1 2018-05-18 Martin Liska <mliska@suse.cz>
4 * coverage.c (coverage_init): Write PWD to .gcno file.
5 * doc/gcov.texi: Document how working directory is printed.
6 * gcov-dump.c (dump_gcov_file): Print PWD.
7 * gcov.c (output_intermediate_file): Likewise.
8 (read_graph_file): Read PWD string.
9 (output_lines): Print PWD.
11 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
15 for retval and return false if all args to phi are zero.
17 2018-05-18 Richard Biener <rguenther@suse.de>
19 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
21 (evrp_dom_walker::before_dom_children): Call it.
23 2018-05-18 Richard Biener <rguenther@suse.de>
25 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
26 results when processing array refs with variable index.
28 2018-05-18 Toon Moene <toon@moene.org>
30 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
31 directly after that of -floop-interchange. Indicate that both
32 options are enabled by default when specifying -O3.
34 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
36 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
37 iterator. Delete separate integer-mode vec_set<mode> expander.
38 (aarch64_simd_vec_setv2di): Delete.
39 (vec_setv2di): Delete.
40 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
41 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
42 the "w, r" alternative.
44 2018-05-18 Martin Liska <mliska@suse.cz>
46 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
47 * tree-pass.h (make_pass_lower_switch_O0): New function.
48 * tree-switch-conversion.c (node_has_low_bound): Remove.
49 (node_has_high_bound): Likewise.
50 (node_is_bounded): Likewise.
51 (class pass_lower_switch): Make it a template type and create
53 (pass_lower_switch::execute): Add template argument.
54 (make_pass_lower_switch): New function.
55 (make_pass_lower_switch_O0): New function.
56 (do_jump_if_equal): Remove.
57 (emit_case_nodes): Simplify to just handle all 3 cases and leave
58 all the hard work to tree optimization passes.
60 2018-05-18 Martin Liska <mliska@suse.cz>
62 * dbgcnt.c (limit_low): Renamed from limit.
63 (limit_high): New variable.
64 (dbg_cnt_is_enabled): Check for upper limit.
65 (dbg_cnt): Adjust dumping.
66 (dbg_cnt_set_limit_by_index): Add new argument for high
68 (dbg_cnt_set_limit_by_name): Likewise.
69 (dbg_cnt_process_single_pair): Parse new format.
70 (dbg_cnt_process_opt): Use strtok.
71 (dbg_cnt_list_all_counters): Remove 'value' and add
73 * doc/invoke.texi: Document changes.
75 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
77 * doc/sourcebuild.texi (scalar_all_fma): Document.
78 * tree.def (FMA_EXPR): Delete.
79 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
80 * internal-fn.c (ternary_direct): New macro.
81 (expand_ternary_optab_fn): Likewise.
82 (direct_ternary_optab_supported_p): Likewise.
83 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
84 * builtins.c (fold_builtin_fma): Delete.
85 (fold_builtin_3): Don't call it.
86 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
87 * expr.c (expand_expr_real_2): Likewise.
88 * fold-const.c (operand_equal_p): Likewise.
89 (fold_ternary_loc): Likewise.
90 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
91 * gimple.c (DEFTREECODE): Likewise.
92 * gimplify.c (gimplify_expr): Likewise.
93 * optabs-tree.c (optab_for_tree_code): Likewise.
94 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
95 * tree-eh.c (operation_could_trap_p): Likewise.
96 (stmt_could_throw_1_p): Likewise.
97 * tree-inline.c (estimate_operator_cost): Likewise.
98 * tree-pretty-print.c (dump_generic_node): Likewise.
99 (op_code_prio): Likewise.
100 * tree-ssa-loop-im.c (stmt_cost): Likewise.
101 * tree-ssa-operands.c (get_expr_operands): Likewise.
102 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
103 * fold-const-call.h (fold_fma): Delete.
104 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
105 CFN_FNMA and CFN_FNMS.
107 * genmatch.c (combined_fn): New enum.
108 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
109 (commutative_op): New function.
110 (commutate): Use it. Handle more than 2 operands.
111 (dt_operand::gen_gimple_expr): Use commutative_op.
112 (parser::parse_expr): Allow :c to be used with non-binary
113 operators if the commutative operand is known.
114 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
115 CFN_FMS, CFN_FNMA and CFN_FNMS.
116 (backprop::process_assign_use): Remove FMA_EXPR handling.
117 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
118 (gen_hsa_fma): New function.
119 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
120 IFN_FNMA and IFN_FNMS.
121 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
122 * gimple-fold.h (follow_all_ssa_edges): Declare.
123 * gimple-fold.c (follow_all_ssa_edges): New function.
124 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
125 gimple_build interface and use follow_all_ssa_edges to fold the result.
126 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
127 instead of checking for optabs directly.
128 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
129 rather than FMA_EXPRs.
130 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
131 call to IFN_FMA instead of an FMA_EXPR.
133 2018-05-17 Jim Wilson <jimw@sifive.com>
135 * expr.c (do_tablejump): When converting index to Pmode, if we have a
136 sign extended promoted subreg, and the range does not have the sign bit
137 set, then do a sign extend.
139 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
140 test, check for sign extended subreg and/or constant operands, and
141 do a sign extend in that case.
143 2018-05-17 Steve Ellcey <sellcey@cavium.com>
145 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
146 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
148 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
149 Change logics_shift_reg to logics_shift_imm.
150 (thunderx2t99_fp_loadpair_basic): Delete.
151 (thunderx2t99_fp_storepair_basic): Delete.
152 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
153 (thunderx2t99_asimd_polynomial): Delete.
154 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
155 and neon_fp_mul_d_scalar_q.
156 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
157 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
158 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
159 (thunderx2t99_asimd_lut): Add missing tbl types.
160 (thunderx2t99_asimd_ext): Delete.
161 (thunderx2t99_asimd_load1_1_mult): Delete.
162 (thunderx2t99_asimd_load1_2_mult): Delete.
163 (thunderx2t99_asimd_load1_ldp): New.
164 (thunderx2t99_asimd_load1): New.
165 (thunderx2t99_asimd_load2): Add missing *load2* types.
166 (thunderx2t99_asimd_load3): New.
167 (thunderx2t99_asimd_load4): New.
168 (thunderx2t99_asimd_store1_1_mult): Delete.
169 (thunderx2t99_asimd_store1_2_mult): Delete.
170 (thunderx2t99_asimd_store2_mult): Delete.
171 (thunderx2t99_asimd_store2_onelane): Delete.
172 (thunderx2t99_asimd_store_stp): New.
173 (thunderx2t99_asimd_store1): New.
174 (thunderx2t99_asimd_store2): New.
175 (thunderx2t99_asimd_store3): New.
176 (thunderx2t99_asimd_store4): New.
178 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
180 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
181 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
183 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
184 Segher Boessenkool <segher@kernel.crashing.org>
187 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
190 2018-05-17 Richard Biener <rguenther@suse.de>
192 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
193 for pruning loop and prune defs feeding only already visited PHIs.
195 2018-05-17 Richard Biener <rguenther@suse.de>
197 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
199 2018-05-17 Bin Cheng <bin.cheng@arm.com>
200 Richard Biener <rguenther@suse.de>
202 PR tree-optimization/85793
203 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
204 for VMAT_ELEMENTWISE.
206 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
208 * internal-fn.h (lookup_internal_fn): Declare
209 * internal-fn.c (lookup_internal_fn): New function.
210 * gimple.c (gimple_build_call_from_tree): Handle calls to
212 * gimple-pretty-print.c (dump_gimple_call): Print "." before
213 internal function names.
214 * tree-pretty-print.c (dump_generic_node): Likewise.
215 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
217 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
219 * gimple-fold.h (gimple_build): Make the function forms take
220 combined_fn rather than built_in_function.
221 (gimple_simplify): Likewise.
222 * gimple-match-head.c (gimple_simplify): Likewise.
223 * gimple-fold.c (gimple_build): Likewise.
224 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
225 rather than gimple_build_call_internal.
226 (get_initial_defs_for_reduction): Likewise.
227 (vect_create_epilog_for_reduction): Likewise.
228 (vectorizable_live_operation): Likewise.
230 2018-05-17 Martin Liska <mliska@suse.cz>
232 * gimple-ssa-sprintf.c (format_directive): Do not use
233 space in between 'G_' and '('.
235 2018-05-17 Jakub Jelinek <jakub@redhat.com>
238 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
239 even if the mask is not all ones.
242 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
244 (ix86_gimple_fold_builtin): Likewise.
247 * config/i386/i386.c: Include tree-vector-builder.h.
248 (ix86_vector_shift_count): New function.
249 (ix86_fold_builtin): Fold shift builtins by scalar count.
250 (ix86_gimple_fold_builtin): Likewise.
252 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
253 _mm512_setzero): New intrinsics.
255 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
256 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
258 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
259 code generation for cases where splatting a value is not useful.
260 * simplify-rtx.c (simplify_ternary_operation): Simplify
261 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
262 mode to a vec_concat.
264 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
266 * config.gcc: Support "goldmont-plus".
267 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
269 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
270 PROCESSOR_GOLDMONT_PLUS.
271 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
272 (processor_target_table): Add "goldmont-plus".
273 (PTA_GOLDMONT_PLUS): Define.
274 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
275 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
276 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
277 (fold_builtin_cpu): Add "goldmont-plus".
278 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
279 (ix86_option_override_internal): Add "goldmont-plus".
280 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
281 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
282 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
283 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
285 2018-05-17 Richard Biener <rguenther@suse.de>
287 PR tree-optimization/85757
288 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
289 remove defs that only feed that PHI from further processing.
291 2018-05-16 Jim Wilson <jimw@sifive.com>
293 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
295 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
297 2018-05-16 Mark Wielaard <mark@klomp.org>
299 * dwarf2out.c (count_index_strings): New function.
300 (output_indirect_strings): Call count_index_strings and generate
301 header for dwarf_version >= 5.
303 2018-05-16 Mark Wielaard <mark@klomp.org>
305 * dwarf2out.c (dwarf_FORM): New function.
306 (set_indirect_string): Use dwarf_FORM.
307 (reset_indirect_string): Likewise.
308 (size_of_die): Likewise.
309 (value_format): Likewise.
310 (output_die): Likewise.
311 (add_skeleton_AT_string): Likewise.
312 (output_macinfo_op): Likewise.
313 (index_string): Likewise.
314 (output_index_string_offset): Likewise.
315 (output_index_string): Likewise.
316 (count_index_strings): Likewise.
318 2018-05-16 Carl Love <cel@us.ibm.com>
320 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
321 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
323 2018-05-16 Martin Jambor <mjambor@suse.cz>
325 * ipa-prop.c (ipa_free_all_edge_args): Remove.
326 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
328 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
330 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
331 (fnma<mode>4): Likewise.
332 (fms<mode>4): Likewise.
333 (fnms<mode>4): Likewise.
334 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
335 (aarch64_fnma<mode>4): Likewise.
336 (aarch64_fms<mode>4): Likewise.
337 (aarch64_fnms<mode>4): Likewise.
338 (aarch64_fnmadd<mode>4): Likewise.
340 2018-05-16 Jason Merrill <jason@redhat.com>
342 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
344 2018-05-16 Richard Biener <rguenther@suse.de>
346 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
347 (dump_stmt_cost): Declare.
348 (add_stmt_cost): Dump cost we add.
349 (add_stmt_costs): New function.
350 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
352 (vect_analyze_stmt): Adjust prototype.
353 (vectorizable_condition): Likewise.
354 (vectorizable_live_operation): Likewise.
355 (vectorizable_reduction): Likewise.
356 (vectorizable_induction): Likewise.
357 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
358 cost vector to pass to vectorizable_ and record afterwards.
359 (vect_model_reduction_cost): Take cost vector argument and adjust.
360 (vect_model_induction_cost): Likewise.
361 (vectorizable_reduction): Likewise.
362 (vectorizable_induction): Likewise.
363 (vectorizable_live_operation): Likewise.
364 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
365 SLP_TREE_NUMBER_OF_VEC_STMTS.
366 (vect_analyze_slp_cost_1): Remove.
367 (vect_analyze_slp_cost): Likewise.
368 (vect_slp_analyze_node_operations): Take visited args and
369 a target cost vector. Avoid processing already visited stmt sets.
