_input_clauses): Likewise.
12821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12822 Alan Hayward <alan.hayward@arm.com>
12823 David Sherwood <david.sherwood@arm.com>
12825 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12826 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12827 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12828 (vect_analyze_slp_cost): Likewise.
12829 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12830 (vect_model_load_cost): Likewise.
12832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12833 Alan Hayward <alan.hayward@arm.com>
12834 David Sherwood <david.sherwood@arm.com>
12836 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12837 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12838 from an unsigned int * to a poly_uint64_pod *.
12839 (calculate_unrolling_factor): New function.
12840 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12843 Alan Hayward <alan.hayward@arm.com>
12844 David Sherwood <david.sherwood@arm.com>
12846 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12847 from an unsigned int to a poly_uint64.
12848 (_loop_vec_info::slp_unrolling_factor): Likewise.
12849 (_loop_vec_info::vectorization_factor): Change from an int
12851 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12852 (vect_get_num_vectors): New function.
12853 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12854 (vect_get_num_copies): Use vect_get_num_vectors.
12855 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12856 to an unsigned int *.
12857 (vect_analyze_data_refs): Change min_vf from an int * to a
12859 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12860 than an unsigned HOST_WIDE_INT.
12861 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12862 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12863 to an unsigned int *.
12864 (vect_analyze_data_ref_dependences): Likewise.
12865 (vect_compute_data_ref_alignment): Handle polynomial vf.
12866 (vect_enhance_data_refs_alignment): Likewise.
12867 (vect_prune_runtime_alias_test_list): Likewise.
12868 (vect_shift_permute_load_chain): Likewise.
12869 (vect_supportable_dr_alignment): Likewise.
12870 (dependence_distance_ge_vf): Take the vectorization factor as a
12871 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12872 (vect_analyze_data_refs): Change min_vf from an int * to a
12874 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12875 vfm1 as a poly_uint64 rather than an int. Make the same change
12876 for the returned bound_scalar.
12877 (vect_gen_vector_loop_niters): Handle polynomial vf.
12878 (vect_do_peeling): Likewise. Update call to
12879 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12880 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12882 * tree-vect-loop.c (vect_determine_vectorization_factor)
12883 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12884 (vect_get_known_peeling_cost): Likewise.
12885 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12886 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12887 (vect_transform_loop): Likewise. Use the lowest possible VF when
12888 updating the upper bounds of the loop.
12889 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12890 rather than an int.
12891 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12892 polynomial unroll factors.
12893 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12894 (vect_make_slp_decision): Likewise.
12895 (vect_supported_load_permutation_p): Likewise, and polynomial
12897 (vect_analyze_slp_cost): Handle polynomial vf.
12898 (vect_slp_analyze_node_operations): Likewise.
12899 (vect_slp_analyze_bb_1): Likewise.
12900 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12901 than an unsigned HOST_WIDE_INT.
12902 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12903 (vectorizable_load): Handle polynomial vf.
12904 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12906 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12909 Alan Hayward <alan.hayward@arm.com>
12910 David Sherwood <david.sherwood@arm.com>
12912 * match.pd: Handle bit operations involving three constants
12913 and try to fold one pair.
12915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12917 * tree-vect-loop-manip.c: Include gimple-fold.h.
12918 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12919 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12920 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12921 Add a path that uses a step of VF instead of 1, but disable it
12923 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12924 and niters_no_overflow parameters. Update calls to
12925 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12926 Create a new SSA name if the latter choses to use a ste other
12927 than zero, and return it via niters_vector_mult_vf_var.
12928 * tree-vect-loop.c (vect_transform_loop): Update calls to
12929 vect_do_peeling, vect_gen_vector_loop_niters and
12930 slpeel_make_loop_iterate_ntimes.
12931 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12932 (vect_gen_vector_loop_niters): Update declarations after above changes.
12934 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12936 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12937 128-bit round to integer instructions.
12938 (ceil<mode>2): Likewise.
12939 (btrunc<mode>2): Likewise.
12940 (round<mode>2): Likewise.
12942 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12944 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12945 unaligned VSX load/store on P8/P9.
12946 (expand_block_clear): Allow the use of unaligned VSX
12947 load/store on P8/P9.
12949 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12951 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12953 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12954 swap associated with both a load and a store.
12956 2018-01-02 Andrew Waterman <andrew@sifive.com>
12958 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12959 * config/riscv/riscv.md (clear_cache): Use it.
12961 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12963 * web.c: Remove out-of-date comment.
12965 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12967 * expr.c (fixup_args_size_notes): Check that any existing
12968 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12969 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12970 (emit_single_push_insn): ...here.
12972 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12974 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12975 (const_vector_encoded_nelts): New function.
12976 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12977 (const_vector_int_elt, const_vector_elt): Declare.
12978 * emit-rtl.c (const_vector_int_elt_1): New function.
12979 (const_vector_elt): Likewise.
12980 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12981 of CONST_VECTOR_ELT.
12983 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12985 * expr.c: Include rtx-vector-builder.h.
12986 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12987 directly on the tree encoding.
12988 (const_vector_from_tree): Likewise.
12989 * optabs.c: Include rtx-vector-builder.h.
12990 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12991 sequence of "u" values.
12992 * vec-perm-indices.c: Include rtx-vector-builder.h.
12993 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12994 directly on the vec_perm_indices encoding.
12996 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12998 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12999 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13000 * rtx-vector-builder.h: New file.
13001 * rtx-vector-builder.c: Likewise.
13002 * rtl.h (rtx_def::u2): Add a const_vector field.
13003 (CONST_VECTOR_NPATTERNS): New macro.
13004 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13005 (CONST_VECTOR_DUPLICATE_P): Likewise.
13006 (CONST_VECTOR_STEPPED_P): Likewise.
13007 (CONST_VECTOR_ENCODED_ELT): Likewise.
13008 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13009 (unwrap_const_vec_duplicate): Likewise.
13010 (const_vec_series_p): Check for a non-duplicated vector encoding.
13011 Say that the function only returns true for integer vectors.
13012 * emit-rtl.c: Include rtx-vector-builder.h.
13013 (gen_const_vec_duplicate_1): Delete.
13014 (gen_const_vector): Call gen_const_vec_duplicate instead of
13015 gen_const_vec_duplicate_1.
13016 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13017 (gen_const_vec_duplicate): Use rtx_vector_builder.
13018 (gen_const_vec_series): Likewise.
13019 (gen_rtx_CONST_VECTOR): Likewise.
13020 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13021 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13022 Build a new vector rather than modifying a CONST_VECTOR in-place.
13023 (handle_special_swappables): Update call accordingly.
13024 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13025 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13026 Build a new vector rather than modifying a CONST_VECTOR in-place.
13027 (handle_special_swappables): Update call accordingly.
13029 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13031 * simplify-rtx.c (simplify_const_binary_operation): Use
13032 CONST_VECTOR_ELT instead of XVECEXP.
13034 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13036 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13037 the selector elements to be different from the data elements
13038 if the selector is a VECTOR_CST.
13039 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13040 ssizetype for the selector.
13042 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13044 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13045 before testing each element individually.
13046 * tree-vect-generic.c (lower_vec_perm): Likewise.
13048 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13050 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13051 * selftest-run-tests.c (selftest::run_tests): Call it.
13052 * vector-builder.h (vector_builder::operator ==): New function.
13053 (vector_builder::operator !=): Likewise.
13054 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13055 (vec_perm_indices::all_from_input_p): New function.
13056 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13057 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13058 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13059 instead of reading the VECTOR_CST directly. Detect whether both
13060 vector inputs are the same before constructing the vec_perm_indices,
13061 and update the number of inputs argument accordingly. Use the
13062 utility functions added above. Only construct sel2 if we need to.
13064 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13066 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13067 the broadcast of the low byte.
13068 (expand_mult_highpart): Use an explicit encoding for the permutes.
13069 * optabs-query.c (can_mult_highpart_p): Likewise.
13070 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13071 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13072 (vectorizable_bswap): Likewise.
13073 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13074 explicit encoding for the power-of-2 permutes.
13075 (vect_permute_store_chain): Likewise.
13076 (vect_grouped_load_supported): Likewise.
13077 (vect_permute_load_chain): Likewise.
13079 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13081 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13082 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13083 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13084 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13085 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13086 (vect_gen_perm_mask_any): Likewise.
13088 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13090 * int-vector-builder.h: New file.
13091 * vec-perm-indices.h: Include int-vector-builder.h.
13092 (vec_perm_indices): Redefine as an int_vector_builder.
13093 (auto_vec_perm_indices): Delete.
13094 (vec_perm_builder): Redefine as a stand-alone class.
13095 (vec_perm_indices::vec_perm_indices): New function.
13096 (vec_perm_indices::clamp): Likewise.
13097 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13098 (vec_perm_indices::new_vector): New function.
13099 (vec_perm_indices::new_expanded_vector): Update for new
13100 vec_perm_indices class.
13101 (vec_perm_indices::rotate_inputs): New function.
13102 (vec_perm_indices::all_in_range_p): Operate directly on the
13103 encoded form, without computing elided elements.