370 (vect_slp_analyze_operations): Use a local cost vector to gather
371 costs and register those of non-discarded instances.
372 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
373 (vect_schedule_slp_instance): Remove copying of
374 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
376 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
377 adding cost. Record cost entry location.
378 (vect_prologue_cost_for_slp_op): Function to compute cost of
379 a constant or invariant generated for SLP vect in the prologue,
380 split out from vect_analyze_slp_cost_1.
381 (vect_model_simple_cost): Make static. Adjust for SLP costing.
382 (vect_model_promotion_demotion_cost): Likewise.
383 (vect_model_store_cost): Likewise, make static.
384 (vect_model_load_cost): Likewise.
385 (vectorizable_bswap): Add cost vector arg and adjust.
386 (vectorizable_call): Likewise.
387 (vectorizable_simd_clone_call): Likewise.
388 (vectorizable_conversion): Likewise.
389 (vectorizable_assignment): Likewise.
390 (vectorizable_shift): Likewise.
391 (vectorizable_operation): Likewise.
392 (vectorizable_store): Likewise.
393 (vectorizable_load): Likewise.
394 (vectorizable_condition): Likewise.
395 (vectorizable_comparison): Likewise.
396 (can_vectorize_live_stmts): Likewise.
397 (vect_analyze_stmt): Likewise.
398 (vect_transform_stmt): Adjust calls to vectorizable_*.
399 * tree-vectorizer.c: Include gimple-pretty-print.h.
400 (dump_stmt_cost): New function.
402 2018-05-16 Richard Biener <rguenther@suse.de>
404 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
405 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
406 * tree-ssa-dse.c: Include tree-ssa-loop.h.
407 (check_name): New callback.
408 (dse_classify_store): Track cycles via a visited bitmap of PHI
409 defs and simplify handling of in-loop and across loop dead stores
410 and properly fail for loop-variant refs. Handle byte-tracking with
411 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
414 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
416 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
417 (vect_get_mask_type_for_stmt): Likewise.
418 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
420 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
421 to determine the statement's vector type and the vector type that
422 should be used for calculating nunits. Deal with cases in which
423 the type has to be deferred.
424 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
425 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
426 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
427 (vect_determine_vf_for_stmt): New functions, split out from...
428 (vect_determine_vectorization_factor): ...here.
429 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
430 (vect_get_mask_type_for_stmt): New functions, split out from
431 vect_determine_vectorization_factor.
433 2018-05-16 Richard Biener <rguenther@suse.de>
435 * tree-cfg.c (verify_gimple_assign_ternary): Properly
436 verify the [VEC_]COND_EXPR embedded comparison.
438 2018-05-15 Martin Sebor <msebor@redhat.com>
440 PR tree-optimization/85753
441 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
442 RECORD_TYPE in addition to ARRAY_TYPE.
444 2018-05-15 Martin Sebor <msebor@redhat.com>
447 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
449 2018-05-15 Richard Biener <rguenther@suse.de>
451 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
452 add by_clobber_p one. Change algorithm to collect all defs
453 representing uses we need to walk and try reducing them to
454 a single one before failing.
455 (dse_dom_walker::dse_optimize_stmt): Adjust.
457 2018-05-13 Mark Wielaard <mark@klomp.org>
459 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
460 (size_of_loc_descr): Likewise.
461 (output_loc_operands): Likewise.
462 (output_loc_operands_raw): Likewise.
463 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
464 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
465 (hash_loc_operands): Likewise.
466 (compare_loc_operands): Likewise.
468 2018-05-14 Mark Wielaard <mark@klomp.org>
470 * dwarf2out.c (count_index_addrs): New function.
471 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
473 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
475 PR tree-optimization/83648
476 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
477 return value as malloc candidate.
479 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
482 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
483 as true in call to suggest_attribute.
485 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
487 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
490 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
492 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
494 (aarch64_crypto_aesd_fused): Likewise.
496 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
498 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
499 (movsi_aarch64): Likewise.
500 (load_pairsi): Likewise.
501 (load_pairdi): Likewise.
502 (store_pairsi): Likewise.
503 (store_pairdi): Likewise.
504 (load_pairsf): Likewise.
505 (load_pairdf): Likewise.
506 (store_pairsf): Likewise.
507 (store_pairdf): Likewise.
508 (zero_extend): Likewise.
509 (trunc): Swap alternatives.
510 (fcvt_target): Add '?' to prefer w over r.
512 2018-05-14 Jakub Jelinek <jakub@redhat.com>
515 * config/i386/i386.md: Disallow non-commutative arithmetics in
516 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
517 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
518 in the peephole2 before it.
520 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
522 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
523 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
524 (ix86_handle_option): Handle -mcldemote.
525 * config.gcc: New header.
526 * config/i386/cldemoteintrin.h: New file.
527 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
528 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
530 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
531 OPTION_MASK_ISA_CLDEMOTE.
532 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
533 (ix86_valid_target_attribute_inner_p): Ditto.
534 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
535 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
536 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
537 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
538 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
540 * config/i386/i386.opt: Add -mcldemote.
541 * config/i386/x86intrin.h: New header.
542 * doc/invoke.texi: Add -mcldemote.
544 2018-05-14 Richard Biener <rguenther@suse.de>
546 * doc/match-and-simplify.texi: Adjust :s documentation.
548 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
550 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
551 intended memcpy size.
552 (REORDER_45): Likewise.
554 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
557 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
558 * vec.c (qsort_chk): Use gcc_qsort.
559 * Makefile.in (OBJS-libcommon): Add sort.o.
560 (build/sort.o): New target. Use it...
561 (BUILD_RTL): ... here, and...
562 (build/gencfn-macros): ... here, and...
563 (build/genmatch): ... here.
565 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
566 Chung-Ju Wu <jasonwucj@gmail.com>
568 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
569 * config/nds32/nds32-graywolf.md: New file.
570 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
571 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
573 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
574 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
575 * config/nds32/nds32.md (pipeline_model): Add graywolf.
576 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
577 * config/nds32/pipelines.md: Include n15 settings.
579 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
580 Chung-Ju Wu <jasonwucj@gmail.com>
582 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
583 * config/nds32/nds32-n13.md: New file.
584 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
585 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
587 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
588 * config/nds32/nds32.md (pipeline_model): Add n13.
589 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
590 * config/nds32/pipelines.md: Include n13 settings.
592 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
593 Chung-Ju Wu <jasonwucj@gmail.com>
595 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
596 * config/nds32/nds32-n10.md: New file.
597 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
598 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
600 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
601 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
602 * config/nds32/nds32.md (pipeline_model): Add n10.
603 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
604 * config/nds32/pipelines.md: Include n10 settings.
606 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
607 Kito Cheng <kito.cheng@gmail.com>
608 Chung-Ju Wu <jasonwucj@gmail.com>
610 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
611 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
612 Add enum values for DSP extension instructions.
613 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
615 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
616 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
618 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
619 * config/nds32/nds32-dspext.md: New file for DSP implementation.
620 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
621 * config/nds32/nds32-intrinsic.md: Likewise.
622 * config/nds32/nds32_intrinsic.h: Likewise.
623 * config/nds32/nds32-md-auxiliary.c: Likewise.
624 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
625 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
626 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
627 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
628 * config/nds32/nds32-protos.h: New declarations for DSP extension.
629 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
630 TYPE_DMAC in switch statement.
631 * config/nds32/nds32.c: New checking and implementation for DSP
632 extension instructions.
633 * config/nds32/nds32.h: Likewise.
634 * config/nds32/nds32.md: Likewise.
635 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
636 * config/nds32/predicates.md: Implement new predicates for DSP
639 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
641 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
642 Reformat alternatives and attributes so it is easier to identify
643 which constraints/attributes go with which instruction.
644 (mov<mode>_hardfloat32, FMOVE64): Likewise.
645 (mov<mode>_softfloat32, FMOVE64): Likewise.
646 (mov<mode>_hardfloat64, FMOVE64): Likewise.
647 (mov<mode>_softfloat64, FMOVE64): Likewise.
649 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
651 * doc/extend.texi (PowerPC Built-in Functions): Rename this
653 (Basic PowerPC Built-in Functions): The new name of the
654 subsection previously known as "PowerPC Built-in Functions".
655 (Basic PowerPC Built-in Functions Available on all Configurations):
657 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
658 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
659 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
660 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
662 2018-05-11 Martin Jambor <mjambor@suse.cz>
665 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
668 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
671 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
673 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
675 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
676 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
677 (ix86_handle_option): Handle -mwaitpkg.
678 * config.gcc: New header.
679 * config/i386/cpuid.h (bit_WAITPKG): New bit.
680 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
681 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
683 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
684 OPTION_MASK_ISA_WAITPKG.
685 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
686 (ix86_option_override_internal): Add PTA_WAITPKG.
687 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
688 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
690 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
691 __builtin_ia32_umwait and __builtin_ia32_tpause.
692 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
693 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
694 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
695 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
696 UNSPECV_TPAUSE): New.
697 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
698 * config/i386/i386.opt: Add -mwaitpkg.
699 * config/i386/waitpkgintrin.h: New file.
700 * config/i386/x86intrin.h: New header.
701 * doc/invoke.texi: Add -mwaitpkg.
703 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
706 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
708 (cortex-m0): Use armv6s-m isa.
709 (cortex-m0plus): Likewise.
710 (cortex-m1): Likewise.
711 (cortex-m0.small-multiply): Likewise.
712 (cortex-m0plus.small-multiply): Likewise.
713 (cortex-m1.small-multiply): Likewise.
715 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
716 Jakub Jelinek <jakub@redhat.com>
718 PR tree-optimization/85692
719 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
720 source permute as well.
722 2018-05-11 Martin Liska <mliska@suse.cz>
725 * doc/extend.texi: Document LLVM style format for no_sanitize
728 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
730 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
731 mode_supports_vsx_dform_quad to mode_supports_dq_form.
732 (mode_supports_vsx_dform_quad): Likewise.
733 (mode_supports_vmx_dform): Move these functions to be next to the
734 other mode_supports functions.
735 (mode_supports_dq_form): Likewise.
736 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
737 mode_supports_dq_form.
738 (reg_offset_addressing_ok_p): Likewise.
739 (offsettable_ok_by_alignment): Likewise.
740 (rs6000_legitimate_offset_address_p): Likewise.
741 (legitimate_lo_sum_address_p): Likewise.
742 (rs6000_legitimize_address): Likewise.
743 (rs6000_legitimize_reload_address): Likewise.
744 (rs6000_secondary_reload_inner): Likewise.
745 (rs6000_preferred_reload_class): Likewise.
746 (rs6000_output_move_128bit): Likewise.
748 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
750 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
751 Generate SImode target register for null target.
752 <case IX86_BUILTIN_XGETBV>: Ditto.
753 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
754 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
756 2018-05-10 Carl Love <cel@us.ibm.com>
758 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
759 dcbtt and dcbtstt if operands[2] is 0.
761 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
764 * config/i386/sse.md (usadv64qi): New expander.
766 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
768 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
769 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
770 -maltivec=be support.
771 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
772 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
773 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
774 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
775 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
776 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
777 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
778 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
779 altivec_vsumsws): Adjust.
780 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
781 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
782 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
784 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
785 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
786 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
787 (altivec_lve<VI_char>x): Delete expand.
788 (*altivec_lve<VI_char>x_internal): Rename to...
789 (altivec_lve<VI_char>x): ... this.
790 (altivec_lvxl_<mode>): Delete expand.
791 (*altivec_lvxl_<mode>_internal): Rename to ...
792 (altivec_lvxl_<mode>): ... this.
793 (altivec_stvxl_<mode>): Delete expand.
794 (*altivec_stvxl_<mode>_internal): Rename to ...
795 (altivec_stvxl_<mode>): ... this.
796 (altivec_stve<VI_char>x): Delete expand.
797 (*altivec_stve<VI_char>x_internal): Rename to ...
798 (altivec_stve<VI_char>x): ... this.
799 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
800 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
801 reduc_plus_scal_<mode>): Adjust.
802 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
804 (rs6000_cpu_cpp_builtins): Adjust.
805 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
806 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
807 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
808 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
809 -maltivec=be support.
810 (rs6000_split_vec_extract_var): Adjust.
811 (rs6000_split_v4si_init): Adjust.
812 (swap_selector_for_mode): Delete.
813 (altivec_expand_lvx_be, altivec_expand_stvx_be,
814 altivec_expand_stvex_be): Delete.
815 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
816 -maltivec=be support.
817 (rs6000_gimple_fold_builtin): Ditto.
818 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
820 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
821 (TARGET_DIRECT_MOVE_64BIT): Adjust.
822 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
823 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
824 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
825 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
826 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
827 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
828 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
829 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
830 anonymous split): Adjust.
831 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
832 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
834 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
836 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
837 when --with-gxx-include-dir is also specified.
838 * configure: Regenerate.
840 2018-05-09 Jim Wilson <jimw@sifive.com>
843 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
844 * config/riscv/t-withmultilib: New.
845 * config/riscv/withmultilib.h: New.
846 * doc/install.texi: Document RISC-V --with-multilib-list support.
848 2018-05-09 Richard Biener <rguenther@suse.de>
850 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
852 (vect_bb_vectorization_profitable_p): Adjust. Compute
853 actual scalar cost using the cost vector and the add_stmt_cost
856 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
858 PR rtl-optimization/85645
859 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
860 in the REG_CFA_REGISTER note for LR, don't leave it empty.
862 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
864 PR rtl-optimization/85645
865 * shrink-wrap.c (spread_components): Return a boolean saying if
866 anything was changed.
867 (try_shrink_wrapping_separate): Iterate spread_components until
868 nothing changes anymore.
870 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
872 PR rtl-optimization/85645
873 * regrename.c (build_def_use): Also kill the chains that include the
874 destination of a REG_CFA_REGISTER note.
876 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
878 PR rtl-optimization/85645
879 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
880 insn that has a REG_CFA_REGISTER note.
882 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
884 * cfgexpand.c (expand_clobber): New function.
885 (expand_gimple_stmt_1): Use it.
886 * tree-vect-stmts.c (vect_clobber_variable): New function,
888 (vectorizable_simd_clone_call): ...here.
889 (vectorizable_store): Emit a clobber either side of an
890 IFN_STORE_LANES sequence.
891 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
893 2018-05-09 Tom de Vries <tom@codesourcery.com>
896 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
897 (define_insn "trap_if_false"): Add exit after trap.
899 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
901 PR rtl-optimization/85638
902 * bb-reorder.c: Include common/common-target.h.
903 (create_forwarder_block): New function extracted from...
904 (fix_up_crossing_landing_pad): ...here. Rename into...
905 (dw2_fix_up_crossing_landing_pad): ...this.
906 (sjlj_fix_up_crossing_landing_pad): New function.
907 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
908 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
909 from both partitions and exit the loop after one iteration.
911 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
914 * doc/extend.texi (PowerPC Built-in Functions): Rename this
916 (Basic PowerPC Built-in Functions): The new name of the
917 subsection previously known as "PowerPC Built-in Functions".
918 (Basic PowerPC Built-in Functions Available on all Configurations):
920 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
922 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
924 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
926 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
929 2018-05-08 Jim Wilson <jimw@sifive.com>
931 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
932 (LD_EMUL_SUFFIX): New.
935 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
937 * doc/extend.texi (PowerPC Built-in Functions): Rename this
939 (Basic PowerPC Built-in Functions): The new name of the
940 subsection previously known as "PowerPC Built-in Functions".
941 (Basic PowerPC Built-in Functions Available on all Configurations):
943 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
945 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
947 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
949 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
952 2018-05-08 Jakub Jelinek <jakub@redhat.com>
955 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
956 after cmpelim optimization.
958 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
960 * config.gcc: Support "goldmont".
961 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
962 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
964 * config/i386/i386.c (m_GOLDMONT): Define.
965 (processor_target_table): Add "goldmont".
966 (PTA_GOLDMONT): Define.
967 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
968 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
969 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
970 (fold_builtin_cpu): Add "goldmont".
971 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
972 (ix86_option_override_internal): Add "goldmont".
973 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
974 (processor_type): Add PROCESSOR_GOLDMONT.
975 * config/i386/i386.md: Add CPU "glm".
976 * config/i386/glm.md: New file.
977 * config/i386/x86-tune.def: Add m_GOLDMONT.
978 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
980 2018-05-08 Jakub Jelinek <jakub@redhat.com>
983 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
985 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
986 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
987 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
990 * config/i386/i386.c (ix86_fold_builtin): Handle
991 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
994 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
995 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
997 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1000 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1001 (check_arch): Likewise.
1002 (check_fpu): Return the result rather than printing it.
1003 (end arch): Fix operator precedence.
1004 (end cpu): Likewise.
1005 (END): Print the result from check_fpu.
1007 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1008 Alan Hayward <alan.hayward@arm.com>
1009 David Sherwood <david.sherwood@arm.com>
1011 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1012 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1013 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1014 (*fcmuo<mode>_and): New patterns.
1016 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1018 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1019 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1020 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1021 (cmp_op, sve_imm_con): New code attributes.
1022 (SVE_COND_INT_CMP, imm_con): Delete.
1023 (cmp_op): Remove above unspecs from int attribute.
1024 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1026 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1027 comparison-specific unspecs.
1028 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1029 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1030 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1031 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1032 (*vec_fcm<cmp_op><mode>): Rename to...
1033 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1034 (*vec_fcmuo<mode>): Rename to...
1035 (*fcmuo<mode>): ...this and adjust likewise.
1036 (*pred_fcm<cmp_op><mode>): New pattern.
1037 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1038 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1040 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1042 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1043 (aarch64_emit_sve_predicated_cond): New function.
1044 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1045 (aarch64_emit_unspec_cond_or): Replace with...
1046 (aarch64_emit_sve_or_conds): ...this new function. Use
1047 aarch64_emit_sve_ptrue_op for the individual comparisons and
1048 aarch64_emit_binop to OR them together.
1049 (aarch64_emit_inverted_unspec_cond): Replace with...
1050 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1051 aarch64_emit_sve_ptrue_op for the comparison and
1052 aarch64_emit_unop to invert the result.
1053 (aarch64_expand_sve_vec_cmp_float): Update after the above
1054 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1056 2018-05-07 Nathan Sidwell <nathan@acm.org>
1058 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1059 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1060 (Backwards Compatibility): Likewise.
1062 2018-05-07 Luis Machado <luis.machado@linaro.org>
1066 2018-05-07 Luis Machado <luis.machado@linaro.org>
1068 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1069 <prefetch_dynamic_strides>: New const bool field.
1070 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1071 prefetch_dynamic_strides.
1072 (exynosm1_prefetch_tune): Likewise.
1073 (thunderxt88_prefetch_tune): Likewise.
1074 (thunderx_prefetch_tune): Likewise.
1075 (thunderx2t99_prefetch_tune): Likewise.
1076 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1077 (aarch64_override_options_internal): Update to set
1078 PARAM_PREFETCH_DYNAMIC_STRIDES.
1079 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1080 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1081 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1082 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1083 prefetch-dynamic-strides setting.
1085 2018-05-07 Luis Machado <luis.machado@linaro.org>
1087 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1088 <minimum_stride>: New const int field.
1089 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1090 minimum_stride field.
1091 (exynosm1_prefetch_tune): Likewise.
1092 (thunderxt88_prefetch_tune): Likewise.
1093 (thunderx_prefetch_tune): Likewise.
1094 (thunderx2t99_prefetch_tune): Likewise.
1095 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1096 (aarch64_override_options_internal): Update to set
1097 PARAM_PREFETCH_MINIMUM_STRIDE.
1098 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1099 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1100 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1101 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1102 stride is constant and is below the minimum stride threshold.
1104 2018-05-07 Luis Machado <luis.machado@linaro.org>
1106 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1109 2018-05-07 Luis Machado <luis.machado@linaro.org>
1111 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1112 <prefetch_dynamic_strides>: New const bool field.
1113 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1114 prefetch_dynamic_strides.
1115 (exynosm1_prefetch_tune): Likewise.
1116 (thunderxt88_prefetch_tune): Likewise.
1117 (thunderx_prefetch_tune): Likewise.
1118 (thunderx2t99_prefetch_tune): Likewise.
1119 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1120 (aarch64_override_options_internal): Update to set
1121 PARAM_PREFETCH_DYNAMIC_STRIDES.
1122 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1123 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1124 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1125 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1126 prefetch-dynamic-strides setting.
1128 2018-05-07 Luis Machado <luis.machado@linaro.org>
1130 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1131 <minimum_stride>: New const int field.
1132 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1133 minimum_stride field.
1134 (exynosm1_prefetch_tune): Likewise.
1135 (thunderxt88_prefetch_tune): Likewise.
1136 (thunderx_prefetch_tune): Likewise.
1137 (thunderx2t99_prefetch_tune): Likewise.
1138 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1139 (aarch64_override_options_internal): Update to set
1140 PARAM_PREFETCH_MINIMUM_STRIDE.
1141 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1142 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1143 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1144 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1145 stride is constant and is below the minimum stride threshold.
1147 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1150 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1151 the type is addressable. Don't force op into register if it has
1154 2018-05-05 Roland McGrath <mcgrathr@google.com>
1157 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1158 any section for which we don't know a specific type it should have,
1159 regardless of name. Previously this was done only for the exact
1160 names ".init_array", ".fini_array", and ".preinit_array".
1161 (default_elf_asm_named_section): Add comment about
1162 relationship with default_section_type_flags and SECTION_NOTYPE.
1163 (get_section): Don't consider it a type conflict if one side has
1164 SECTION_NOTYPE and the other doesn't, as long as neither has the
1165 SECTION_BSS et al used in the default_section_type_flags logic.
1167 2018-05-05 Tom de Vries <tom@codesourcery.com>
1170 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1171 (workaround_barsyncs): New function.
1172 (nvptx_reorg): Use workaround_barsyncs.
1173 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1174 (define_expand "nvptx_membar_cta"): New define_expand.
1175 (define_insn "*nvptx_membar_cta"): New insn.
1177 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1179 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1180 To improve optimization opportunities.
1181 * builtin-types.def: The new needed builtin types for the above.
1183 2018-05-04 Richard Biener <rguenther@suse.de>
1185 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1186 * gimple-ssa-store-merging.c
1187 (imm_store_chain_info::output_merged_store): Remove redundant create,
1188 release split_store vector contents on failure.
1189 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1190 scalar stmt vector on cache hit.
1192 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1194 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1196 * config.gcc (powerpc-xilinx-eabi*): Remove.
1197 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1199 (fusion_addis_mem_combo_load): Ditto.
1200 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1202 (rs6000_cpu_cpp_builtins): Ditto.
1203 * config/rs6000/rs6000-linux.c
1204 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1205 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1206 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1208 (rs6000_setup_reg_addr_masks): Ditto.
1209 (rs6000_init_hard_regno_mode_ok): Ditto.
1210 (rs6000_option_override_internal): Ditto.
1211 (legitimate_lo_sum_address_p): Ditto.
1212 (rs6000_legitimize_address): Ditto.
1213 (rs6000_legitimize_reload_address): Ditto.
1214 (rs6000_legitimate_address_p): Ditto.
1215 (abi_v4_pass_in_fpr): Ditto.
1216 (setup_incoming_varargs): Ditto.
1217 (rs6000_gimplify_va_arg): Ditto.
1218 (rs6000_split_multireg_move): Ditto.
1219 (rs6000_savres_strategy): Ditto.
1220 (rs6000_emit_prologue_components): Ditto.
1221 (rs6000_emit_epilogue_components): Ditto.
1222 (rs6000_emit_prologue): Ditto.
1223 (rs6000_emit_epilogue): Ditto.
1224 (rs6000_elf_file_end): Ditto.
1225 (rs6000_function_value): Ditto.
1226 (rs6000_libcall_value): Ditto.
1227 * config/rs6000/rs6000.h: Ditto.
1228 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1229 (TARGET_MINMAX): ... this. New.
1230 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1231 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1232 (*movsi_internal1_single): Delete.
1233 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1234 mfpu=, mxilinx-fpu): Delete.
1235 * config/rs6000/singlefp.h: Delete.
1236 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1237 * config/rs6000/t-rs6000: Ditto.
1238 * config/rs6000/t-xilinx: Delete.
1239 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1240 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1241 (VStype_simple): Delete.
1242 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1243 * config/rs6000/xfpu.h: Delete.
1244 * config/rs6000/xfpu.md: Delete.