13104 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13105 encoding. Update for new vec_perm_indices class.
13106 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13107 the given vec_perm_builder.
13108 (expand_vec_perm_var): Update vec_perm_builder constructor.
13109 (expand_mult_highpart): Use vec_perm_builder instead of
13110 auto_vec_perm_indices.
13111 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13112 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13113 or double series encoding as appropriate.
13114 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13115 vec_perm_indices instead of auto_vec_perm_indices.
13116 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13117 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13118 (vect_permute_store_chain): Likewise.
13119 (vect_grouped_load_supported): Likewise.
13120 (vect_permute_load_chain): Likewise.
13121 (vect_shift_permute_load_chain): Likewise.
13122 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13123 (vect_transform_slp_perm_load): Likewise.
13124 (vect_schedule_slp_instance): Likewise.
13125 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13126 (vectorizable_mask_load_store): Likewise.
13127 (vectorizable_bswap): Likewise.
13128 (vectorizable_store): Likewise.
13129 (vectorizable_load): Likewise.
13130 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13131 vec_perm_indices instead of auto_vec_perm_indices. Use
13132 tree_to_vec_perm_builder to read the vector from a tree.
13133 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13134 vec_perm_builder instead of a vec_perm_indices.
13135 (have_whole_vector_shift): Use vec_perm_builder and
13136 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13137 truncation to calc_vec_perm_mask_for_shift.
13138 (vect_create_epilog_for_reduction): Likewise.
13139 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13140 from auto_vec_perm_indices to vec_perm_indices.
13141 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13142 instead of changing individual elements.
13143 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13144 the vector in d.perm.
13145 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13146 from auto_vec_perm_indices to vec_perm_indices.
13147 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13148 instead of changing individual elements.
13149 (arm_vectorize_vec_perm_const): Use new_vector to install
13150 the vector in d.perm.
13151 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13152 Update vec_perm_builder constructor.
13153 (rs6000_expand_interleave): Likewise.
13154 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13155 (rs6000_expand_interleave): Likewise.
13157 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13159 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13160 to qimode could truncate the indices.
13161 * optabs.c (expand_vec_perm_var): Likewise.
13163 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13165 * Makefile.in (OBJS): Add vec-perm-indices.o.
13166 * vec-perm-indices.h: New file.
13167 * vec-perm-indices.c: Likewise.
13168 * target.h (vec_perm_indices): Replace with a forward class
13170 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13171 * optabs.h: Include vec-perm-indices.h.
13172 (expand_vec_perm): Delete.
13173 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13174 (expand_vec_perm_const): Declare.
13175 * target.def (vec_perm_const_ok): Replace with...
13176 (vec_perm_const): ...this new hook.
13177 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13178 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13179 * doc/tm.texi: Regenerate.
13180 * optabs.def (vec_perm_const): Delete.
13181 * doc/md.texi (vec_perm_const): Likewise.
13182 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13183 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13184 expand_vec_perm for constant permutation vectors. Assert that
13185 the mode of variable permutation vectors is the integer equivalent
13186 of the mode that is being permuted.
13187 * optabs-query.h (selector_fits_mode_p): Declare.
13188 * optabs-query.c: Include vec-perm-indices.h.
13189 (selector_fits_mode_p): New function.
13190 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13191 is defined, instead of checking whether the vec_perm_const_optab
13192 exists. Use targetm.vectorize.vec_perm_const instead of
13193 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13194 fit in the vector mode before using a variable permute.
13195 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13196 vec_perm_indices instead of an rtx.
13197 (expand_vec_perm): Replace with...
13198 (expand_vec_perm_const): ...this new function. Take the selector
13199 as a vec_perm_indices rather than an rtx. Also take the mode of
13200 the selector. Update call to shift_amt_for_vec_perm_mask.
13201 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13202 Use vec_perm_indices::new_expanded_vector to expand the original
13203 selector into bytes. Check whether the indices fit in the vector
13204 mode before using a variable permute.
13205 (expand_vec_perm_var): Make global.
13206 (expand_mult_highpart): Use expand_vec_perm_const.
13207 * fold-const.c: Includes vec-perm-indices.h.
13208 * tree-ssa-forwprop.c: Likewise.
13209 * tree-vect-data-refs.c: Likewise.
13210 * tree-vect-generic.c: Likewise.
13211 * tree-vect-loop.c: Likewise.
13212 * tree-vect-slp.c: Likewise.
13213 * tree-vect-stmts.c: Likewise.
13214 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13216 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13217 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13218 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13219 (aarch64_vectorize_vec_perm_const): ...this new function.
13220 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13221 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13222 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13223 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13224 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13225 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13226 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13228 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13229 check for NEON modes.
13230 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13231 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13232 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13233 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13235 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13236 the old VEC_PERM_CONST conditions.
13237 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13238 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13239 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13240 (ia64_vectorize_vec_perm_const_ok): Merge into...
13241 (ia64_vectorize_vec_perm_const): ...this new function.
13242 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13243 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13244 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13245 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13246 * config/mips/mips.c (mips_expand_vec_perm_const)
13247 (mips_vectorize_vec_perm_const_ok): Merge into...
13248 (mips_vectorize_vec_perm_const): ...this new function.
13249 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13250 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13251 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13252 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13253 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13254 (rs6000_expand_vec_perm_const): Delete.
13255 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13257 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13258 (altivec_expand_vec_perm_const_le): Take each operand individually.
13259 Operate on constant selectors rather than rtxes.
13260 (altivec_expand_vec_perm_const): Likewise. Update call to
13261 altivec_expand_vec_perm_const_le.
13262 (rs6000_expand_vec_perm_const): Delete.
13263 (rs6000_vectorize_vec_perm_const_ok): Delete.
13264 (rs6000_vectorize_vec_perm_const): New function.
13265 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13266 an element count and rtx array.
13267 (rs6000_expand_extract_even): Update call accordingly.
13268 (rs6000_expand_interleave): Likewise.
13269 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13270 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13271 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13272 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13273 (rs6000_expand_vec_perm_const): Delete.
13274 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13275 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13276 (altivec_expand_vec_perm_const_le): Take each operand individually.
13277 Operate on constant selectors rather than rtxes.
13278 (altivec_expand_vec_perm_const): Likewise. Update call to
13279 altivec_expand_vec_perm_const_le.
13280 (rs6000_expand_vec_perm_const): Delete.
13281 (rs6000_vectorize_vec_perm_const_ok): Delete.
13282 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13283 reference to the SPE evmerge intructions.
13284 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13285 an element count and rtx array.
13286 (rs6000_expand_extract_even): Update call accordingly.
13287 (rs6000_expand_interleave): Likewise.
13288 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13289 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13291 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13293 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13295 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13296 vector mode and that that mode matches the mode of the data
13298 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13299 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13300 directly using expand_vec_perm_1 when forcing selectors into
13302 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13304 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13306 * optabs-query.h (can_vec_perm_p): Delete.
13307 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13308 * optabs-query.c (can_vec_perm_p): Split into...
13309 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13310 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13311 particular selector is valid.
13312 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13313 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13314 (vect_grouped_load_supported): Likewise.
13315 (vect_shift_permute_load_chain): Likewise.
13316 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13317 (vect_transform_slp_perm_load): Likewise.
13318 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13319 (vectorizable_bswap): Likewise.
13320 (vect_gen_perm_mask_checked): Likewise.
13321 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13322 implementations of variable permutation vectors into account
13323 when deciding which selector to use.
13324 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13325 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13326 with a false third argument.
13327 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13328 to test whether the constant selector is valid and can_vec_perm_var_p
13329 to test whether a variable selector is valid.
13331 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13333 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13334 * optabs-query.c (can_vec_perm_p): Likewise.
13335 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13336 instead of vec_perm_indices.
13337 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13338 (vect_gen_perm_mask_checked): Likewise,
13339 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13340 (vect_gen_perm_mask_checked): Likewise,
13342 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13344 * optabs-query.h (qimode_for_vec_perm): Declare.
13345 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13346 (qimode_for_vec_perm): ...this new function.
13347 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13349 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13351 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13352 does not have a conditional at the top.
13354 2018-01-02 Richard Biener <rguenther@suse.de>
13356 * ipa-inline.c (big_speedup_p): Fix expression.
13358 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13361 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13364 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13368 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13369 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13370 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13371 cond_taken_branch_cost 3->4.
13373 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13375 PR tree-optimization/83581
13376 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13377 TODO_cleanup_cfg if any changes have been made.
13379 PR middle-end/83608
13380 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13381 convert_modes if target mode has the right side, but different mode
13384 PR middle-end/83609
13385 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13386 last argument when extracting from CONCAT. If either from_real or
13387 from_imag is NULL, use expansion through memory. If result is not
13388 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13389 the parts directly to inner mode, if even that fails, use expansion
13392 PR middle-end/83623
13393 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13394 check for bswap in mode rather than HImode and use that in expand_unop
13397 Copyright (C) 2018 Free Software Foundation, Inc.
13399 Copying and distribution of this file, with or without modification,
13400 are permitted in any medium without royalty provided the copyright
13401 notice and this notice are preserved.