1245 * config/rs6000/xilinx.h: Delete.
1246 * config/rs6000/xilinx.opt: Delete.
1247 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1248 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1250 2018-05-04 Tom de Vries <tom@codesourcery.com>
1253 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1256 2018-05-04 Richard Biener <rguenther@suse.de>
1259 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1260 (expand_complex_div_wide): Likewise.
1261 (expand_complex_operations_1): Likewise.
1262 (expand_complex_libcall): Preserve EH info of the original stmt.
1263 (tree_lower_complex): Handle removed blocks.
1264 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1265 on complex multiplication and division libcall builtins.
1267 2018-05-04 Richard Biener <rguenther@suse.de>
1270 * fold-const.c (negate_expr_p): Restrict negation of operand
1271 zero of a division to when we know that can happen without
1273 (fold_negate_expr_1): Likewise.
1275 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1278 * real.h (real_nextafter): Declare.
1279 * real.c (real_nextafter): New function.
1280 * fold-const-call.c (fold_const_nextafter): New function.
1281 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1282 CASE_CFN_NEXTTOWARD.
1283 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1284 even when arg1_mode is different from arg0_mode.
1286 2018-05-03 Nathan Sidwell <nathan@acm.org>
1288 * doc/extend.texi (Deprecated Features): Remove
1290 (Backwards Compatibility): Likewise.
1291 * doc/invoke.texi (C++ Language Options): Likewise.
1292 (C++ Dialect Options): Likewise.
1294 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1297 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1298 _mm512_mask_mullox_epi64): New intrinsics.
1300 2018-05-03 Tom de Vries <tom@codesourcery.com>
1303 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1304 dump files): Add offload-tree.
1306 2018-05-03 Richard Biener <rguenther@suse.de>
1308 PR tree-optimization/85615
1309 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1310 to loops not nested in BBs loop father to avoid creating multi-entry
1313 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1315 PR tree-optimization/70291
1316 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1317 arguments. Change return type to tree. Emit libcall as a new
1318 statement rather than replacing existing one when inplace_p is true.
1319 (expand_complex_multiplication_components): New function.
1320 (expand_complex_multiplication): Expand floating-point complex
1321 multiplication using the above.
1322 (expand_complex_division): Rename inner_type parameter to type.
1323 Update expand_complex_libcall call-site.
1324 (expand_complex_operations_1): Update expand_complex_multiplication
1325 and expand_complex_division call-sites.
1327 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1330 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1331 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1332 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1333 the highest significant bit of the shift count mask is clear. In
1334 check whether and[sq]i3 is needed verify that all significant bits
1335 of the shift count other than the highest are set.
1337 2018-05-02 Tom de Vries <tom@codesourcery.com>
1340 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1341 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1342 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1343 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1344 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1345 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1346 __builtin_goacc_parlevel_size.
1348 2018-05-02 Richard Biener <rguenther@suse.de>
1350 PR tree-optimization/85597
1351 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1352 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1355 2018-05-02 Tom de Vries <tom@codesourcery.com>
1358 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1359 dump files): Add ltrans-tree.
1361 2018-05-02 Tom de Vries <tom@codesourcery.com>
1364 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1365 dump files): Add wpa-ipa.
1367 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1369 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1370 powerpc*-*-linux*paired* target.
1371 * config/rs6000/750cl.h: Delete.
1372 * config/rs6000/paired.h: Delete.
1373 * config/rs6000/paired.md: Delete.
1374 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1376 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1377 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1378 comment. Remove paired float support.
1379 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1380 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1382 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1383 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1385 * config/rs6000/rs6000.c: Remove paired float support.
1386 (paired_expand_vector_init, paired_expand_vector_move,
1387 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1388 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1389 paired_expand_builtin, paired_expand_predicate_builtin,
1390 paired_init_builtins): Delete.
1391 * config/rs6000/rs6000.h: Remove paired float support.
1392 * config/rs6000/rs6000.md: Remove paired float support.
1393 (move_from_CR_ov_bit): Delete.
1394 * config/rs6000/rs6000.opt (mpaired): Delete.
1395 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1396 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1398 2018-05-02 Richard Biener <rguenther@suse.de>
1401 * gimplify.c (gimplify_save_expr): When in SSA form allow
1402 SAVE_EXPRs to compute to SSA vars.
1404 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1407 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1408 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1409 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1410 clobber operands[2], instead use a new pseudo. Formatting fixes.
1412 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1414 PR tree-optimization/85586
1415 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1416 exit early for statements in the same group if the accesses are
1419 2018-05-02 Tom de Vries <tom@codesourcery.com>
1422 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1425 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1427 PR tree-optimization/85143
1428 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1430 2018-05-01 Tom de Vries <tom@codesourcery.com>
1433 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1434 not found" error message.
1436 2018-05-01 Tom de Vries <tom@codesourcery.com>
1439 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1440 * vec.c (test_ordered_remove_if): New function.
1441 (vec_c_tests): Call test_ordered_remove_if.
1442 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1443 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1444 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1445 VEC_ORDERED_REMOVE_IF.
1447 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1449 PR tree-optimization/82665
1450 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1451 pointer subtraction where arguments come from a memchr call.
1453 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1455 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1456 --push-state --as-needed and --pop-state instead of --as-needed and
1457 --no-as-needed if ld supports it.
1458 * configure: Regenerated.
1461 * doc/install.texi2html: Replace _002d with - and _002a with * in
1462 generated html files using sed.
1464 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1467 * gcc-rich-location.c (blank_line_before_p): New function.
1468 (use_new_line): New function.
1469 (gcc_rich_location::add_fixit_insert_formatted): New function.
1470 * gcc-rich-location.h
1471 (gcc_rich_location::add_fixit_insert_formatted): New function.
1473 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1475 * selftest.c (assert_streq): Rename "expected" and "actual" to
1476 "val1" and "val2". Extend NULL-handling to cover both inputs
1477 symmetrically, while still requiring both to be non-NULL for a pass.
1478 * selftest.h (assert_streq): Rename "expected" and "actual" to
1480 (ASSERT_EQ): Likewise.
1481 (ASSERT_EQ_AT): Likewise.
1482 (ASSERT_KNOWN_EQ): Likewise.
1483 (ASSERT_KNOWN_EQ_AT): Likewise.
1484 (ASSERT_NE): Likewise.
1485 (ASSERT_MAYBE_NE): Likewise.
1486 (ASSERT_MAYBE_NE_AT): Likewise.
1487 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1488 the assertion to pass.
1489 (ASSERT_STREQ_AT): Likewise.
1491 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1493 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1494 interaction with -pie.
1496 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1498 * selftest.h: Fix alphabetization of per-source-file selftest
1501 2018-04-30 Jason Merrill <jason@redhat.com>
1503 PR c++/61982 - dead stores to destroyed objects.
1504 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1507 2018-04-30 Jason Merrill <jason@redhat.com>
1509 * tree.c (build_clobber): New.
1510 * tree.h: Declare it.
1511 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1513 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1515 * diagnostic-show-locus.c (layout::layout): Update for
1516 location_get_source_line returning a char_span.
1517 (struct char_span): Move to input.h.
1518 (struct correction): Update for fields in char_span becoming
1520 (struct source_line): Update for location_get_source_line
1521 returning a char_span.
1522 (layout::print_line): Likewise.
1523 * edit-context.c (edited_file::print_content): Likewise.
1524 (edited_file::print_diff_hunk): Likewise.
1525 (edited_file::print_run_of_changed_lines): Likewise.
1526 (edited_file::get_num_lines): Likewise.
1527 (edited_line::edited_line): Likewise.
1528 * final.c (asm_show_source): Likewise.
1529 * input.c (location_get_source_line): Convert return type
1530 from const char * to char_span, losing the final "line_len"
1532 (dump_location_info): Update for the above.
1533 (get_substring_ranges_for_loc): Likewise. Use a char_span
1534 when handling the literal within the line.
1535 (test_reading_source_line): Update for location_get_source_line
1536 returning a char_span.
1537 * input.h (class char_span): Move here from
1538 diagnostic-show-locus.c, converting from a struct to a class.
1539 Make data members private.
1540 (char_span::operator bool): New.
1541 (char_span::length): New.
1542 (char_span::get_buffer): New.
1543 (char_span::operator[]): New.
1544 (char_span::subspan): Make const.
1545 (char_span::xstrdup): New.
1546 (location_get_source_line): Convert return type from const char *
1547 to char_span, losing the final "line_size" param.
1549 2018-04-30 Jan Hubicka <jh@suse.cz>
1551 * lto-wrapper.c (ltrans_priorities): New static var.
1552 (cmp_priority): New.
1553 (run_gcc): Read priorities and if doing parallel build order
1554 the Makefile by them.
1556 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1558 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1560 2018-04-30 Richard Biener <rguenther@suse.de>
1562 * tree-cfg.c (verify_address): Remove base argument, add
1563 flag whether to check TREE_ADDRESSABLE and do that.
1564 (verify_expr): Remove.
1565 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1566 (verify_gimple_assign_single): Likewise.
1567 (verify_gimple_switch): Likewise.
1568 (verify_expr_location_1): Dereference tp once. Add (disabled)
1569 piece from verify_expr.
1570 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1572 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1574 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1576 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1578 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1579 (small_data_pattern): Likewise.
1580 (arc_rewrite_small_data): Likewise.
1581 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1582 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1583 (get_symbol_alignment): New function.
1584 (legitimate_small_data_address_p): Likewise.
1585 (legitimate_scaled_address): Update, call
1586 legitimate_small_data_address_p.
1587 (output_sdata): New static variable.
1588 (arc_print_operand): Update how we handle small data operands.
1589 (arc_print_operand_address): Likewise.
1590 (arc_legitimate_address_p): Update, use
1591 legitimate_small_data_address_p.
1592 (arc_rewrite_small_data_p): Remove.
1593 (arc_rewrite_small_data_1): Likewise.
1594 (arc_rewrite_small_data): Likewise.
1595 (small_data_pattern): Likewise.
1596 (compact_sda_memory_operand): Update to use
1597 legitimate_small_data_address_p and get_symbol_alignment.
1598 (prepare_move_operands): Don't rewite sdata pattern.
1599 (prepare_extend_operands): Remove.
1600 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1602 (zero_extendqisi2): Likewise.
1603 (zero_extendhisi2): Likewise.
1604 (extendqihi2): Likewise.
1605 (extendqisi2): Likewise.
1606 (extendhisi2): Likewise.
1610 * config/arc/constraints.md (Usd): Change it to memory constraint.
1612 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1614 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1615 as source of std instructions.
1616 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1617 allow 6-bit constants as source for store instructions.
1618 (movdi_insn): Update instruction pattern to allow 6-bit constants
1619 as source for store instructions.
1621 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1623 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1625 2018-04-30 Nathan Sidwell <nathan@acm.org>
1626 Sandra Loosemore <sandra@codesourcery.com>
1628 * dumpfile.c (dump_open): Allow '-' for stdout.
1629 * doc/invoke.texi (Developer Options): Document dump filename
1630 determination early. Document stdin/stdout selection.
1632 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1634 Microblaze Target: PIC data text relative
1636 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1637 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1639 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1640 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1641 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1642 New addressing mode for data-text relative position indepenedent code.
1643 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1644 'ADDRESS_SYMBOLIC_TXT_REL'.
1645 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1646 (microblaze_legitimate_pic_operand): Exclude function calls from
1647 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1648 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1650 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1651 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1652 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1653 for 'address + offset'.
1654 (microblaze_expand_prologue): Add new function prologue call for
1656 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1657 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1658 table in case of TARGET_PIC_DATA_TEXT_REL.
1659 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1660 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1661 Add new macros 'UNSPEC_TEXT',
1662 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1663 + exclude function calls from 'UNSPEC_PLT' in case of data text
1665 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1666 new target hook for generating address diff vector tables in case of
1668 * doc/tm.texi : Regenerate.
1669 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1670 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1671 of addr diff vector generation.
1672 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1673 target hook definition.
1674 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1675 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1676 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1677 option for data text relative.
1679 2018-04-30 Richard Biener <rguenther@suse.de>
1681 * tree-chrec.h (evolution_function_is_constant_p): Remove
1683 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1685 2018-04-30 Richard Biener <rguenther@suse.de>
1688 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1690 2018-04-30 Richard Biener <rguenther@suse.de>
1692 PR tree-optimization/28364
1693 PR tree-optimization/85275
1694 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1695 copying first exit test.
1697 2018-04-28 Mark Wielaard <mark@klomp.org>
1699 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1701 (dwarf_AT): Handle DW_AT_addr_base.
1702 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1704 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1707 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1708 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1709 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1710 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1712 2018-04-28 Richard Biener <rguenther@suse.de>
1714 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1715 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1716 to reflect use. Only add interesting stmts.
1718 2018-04-27 Martin Jambor <mjambor@suse.cz>
1721 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1722 the jump function allows for passing through aggregate values.
1724 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1726 * input.h (in_system_header_at): Convert from macro to inline
1728 (from_macro_expansion_at): Likewise.
1729 (from_macro_definition_at): Likewise.
1731 2018-04-27 Jeff Law <law@redhat.com>
1733 * config.gcc: Mark tile* targets as deprecated/obsolete.
1735 2018-04-27 Richard Biener <rguenther@suse.de>
1737 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1740 2018-04-27 Richard Biener <rguenther@suse.de>
1742 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1744 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1746 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1747 with Yd constraint. Set "preferred_for_speed" attribute from
1748 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1750 (*movdi_internal): Ditto.
1751 (movti_interunit splitters): Remove
1752 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1753 (movdi_interunit splitters): Ditto.
1754 * config/i386/constraints.md (Ye): Remove.
1755 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1757 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1760 * config/aarch64/constraints.md (Usg): Limit to 31.
1763 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1765 PR tree-optimization/85529
1766 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1767 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1768 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1769 zero extension or masking of the MSB bit.
1770 (optimize_range_tests): Add FIRST_BB argument, pass it through
1771 to optimize_range_tests_var_bound.
1772 (maybe_optimize_range_tests, reassociate_bb): Adjust
1773 optimize_range_tests callers.
1775 2018-04-26 Richard Biener <rguenther@suse.de>
1776 Jakub Jelinek <jakub@redhat.com>
1778 * cgraph.h (symbol_table): Just declare debug method here.
1779 * symtab.c (symbol_table::debug): Define.
1781 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1783 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1785 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1787 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1788 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1789 (*movdi_internal): Substitute Yi and Yj constraint with x
1790 and Ym and Yn constraint with y constraint. Update "isa"
1791 attribute and set "preferred_for_speed" attribute from
1792 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1793 (*movsi_internal): Ditto.
1794 (*movdf_internal): Ditto.
1795 (*movsf_internal): Ditto.
1796 (*zero_extendsidi2): Ditto.
1797 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1798 (sse2_loadld): Ditto.
1799 (*vec_extract<ssevecmodelower>_0): Ditto.
1800 (*vec_extractv4si_0_zext_sse4): Ditto.
1801 (vec_concatv2di): Ditto.
1802 (*vec_dup<mode>): Ditto.
1803 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1804 * config/i386/constraints.md (Yi): Remove.
1809 2018-04-26 Nathan Sidwell <nathan@acm.org>
1811 * dumpfile.c (dump_open): New.
1812 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1813 (dump_finish): Detect stdio/stderr by value not name.
1815 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1817 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1819 2018-04-26 Tom de Vries <tom@codesourcery.com>
1822 * config/nvptx/nvptx.c (verify_neutering_jumps)
1823 (verify_neutering_labels): New function
1824 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1826 2018-04-26 Tom de Vries <tom@codesourcery.com>
1829 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1830 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1833 2018-04-26 Richard Biener <rguenther@suse.de>
1834 Tom de Vries <tom@codesourcery.com>
1837 * lto-streamer-out.c (output_function): Fixup loops if required to match
1838 discovery done in the reader.
1840 2018-04-26 Richard Biener <rguenther@suse.de>
1842 PR tree-optimization/85116
1843 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1844 have a loop exit from the single latch predecessor. Remove
1845 case of header with just condition.
1846 (ch_base::copy_headers): Exclude infinite loops from any
1848 (pass_ch::execute): Record exits.
1850 2018-04-26 Richard Biener <rguenther@suse.de>
1852 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1853 prologue cost vector and pass it to vect_get_load_cost.
1854 (vect_get_peeling_costs_all_drs): Likewise.
1855 (vect_peeling_hash_get_lowest_cost): Likewise.
1856 (vect_enhance_data_refs_alignment): Likewise.
1858 2018-04-26 Richard Biener <rguenther@suse.de>
1861 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1862 checking of integer<->pointer conversions.
1863 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1864 sign-/zero-extending pointer types.
1865 (expand_omp_for_static_chunk): Likewise.
1867 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1868 Jean Lee <xiaoyur347@gmail.com>
1870 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1871 (TARGET_ASAN_SHADOW_OFFSET): Define.
1872 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1873 true for -fsanitize=address.
1875 2018-04-25 Mark Wielaard <mark@klomp.org>
1877 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1880 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1882 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1883 than "alu", remove explicit "memory" and "imm_disp" attributes.
1884 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1887 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1888 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1889 gen_lowpart_no_emit.
1891 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1894 * config/i386/i386.c (ix86_expand_builtin): Change memory
1895 operand to XI, extend p0 to Pmode.
1896 * config/i386/i386.md: Change unspec volatile and operand
1897 1 mode to XI, change operand 0 mode to P.
1899 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1901 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1902 GET_MODE_MASK before any checking.
1903 (nds32_can_use_bset_p): Likewise.
1904 (nds32_can_use_btgl_p): Likewise.
1906 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1908 * config/nds32/nds32-doubleword.md: New define_split pattern for
1909 illegal register number.
1911 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1913 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1915 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1917 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1919 2018-04-25 Richard Biener <rguenther@suse.de>
1921 * lto-streamer.h (LTO_major_version): Bump to 8.
1923 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1925 * BASE-VER: Set to 9.0.0.
1927 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1929 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1930 in __abskf2 and __powikf2.
1932 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1935 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1936 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1937 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1938 Use the above on operand 2. Reindent.
1939 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1941 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1944 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1946 * config/i386/i386.opt (mcet): Removed.
1947 * doc/install.texi: Remove -mcet documentation.
1948 * doc/invoke.texi: Likewise.
1950 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1953 * doc/install.texi: Remove -mcet from bootstrap-cet.
1955 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1958 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1959 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1963 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1964 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1965 containing a CONST_VECTOR.
1967 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1969 * doc/install.texi: Update newlib dependency for nvptx.
1971 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1974 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1975 instead of INTVAL when shifting x left.
1977 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1979 PR tree-optimization/85478
1980 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1981 vect_grouped_store_supported for single element vectors.
1983 2018-04-24 Richard Biener <rguenther@suse.de>
1986 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1987 load cost increase to the case of non-constant step.
1989 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1992 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1993 destination if any_malformed_asm.
1995 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1998 * expr.c (store_field): In the bitfield case, if the value comes from
1999 a function call and is returned in registers by means of a PARALLEL,
2000 do not change the mode of the temporary unless BLKmode and VOIDmode.
2002 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2004 PR rtl-optimization/85423
2005 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2006 dependencies to debug insns when the previous insn is non-debug.
2008 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2010 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2011 enums into a single definition.
2012 (fls): Fix predicates and printing.
2015 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2017 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2018 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2019 and short u6 immediate.
2020 (check_if_valid_sleep_operand): Remove.
2021 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2023 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2025 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2026 flag_always_save_lp condition.
2027 * config/nds32/nds32.opt (malways-save-lp): New option.
2029 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2031 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2032 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2033 * config/nds32/nds32.h
2034 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2035 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2037 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2039 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2040 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2042 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2043 Chung-Ju Wu <jasonwucj@gmail.com>
2045 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2047 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2048 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2050 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2052 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2054 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2056 * config/nds32/nds32-protos.h (nds32_data_alignment,
2057 nds32_local_alignment): Declare.
2058 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2059 nds32_local_alignment): New functions.
2060 (TARGET_CONSTANT_ALIGNMENT): Define.
2061 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2063 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2065 * config/nds32/nds32.c
2066 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2067 (TARGET_MODES_TIEABLE_P): Likewise.
2069 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2071 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2074 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2075 Chung-Ju Wu <jasonwucj@gmail.com>
2077 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2078 for unaligned access.
2079 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2081 * config/nds32/nds32-intrinsic.md: Likewise.
2082 * config/nds32/nds32_intrinsic.h: Likewise.
2083 * config/nds32/nds32.h (nds32_builtins): Likewise.
2084 * config/nds32/nds32.opt (munaligned-access): New option.
2085 * config/nds32/nds32.c (nds32_asm_file_start): Display
2086 flag_unaligned_access status.
2088 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2090 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2091 -mno-relax is present.
2092 * config/riscv/linux.h (LINK_SPEC): Ditto.
2094 2018-04-20 Martin Sebor <msebor@redhat.com>
2097 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2099 (gimple_fold_builtin_stxcpy_chk): Same.
2100 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2102 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2105 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2106 __powikf2 when long double is IEEE 128-bit.
2108 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2110 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2111 step to make sure stack always aligned.
2113 2018-04-20 Carl Love <cel@us.ibm.com>
2116 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2117 size check for arg0.
2119 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2120 Tom de Vries <tom@codesourcery.com>
2123 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2124 Emit insns for calls too.
2125 (nvptx_find_par): Always look for worker-level predecessor insn.
2126 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2128 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2129 (nvptx_process_pars): Propagate frames for calls.
2131 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2134 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2136 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2137 (ix86_handle_option): Don't handle OPT_mibt.
2138 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2140 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2142 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2143 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2144 (ix86_target_macros): Define __CET__ with flag_cf_protection
2145 for -fcf-protection.
2146 * config/i386/i386.c (isa2_opts): Remove -mibt.
2147 * config/i386/i386.h (TARGET_IBT): Removed.
2148 (TARGET_IBT_P): Likewise.
2149 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2150 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2151 * config/i386/i386.opt (mcet): Update help message.
2154 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2155 -mcet as an alias for -mshstk.
2157 2018-04-20 Richard Biener <rguenther@suse.de>
2160 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2161 complexity by forcing a single use of the multiply operand.
2163 2018-04-20 Martin Jambor <mjambor@suse.cz>
2166 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2167 recursion dependency to only apply to non-clones.
2169 2018-04-20 Martin Jambor <mjambor@suse.cz>
2172 * ipa-cp.c (create_specialized_node): Check that clones of
2173 self-recursive edges exist during IPA-CP.
2175 2018-04-19 Toon Moene <toon@moene.org>
2177 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2180 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2182 PR tree-optimization/85467
2183 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2184 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2185 VECTOR_CST element to type.
2187 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2190 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2191 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2192 (builtin_longjmp): Likewise.
2193 (save_stack_nonlocal): New pattern.
2194 (restore_stack_nonlocal): Likewise.
2196 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2199 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2200 Replace ASM_OUTPUT_LABEL with fprintf.
2202 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2205 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2206 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2207 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2208 define __IBT__ and __SHSTK__ for -fcf-protection.
2209 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2211 (ix86_trampoline_init): Likewise.
2212 (x86_output_mi_thunk): Likewise.
2213 (ix86_notrack_prefixed_insn_p): Likewise.
2214 (ix86_option_override_internal): Don't disallow -fcf-protection.
2215 * config/i386/i386.md (rdssp<mode>): Also enable for
2217 (incssp<mode>): Likewise.
2218 (nop_endbr): Likewise.
2219 * config/i386/i386.opt (mcet): Change help message to built-in
2223 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2224 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2225 enable CET built-in functions.
2227 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2229 * common/config/i386/i386-common.c
2230 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2231 OPTION_MASK_ISA_MOVDIRI_UNSET,
2232 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2233 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2234 * config.gcc (movdirintrin.h): New header.
2235 * config/i386/cpuid.h (bit_MOVDIRI,
2236 bit_MOVDIR64B): New bits.
2237 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2239 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2240 (VOID, PVOID, PCVOID)): New function types.
2241 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2242 __builtin_ia32_directstoreu_u64,
2243 __builtin_ia32_movdir64b): New builtins.
2244 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2245 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2247 (ix86_valid_target_attribute_inner_p): Ditto.
2248 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2249 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2250 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2251 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2252 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2253 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2254 (movdiri<mode>, movdir64b_<mode>): New.
2255 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2256 * config/i386/immintrin.h: Include movdirintrin.h.
2257 * config/i386/movdirintrin.h: New file.
2258 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2260 2018-04-19 Richard Biener <rguenther@suse.de>
2263 * cfg.c (clear_bb_flags): When loop state says we have
2264 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2266 2018-04-19 Richard Biener <rguenther@suse.de>
2268 PR tree-optimization/84737
2269 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2270 copying restrict info.
2271 (vect_setup_realignment): Use it.
2272 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2273 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2274 the first DR to all generated stores.
2275 (vectorizable_load): Likewise for loads.
2277 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2279 PR tree-optimization/85446
2280 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2281 the integral and pointer types to have the same precision.
2283 * doc/install.texi: Document --disable-cet being the default and
2286 2018-04-18 Martin Liska <mliska@suse.cz>
2288 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2291 2018-04-18 Martin Liska <mliska@suse.cz>
2294 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2297 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2298 arguments if they are comparable.
2300 2018-04-18 Martin Liska <mliska@suse.cz>
2303 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2306 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2309 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2312 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2313 ENDBR after calling __morestack.
2315 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2318 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2319 by using gcc_base_ver to generate a gcc_driver_version, and use
2320 it when generating GCC_DRIVER_NAME.
2321 * configure: Regenerate.
2323 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2326 * config.gcc: Obsolete powerpc*-*-*spe*.
2328 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2331 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2332 (stabstr_D): Change type of unum from unsigned int to
2333 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2336 2018-04-17 Jim Wilson <jimw@sifive.com>
2339 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2340 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2341 Set arg_pointer_offset after using pretend_args_size.
2343 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2345 PR rtl-optimization/85431
2346 * dse.c (record_store): Ignore zero width stores.
2349 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2350 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2351 __builtin_stack_restore rather than after it.
2352 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2353 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2354 argument instead of virtual_dynamic_stack_rtx.
2356 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2358 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2360 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2361 Add note to error message to explain internal mapping of overloaded
2362 built-in function name to non-overloaded built-in function name.
2363 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2366 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2369 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2370 where the inputs overlap with the output.
2372 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2374 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2375 (=v, v) alternative and explicit "memory" attribute.
2376 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2377 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2379 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2380 "sselog1" type instead of "sselog".
2381 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2382 "sselog". Remove explicit "memory" attribute.
2383 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2384 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2386 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2387 "sselog1" type instead of "sselog". Remove explicit "memory"
2389 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2390 use "sselog1" type instead of "sselog". Remove explicit "memory"
2392 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2393 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2395 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2396 "sselog1" type instead of "sselog". Remove explicit "memory"
2398 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2399 use "sselog1" type instead of "sselog". Remove explicit "memory"
2403 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2406 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2409 2018-04-17 Martin Jambor <mjambor@suse.cz>
2412 * ipa-cp.c (create_specialized_node): Call
2413 expand_all_artificial_thunks if necessary.
2415 2018-04-17 Martin Liska <mliska@suse.cz>
2418 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2419 in message, remote space in between '_G' and '('.
2421 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2424 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2425 avx512f_vmcmp<mode>3<round_saeonly_name>,
2426 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2427 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2428 avx512f_rndscale<mode><round_saeonly_name>,
2429 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2430 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2431 Use %<iptr>2 instead of %2 for -masm=intel.
2432 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2433 avx512f_vcvttss2usi<round_saeonly_name>,
2434 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2436 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2437 avx512f_vcvttsd2usi<round_saeonly_name>,
2438 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2439 Use %q1 instead of %1 for -masm=intel.
2440 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2441 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2442 of %3 for -masm=intel.
2443 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2445 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2447 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2449 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2450 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2452 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2454 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2455 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2456 %g1 and one with %0 and %1.
2457 (avx512er_vmrcp28<mode><round_saeonly_name>,
2458 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2460 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2461 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2462 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2463 of %0 and %{%4%} for -masm=intel.
2464 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2465 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2466 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2467 order of %0 and %{%5%}%{z%} for -masm=intel.
2469 2018-04-17 Jan Hubicka <jh@suse.cz>
2472 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2474 2018-04-17 Martin Liska <mliska@suse.cz>
2477 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2478 for target_clone error message. Make default implementation
2479 clone to be a local declaration.
2480 (separate_attrs): Add new argument and check for an empty
2482 (expand_target_clones): Handle it.
2483 (ipa_target_clone): Make redirection just for target_clones
2486 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2487 Tom de Vries <tom@codesourcery.com>
2490 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2491 tiled basic blocks without omp continue statements.
2493 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2496 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2497 vec_extract expression as having side effects to make sure it gets
2500 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2503 * config/i386/i386.c (get_builtin_code_for_version): Check
2506 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2509 * gcc/config.gcc: Support "skylake".
2510 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2512 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2513 (processor_target_table): Add "skylake".
2514 (ix86_option_override_internal): Add "skylake".
2515 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2516 PROCESSOR_CANNONLAKE.
2517 (get_builtin_code_for_version): Fix priority for
2518 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2519 PROCESSOR_SKYLAKE-AVX512.
2520 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2521 (processor_type): Add PROCESSOR_SKYLAKE.
2523 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2524 Jason Merrill <jason@redhat.com>
2527 * convert.c (convert_to_integer_1): Use direct recursion for
2528 enumeral types and types with a precision less than the number
2529 of bits in their mode.
2531 2018-04-16 Julia Koval <julia.koval@intel.com>
2534 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2535 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2537 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2540 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2541 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2542 and -mno-direct-move.
2544 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2547 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2548 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2549 Ensure negative shifts result in {0}.
2551 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2553 PR rtl-optimization/79916
2554 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2555 regs (if any) to define how to gnerate SD moves when LRA is in
2558 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2560 PR rtl-optimization/85393
2561 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2562 * except.c (expand_dw2_landing_pad_for_region): Make static.
2563 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2564 a label and unconditional jump to old_bb, rather than
2565 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2568 PR rtl-optimization/85376
2569 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2570 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2571 instead of a specific value.
2573 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2574 Bin Cheng <bin.cheng@arm.com>
2576 PR tree-optimization/82965
2577 PR tree-optimization/83991
2578 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2579 by_profile_only parameter.
2580 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2581 information if the loop was predicted to iterate too many times.
2582 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2584 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2587 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2590 2018-04-13 Martin Liska <mliska@suse.cz>
2591 Jakub Jelinek <jakub@redhat.com>
2594 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2595 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2596 * builtins.c (expand_builtin_memory_copy_args): Use
2597 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2598 handle dest_addr == pc_rtx.
2600 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2603 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2604 asked to not generate direct moves.
2605 (fix_trunc<mode>si2_stfiwx): Similar.
2606 (fix_trunc<mode>si2_internal): Similar.
2608 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2611 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2612 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2613 lookup if dest in some wider mode is known to be const0_rtx and
2614 if so, record permanent equivalence for it to be ZERO_EXTEND of
2615 the narrower mode destination.
2617 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2619 * lto-streamer-out.c (output_function): Revert 259346.
2620 * omp-expand.c (expand_oacc_for): Likewise.
2622 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2624 PR rtl-optimization/85354
2625 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2626 * sel-sched.c (sel_global_init): ... here.
2628 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2631 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2632 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2633 mode for PE-COFF targets.
2634 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2635 (i386_pe_asm_lto_end): Likewise.
2636 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2637 (TARGET_ASM_LTO_END): Likewise.
2638 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2639 (i386_pe_asm_lto_start): New function.
2640 (i386_pe_asm_lto_end): Likewise.
2642 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2643 Richard Biener <rguenther@suse.de>
2646 * lto-streamer-out.c (output_function): Fix CFG loop state before
2648 * omp-expand.c (expand_oacc_for): Handle calls to internal
2649 functions like regular functions.
2651 2018-04-12 Richard Biener <rguenther@suse.de>
2654 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2655 for the early LTO debug to properly generate references to it
2656 during DIE emission. Do not re-use that for the skeleton for
2658 (dwarf2out_early_finish): Likewise.
2660 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2663 * config/i386/sse.md
2664 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2665 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2666 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2667 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2668 and output is a reg, avoid creating invalid lowpart subreg, but
2669 instead split into a 512-bit move. Don't split if not AVX512VL,
2670 input is xmm16+ reg and output is a mem.
2671 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2672 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2673 xmm16+ reg and output is a mem.
2675 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2677 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2678 also for flag_dwarf2_cfi_asm.
2680 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2682 PR rtl-optimization/85342
2683 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2684 a bool scalar var inside of the loop instead. Don't try to update
2685 recog_data.operand after failed apply_change_group.
2687 2018-04-12 Tom de Vries <tom@codesourcery.com>
2690 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2691 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2692 array with flexible array member as array without given dimension.
2693 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2694 argument for undefined param to true.
2696 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2699 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2700 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2701 from PowerPC section.
2702 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2703 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2704 help text that is too long.
2705 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2706 help text that is too long.
2707 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2708 help text that is too long.
2710 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2712 * config/alpha/alpha.md (stack_probe_internal): Rename
2713 from "probe_stack". Update all callers.
2715 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2717 PR rtl-optimization/84566
2718 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2719 sched_macro_fuse_insns.
2721 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2724 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2725 (compute_block_dependences): ... from here.
2727 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2729 PR tree-optimization/85331
2730 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2731 from int to HOST_WIDE_INT.
2733 2018-04-11 Martin Jambor <mjambor@suse.cz>
2736 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2737 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2738 not the same as the source val.
2739 (cgraph_edge_brings_value_p): New parameter.
2740 (gather_edges_for_value): Pass destination value to
2741 cgraph_edge_brings_value_p.
2742 (perhaps_add_new_callers): Likewise.
2743 (get_info_about_necessary_edges): Likewise and exclude values brought
2744 only by self-recursive edges.
2745 (create_specialized_node): Redirect only clones of self-calling edges.
2746 (+self_recursive_pass_through_p): New function.
2747 (find_more_scalar_values_for_callers_subset): Use it.
2748 (find_aggregate_values_for_callers_subset): Likewise.
2749 (known_aggs_to_agg_replacement_list): Removed.
2750 (decide_whether_version_node): Re-calculate known constants for all
2751 remaining context clones.
2753 2018-04-11 Richard Biener <rguenther@suse.de>
2756 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2757 from early DWARF output.
2758 (dwarf2out_early_finish): Output line info unconditionally into
2759 early DWARF and add reference to it.
2761 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2764 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2765 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2766 other than V2DFmode using iptr mode attribute.
2767 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2769 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2771 PR rtl-optimization/84659
2772 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2774 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2777 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2779 (output_loc_list): Pass address of a dummy size variable even in the
2780 locview handling loop.
2781 (index_location_lists): Add comment on why skip_loc_list_entry can't
2784 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2787 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2790 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2793 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2794 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2795 and -mstring-compare-inline-limit.
2797 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2800 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2801 for stack clash protection in a register whenever we need it to be in
2804 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2806 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2807 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2809 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2812 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2814 (mlong-double-): Ditto.
2815 * config/rs6000/sysv4.opt (msdata=): Ditto.
2816 (mtls-size=): Ditto.
2818 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2820 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2821 erroneous entries for
2822 "vector int vec_ldl (int, long int *)", and
2823 "vector unsigned int vec_ldl (int, unsigned long int *)".
2824 Add comments and entries for
2825 "vector bool char vec_ldl (int, bool char *)",
2826 "vector bool short vec_ldl (int, bool short *)",
2827 "vector bool int vec_ldl (int, bool int *)",
2828 "vector bool long long vec_ldl (int, bool long long *)",
2829 "vector pixel vec_ldl (int, pixel *)",
2830 "vector long long vec_ldl (int, long long *)",
2831 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2832 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2833 type tree bool_long_long_type_node and correct definition of
2834 bool_V2DI_type_node to make reference to this new type tree.
2835 (rs6000_mangle_type): Replace erroneous reference to
2836 bool_long_type_node with bool_long_long_type_node.
2837 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2838 comments to emphasize sign distinctions for char and int types and
2839 replace RS6000_BTI_bool_long constant with
2840 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2841 use of RS6000_BTI_pixel.
2842 (bool_long_type_node): Remove this macro definition.
2843 (bool_long_long_type_node): New macro definition
2845 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2847 PR rtl-optimization/85300
2848 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2849 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2850 simplify_unary_operation fails.
2852 2018-04-10 Martin Liska <mliska@suse.cz>
2854 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2855 cgraph_edge and ipa_ref.
2857 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2861 * config/i386/sse.md
2862 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2863 computation of the VEC_MERGE selector from mask.
2864 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2865 Fix decoding of the VEC_MERGE selector into mask.
2867 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2869 PR tree-optimization/85286
2870 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2872 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2874 * final.c (final_1): Set insn_last_address as well as
2875 insn_current_address.
2877 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2880 * explow.c (emit_stack_probe): Call validize_mem on memory location
2881 before passing it to gen_probe_stack. Create address operand and
2882 legitimize it for the probe_stack_address case.
2884 2018-04-09 Jan Hubicka <jh@suse.cz>
2887 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2888 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2889 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2891 (free_lang_data): Rebuild type inheritance graph.
2893 2018-04-09 Martin Sebor <msebor@redhat.com>
2895 * invoke.texi (-finline-small-functions): Mention other optimization
2897 (-findirect-inlining, -fpartial-inlining): Same.
2898 (-finline-functions-called-once): Same.
2899 (-freorder-blocks-and-partition): Same.
2901 2018-04-09 Jan Hubicka <jh@suse.cz>
2904 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2905 jumps; choose last target that matches the criteria (i.e.
2906 no partition changes for non-crossing jumps).
2907 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2908 support for redirecting crossing jumps to non-crossing.
2910 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2912 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2913 also for naked functions.
2915 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2917 * config/arc/arc.md (add_shift): New pattern.
2918 (add_shift2): Likewise.
2919 (sub_shift): Likewise.
2920 (sub_shift_cmp0_noout): Likewise.
2921 (compare_si_ashiftsi): Likewise.
2922 (xbfu_cmp0_noout): New combine pattern.
2923 (xbfu_cmp0"): Likewise.
2924 (movsi_set_cc_insn): Place the predicable variant first.
2925 (commutative_binary_cmp0_noout): Remove clobber.
2926 (commutative_binary_cmp0): New pattern.
2927 (noncommutative_binary_cmp0): Likewise.
2928 (noncommutative_binary_cmp0_noout): Likewise.
2929 (noncommutative_binary_comparison_result_used): Removed.
2930 (rsub_cmp0): New pattern.
2931 (rsub_cmp0_noout): Likewise.
2932 (extzvsi): Changed, keep only meaningful variants.
2933 (SQH, SEZ): New iterators.
2934 (SQH_postfix): New mode attribute.
2935 (SEZ_prefix): New code attribute.
2936 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2937 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2938 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2940 (noncommutative_operator): Check the availability of barrel
2943 2018-04-09 Richard Biener <rguenther@suse.de>
2945 PR tree-optimization/85284
2946 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2947 Only use the niter constraining form of simple_iv when the exit
2950 2018-04-09 Tom de Vries <tom@codesourcery.com>
2953 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2954 (define_expand "*memory_barrier"): New define_expand.
2955 (define_insn "memory_barrier"): New insn.
2957 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2959 PR rtl-optimization/80463
2960 PR rtl-optimization/83972
2961 PR rtl-optimization/83480
2963 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2964 correct producer for the insn.
2965 (tidy_control_flow): Fixup seqnos in case of debug insns.
2967 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2969 PR rtl-optimization/83913
2971 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2972 different sched-times when merging exprs.
2974 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2976 PR rtl-optimization/83962
2978 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2979 tidy_fallthru_edge and tidy_control_flow.
2981 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2983 PR rtl-optimization/83530
2985 * sel-sched.c (force_next_insn): New global variable.
2986 (remove_insn_for_debug): When force_next_insn is true, also leave only
2987 next insn in the ready list.
2988 (sel_sched_region): When the region wasn't scheduled, make another pass
2989 over it with force_next_insn set to 1.
2991 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2993 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2995 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2996 for interrupt control.
2997 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2998 functions for interrupt control.
2999 * config/nds32/nds32-intrinsic.md: Likewise.
3000 * config/nds32/nds32_intrinsic.h: Likewise.
3001 * config/nds32/nds32.h (nds32_builtins): Likewise.
3003 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3005 * config/nds32/nds32.c (nds32_init_machine_status,
3006 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3007 strict_aligned_p field.
3008 (nds32_expand_to_rtl_hook): New function.
3009 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3010 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3012 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3013 Chung-Ju Wu <jasonwucj@gmail.com>
3015 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3016 * config/nds32/nds32-n7.md: New file.
3017 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3018 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3020 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3021 * config/nds32/nds32.md (pipeline_model): Add n7.
3022 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3023 * config/nds32/pipelines.md: Include n7 settings.
3025 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3026 Chung-Ju Wu <jasonwucj@gmail.com>
3028 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3029 * config/nds32/nds32-e8.md: New file.
3030 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3031 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3033 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3034 * config/nds32/nds32.md (pipeline_model): Add e8.
3035 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3036 * config/nds32/pipelines.md: Include e8 settings.
3038 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3039 Chung-Ju Wu <jasonwucj@gmail.com>
3041 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3042 * config/nds32/nds32-n8.md: New file.
3043 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3044 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3046 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3047 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3048 * config/nds32/nds32.md (pipeline_model): Add n8.
3049 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3050 * config/nds32/pipelines.md: Include n8 settings.
3052 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3053 Chung-Ju Wu <jasonwucj@gmail.com>
3055 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3056 * config/nds32/nds32-n9-2r1w.md: New file.
3057 * config/nds32/nds32-n9-3r2w.md: New file.
3058 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3059 nds32_register_ports): New or modify for cpu n9.
3060 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3062 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3063 * config/nds32/nds32-utils.c: New file.
3064 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3065 TARGET_MUL_SLOW): Define.
3066 * config/nds32/nds32.md (pipeline_model): New attribute.
3067 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3068 New options that support cpu n9.
3069 * config/nds32/pipelines.md: Include n9 settings.
3070 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3072 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3074 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3075 information if necessary.
3076 (output_cond_branch_compare_zero): Likewise.
3077 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3078 (nds32_target_alignment): Refine for alignment.
3079 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3080 (FUNCTION_BOUNDARY): Modify.
3081 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3083 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3085 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3087 * config/nds32/constants.md (unspec_volatile_element): Add values for
3088 TLB operation and data prefetch.
3089 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3090 functions for TLB operation and data prefetch.
3091 * config/nds32/nds32-intrinsic.md: Likewise.
3092 * config/nds32/nds32_intrinsic.h: Likewise.
3093 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3094 (nds32_print_operand): Likewise.
3095 * config/nds32/nds32.h (nds32_builtins): Likewise.
3097 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3098 Andrew Pinski <pinsika@gcc.gnu.org>
3101 * match.pd: Use constant_boolean_node of correct type instead of
3102 boolean_true_node or boolean_false_node for simplifying
3103 pointer comparisons to zero.
3105 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3107 PR tree-optimization/80021
3108 * tree.c (verify_type_variant): Make error call in verify_variant_match
3109 translatable and remove final full stop.
3111 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3113 * config/nds32/constants.md (unspec_volatile_element): Add
3114 UNSPEC_VOLATILE_EH_RETURN.
3115 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3116 nds32_output_stack_pop): Support dwarf exception handling process.
3117 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3118 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3119 exception handling process.
3120 (nds32_compute_stack_frame): Likewise.
3121 (nds32_return_addr_rtx): Likewise.
3122 (nds32_initial_elimination_offset): Likewise.
3123 (nds32_expand_prologue): Likewise.
3124 (nds32_expand_epilogue): Likewise.
3125 (nds32_dynamic_chain_address): New function.
3126 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3128 (DYNAMIC_CHAIN_ADDRESS): Define.
3129 (EH_RETURN_DATA_REGNO): Define.
3130 (EH_RETURN_STACKADJ_RTX): Define.
3131 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3132 patterns for dwarf exception handling.
3134 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3136 * config/nds32/nds32.h: Clean up obsolete macros.
3138 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3140 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3141 Add enum values for particular instructions.
3142 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3143 particular intrinsic functions.
3144 * config/nds32/nds32-intrinsic.md: Likewise.
3145 * config/nds32/nds32_intrinsic.h: Likewise.
3146 * config/nds32/nds32.h (nds32_builtins): Likewise.
3147 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3148 (btst, ave): New patterns for particular instructions.
3150 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3152 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3153 Add enum values for atomic load/store and memory sync.
3154 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3156 * config/nds32/nds32-intrinsic.md: Likewise.
3157 * config/nds32/nds32_intrinsic.h: Likewise.
3158 * config/nds32/nds32.h (nds32_builtins): Likewise.
3160 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3162 PR tree-optimization/85257
3163 * fold-const.c (native_encode_vector): If not all elts could fit
3164 and off is -1, return 0 rather than offset.
3165 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3166 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3167 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3168 adjust buffer in native_interpret_expr call.
3170 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3172 * config/nds32/constants.md (unspec_volatile_element): Add cache
3173 control enum values.
3174 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3175 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3176 * config/nds32/nds32.c (nds32_cctl_names): New.
3177 (nds32_print_operand): Handle cache control register names.
3178 * config/nds32/nds32.h (nds32_builtins): New enum values.
3179 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3181 * config/nds32/nds32.md (type): Add mmu.
3182 * config/nds32/pipelines.md (simple_insn): Add mmu.
3184 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3186 * config/nds32/nds32.md (type): Remove call.
3187 * config/nds32/pipelines.md (simple_insn): Likewise.
3189 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3191 * config/nds32/constants.md (unspec_volatile_element): Add
3192 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3193 UNSPEC_VOLATILE_FMFCFG.
3194 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3195 description for fmfcfg and fmfcsr.
3196 (bdesc_1arg): Add fmtcsr.
3197 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3198 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3199 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3200 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3201 unspec_fmfcfg): New patterns.
3202 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3203 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3204 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3205 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3206 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3207 __nds32__fmfcfg): Define.
3209 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3211 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3212 intrinsic register names.
3213 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3214 intrinsic register enum values and macros.
3216 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3218 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3219 for load/store addressing form.
3220 (nds32_print_operand_address): Likewise.
3222 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3225 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3226 based on LABEL_REF. Remove useless assertion.
3227 (pic_address_needs_scratch): Fix formatting.
3228 (sparc_legitimize_pic_address): Minor tweaks.
3229 (sparc_delegitimize_address): Adjust assertion accordingly.
3230 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3231 into symbolic_operand.
3232 (movsi_high_pic_label_ref): Likewise.
3233 (movsi_lo_sum_pic_label_ref): Likewise.
3234 (movdi_pic_label_ref): Likewise.
3235 (movdi_high_pic_label_ref): Likewise.
3236 (movdi_lo_sum_pic_label_ref): Likewise.
3238 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3240 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3241 custom LIB_SPEC setup.
3243 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3244 Kito Cheng <kito.cheng@gmail.com>
3246 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3247 * config/riscv/freebsd.h: New.
3249 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3251 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3252 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3255 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3256 Kito Cheng <kito.cheng@gmail.com>
3258 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3259 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3260 * config/nds32/nds32-protos.h (nds32_output_call,
3261 nds32_output_return): Declare.
3262 * config/nds32/nds32.md: Refine all the call and return patterns.
3264 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3267 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3268 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3270 PR rtl-optimization/84872
3271 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3272 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3275 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3277 * expr.c (copy_blkmode_to_reg): Revert 254862.
3278 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3280 2018-04-06 Richard Biener <rguenther@suse.de>
3283 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3284 after seeing a component reference with an adjacent field. Treat
3285 refs to arrays at struct end of external decls similar to
3286 refs to unconstrained commons.
3288 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3291 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3292 look through SAVE_EXPRs with non-side-effects argument. Adjust
3294 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3297 2018-04-06 Richard Biener <rguenther@suse.de>
3300 * alias.c (find_base_term): New wrapper around find_base_term
3301 unwinding CSELIB_VAL_PTR changes.
3302 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3305 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3307 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3309 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3310 constant definitions.
3311 ("nop"): lr 0,0 -> nopr r0
3312 ("nop_lr0", "nop_lr1"): New insn definitions.
3314 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3316 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3317 NDS32_V3PUSH_AVAILABLE_P macro.
3319 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3320 Chung-Ju Wu <jasonwucj@gmail.com>
3322 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3323 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3324 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3325 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3326 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3327 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3328 * config/nds32/constraints.md: New constraints and checking for hard
3329 float configuration.
3330 * config/nds32/iterators.md: New mode iterator and attribute for hard
3331 float configuration.
3332 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3334 * config/nds32/nds32-fpu.md: New file.
3335 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3336 deal with hard float code generation.
3337 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3339 (abi_type, float_reg_number): New enum type.
3340 * config/nds32/nds32-predicates.c: New predicates for hard float.
3341 * config/nds32/nds32-protos.h: Declare functions for hard float.
3342 * config/nds32/nds32.c: Implementation for hard float configuration.
3343 * config/nds32/nds32.h: Definitions for hard float configuration.
3344 * config/nds32/nds32.md: Include hard float machine description and
3345 modify patterns for hard float configuration.
3346 * config/nds32/nds32.opt: New options for hard float configuration.
3347 * config/nds32/predicates.md: New predicates for hard float
3350 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3352 * common/config/nds32/nds32-common.c
3353 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3355 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3358 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3359 CONSTRUCTOR_ELT (ctor, ...)->value.
3361 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3364 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3366 2018-04-05 Tom de Vries <tom@codesourcery.com>
3369 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3372 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3373 Kito Cheng <kito.cheng@gmail.com>
3375 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3376 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3377 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3378 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3380 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3381 Kito Cheng <kito.cheng@gmail.com>
3383 * config/nds32/constraints.md (Ufe): New memory constraint.
3384 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3385 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3386 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3388 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3389 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3391 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3393 * config/nds32/nds32.md: Use optimize_size in the condition for
3394 alu-shift instructions.
3396 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3398 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3400 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3402 * config/nds32/nds32.md (negsi2): Refine pattern.
3404 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3405 Chung-Ju Wu <jasonwucj@gmail.com>
3407 * config/nds32/iterators.md (shift_rotate): New code iterator.
3408 (shift): New code attribute.
3409 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3410 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3411 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3412 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3413 bit-wise operations.
3414 (andsi3, *andsi3): Ditto.
3415 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3416 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3417 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3418 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3419 nds32_ior_operand, nds32_xor_operand): New predicates.
3421 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3423 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3424 (addsi3, subsi3): ... this.
3426 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3428 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3430 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3432 * config/nds32/nds32.md: Adjust indention.
3434 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3436 * config/nds32/nds32.md (feature): New attribute.
3438 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3440 * config/nds32/nds32.md (subtype): New attribute.
3442 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3445 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3446 expansion to perform a bitwise AND of the argument followed by a
3447 boolean negation of the result.
3449 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3451 PR rtl-optimization/84878
3452 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3453 the basic block. Assert the use reference is not artificial and that
3454 it has an associated insn.
3456 2018-04-04 Michael Matz <matz@suse.de>
3458 * builtins.c (compute_objsize): Pass correct operand
3459 to array_at_struct_end_p.
3461 2018-04-04 Richard Biener <rguenther@suse.de>
3464 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3465 from contexts for DINFO_LEVEL_TERSE and below.
3467 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3469 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3470 resiter_operand condition.
3471 * config/nds32/nds32.md (*move<mode>): Ditto.
3473 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3474 Monk Chiang <sh.chiang04@gmail.com>
3476 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3478 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3480 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3482 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3483 Kito Cheng <kito.cheng@gmail.com>
3485 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3486 nds32_cond_code_str, output_cond_branch,
3487 output_cond_branch_compare_zero, nds32_expand_cbranch,
3488 nds32_expand_cstore, nds32_expand_movcc,
3489 nds32_output_cbranchsi4_equality_zero,
3490 nds32_output_cbranchsi4_equality_reg,
3491 nds32_output_cbranchsi4_equality_reg_or_const_int,
3492 nds32_output_cbranchsi4_greater_less_zero: New functions.
3493 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3494 nds32_expand_cstore, nds32_expand_movcc,
3495 nds32_output_cbranchsi4_equality_zero,
3496 nds32_output_cbranchsi4_equality_reg,
3497 nds32_output_cbranchsi4_equality_reg_or_const_int,
3498 nds32_output_cbranchsi4_greater_less_zero): Declare.
3499 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3500 nds32_rimm11s_operand): New predicates.
3501 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3502 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3505 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3507 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3508 * config/nds32/nds32.md: Ditto.
3509 * config/nds32/pipelines.md: Ditto.
3511 2018-04-04 Richard Biener <rguenther@suse.de>
3513 PR tree-optimization/85168
3514 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3515 propagating abnormals.
3517 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3519 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3521 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3522 Kito Cheng <kito.cheng@gmail.com>
3524 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3525 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3526 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3527 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3528 * config/nds32/nds32.md (sibcall_internal): New.
3529 (sibcall_register): Remove.
3530 (sibcall_immediate): Remove.
3531 (sibcall_value_internal): New.
3532 (sibcall_value_register): Remove.
3533 (sibcall_value_immediate): Remove.
3534 * config/nds32/predicates.md (nds32_general_register_operand): New.
3535 (nds32_call_address_operand): New.
3537 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3539 PR rtl-optimization/85167
3540 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3541 bb_defs if *split_p, instead preinitialize it to NULL.
3543 PR tree-optimization/85156
3544 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3545 evaluating the argument multiple times.
3547 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3549 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3551 (_mm_cvtpd_ps): Likewise.
3552 (_mm_cvttpd_epi32): Likewise.
3553 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3554 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3555 vector, pixel, and bool following altivec.h include.
3557 2018-04-03 Martin Sebor <msebor@redhat.com>
3559 * doc/extend.texi (Common Function Attributes): Clarify.
3560 (const attribute): Likewise.
3561 (pure attribute): Likewise.
3563 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3566 * config/i386/i386.c (ix86_expand_vector_set): Use
3567 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3569 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3571 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3572 instructions when changing rounding bits to preserve precision bits
3573 in the x87 control word.
3575 2018-04-03 Martin Liska <mliska@suse.cz>
3577 PR tree-optimization/82491
3578 * rtl.h (strip_offset_and_add): Replace += suboffset with
3579 poly_uint64 () + suboffset.
3581 2018-03-29 Martin Liska <mliska@suse.cz>
3582 Martin Jambor <mjambor@suse.cz>
3585 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3586 param_type is not an integral or pointer type.
3588 2018-04-03 Richard Biener <rguenther@suse.de>
3590 * sese.h (recompute_all_dominators): Remove.
3592 2018-04-02 Martin Sebor <msebor@redhat.com>
3594 * doc/invoke.texi (-Wrestrict): Fix typos.
3596 2018-04-02 Jim Wilson <jimw@sifive.com>
3598 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3599 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3600 (<optab>di3, <optab>si3_extend): Likewise.
3601 (<optab>si3_mask, <optab>si3_mask_1): New.
3602 (<optab>di3_mask, <optab>di3_mask_1): New.
3603 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3604 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3605 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3607 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3609 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3612 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3614 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3615 (nds32_canonicalize_comparison): New function.
3617 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3618 Kito Cheng <kito.cheng@gmail.com>
3619 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3621 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3622 * config/nds32/constants.md (unspec_volatile_element): Add
3623 UNSPEC_VOLATILE_RELAX_GROUP.
3624 * config/nds32/nds32-relax-opt.c: New file.
3625 * config/nds32/nds32-predicates.c
3626 (nds32_symbol_load_store_p): New function.
3627 * config/nds32/nds32-protos.h
3628 (nds32_symbol_load_store_p): Declare function.
3629 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3630 * config/nds32/nds32.c
3631 (nds32_register_pass): New function to register pass.
3632 (nds32_register_passes): New function to register passes.
3633 * config/nds32/nds32.md (relax_group): New pattern.
3634 * config/nds32/nds32.opt (mrelax-hint): New option.
3635 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3637 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3639 * config/nds32/t-nds32: Modify files dependency.
3641 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3643 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3644 (PROFILE_HOOK): Define its implementation.
3646 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3648 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3649 type and 32-bit size.
3651 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3654 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3655 (V_128_256): New mode iterator.
3656 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3657 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3658 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3660 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3661 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3663 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3666 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3667 NaN inputs correctly.
3669 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3672 * config/rs6000/vsx.md (??r): New mode attribute.
3673 (*vsx_mov<mode>_64bit): Use it.
3674 (*vsx_mov<mode>_32bit): Likewise.
3676 2018-03-30 Martin Sebor <msebor@redhat.com>
3678 PR tree-optimization/84818
3679 * builtins.c (check_access): Use warning_n.
3681 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3684 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3686 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3689 2018-03-30 Julia Koval <julia.koval@intel.com>
3692 * x86-tune.def (movx, partial_reg_dependency): Enable for
3695 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3698 * lra-constraints.c (process_alt_operands): Move setting
3699 this_alternative_matches below.
3701 2018-03-29 Martin Liska <mliska@suse.cz>
3704 * doc/invoke.texi: Document how LTO works with debug info.
3705 Describe auto-load support of binutils. Mention 'x86-64'
3706 as valid option value of -march option.
3708 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3710 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3713 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3714 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3715 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3718 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3721 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3722 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3723 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3724 for RS6000_BTM_POWERPC64.
3725 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3726 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3727 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3732 2018-03-28 Carl Love <cel@us.ibm.com>
3735 2017-09-27 Carl Love <cel@us.ibm.com>
3737 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3738 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3739 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3742 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3744 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3745 instead of __vector bool.
3746 (_mm_max_pu8): Likewise.
3747 (_mm_min_pi16): Likewise.
3749 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3752 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3753 (DIVWEUO): Likewise.
3755 (DIVDEUO): Likewise.
3756 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3757 DIVWEUO and DIVDEUO.
3758 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3759 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3760 (div_extend): Likewise.
3761 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3763 (__builtin_divweuo): Likewise.
3764 (__builtin_divdeo): Likewise.
3765 (__builtin_divdeuo): Likewise.
3767 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3770 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3771 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3773 PR tree-optimization/82004
3774 * gimple-match-head.c (optimize_pow_to_exp): New function.
3775 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3776 Don't fold to exp if optimize_pow_to_exp is false.
3778 2018-03-28 Martin Liska <mliska@suse.cz>
3781 * calls.c (initialize_argument_information): Fix trailing space.
3782 * common.opt: Fix typo and provide better explanation for
3783 -fsanitize-coverage option.
3784 * config/i386/i386.opt: Fix typo.
3786 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3787 Martin Liska <mliska@suse.cz>
3790 * gimplify.c (asan_poison_variable): Don't do the check for
3791 gimplify_omp_ctxp here.
3792 (gimplify_decl_expr): Do it here.
3793 (gimplify_target_expr): Likewise.
3795 2018-03-28 Martin Liska <mliska@suse.cz>
3798 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3799 chkp_type_bounds_count if MPX is not enabled.
3801 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3803 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3805 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3808 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3809 function to create the function decl for complex long double
3810 multiply and divide for -mabi=ieeelongdouble.
3811 (init_float128_ieee): Call it.
3813 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3816 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3817 -fcf-protection=branch -mibt.
3818 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3820 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3823 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3825 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3828 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3829 extern array declarations.
3831 2018-03-27 Richard Biener <rguenther@suse.de>
3834 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3835 explicit single_use checks.
3837 2018-03-27 Richard Biener <rguenther@suse.de>
3839 PR tree-optimization/85082
3840 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3843 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3845 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3846 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3847 Turn on fasynchronous-unwind-tables and funwind-tables.
3849 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3852 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3853 (*bmi_blsr_<mode>_ccz): Ditto.
3855 2018-03-26 Tom de Vries <tom@codesourcery.com>
3857 PR tree-optimization/85063
3858 * omp-general.c (offloading_function_p): New function. Factor out
3860 * omp-offload.c (pass_omp_target_link::gate): ... here.
3861 * omp-general.h (offloading_function_p): Declare.
3862 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3863 with attribute omp declare target for offloading functions.
3865 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3867 PR tree-optimization/84005
3868 * tree-data-ref.h (get_base_for_alignment): Declare.
3869 * tree-data-ref.c (get_base_for_alignment_1): New function.
3870 (get_base_for_alignment): Likewise.
3871 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3872 get_base_for_alignment to find a suitable base object, instead
3873 of always using drb->base_address.
3875 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3878 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3879 known size by default.
3881 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3884 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3885 and non BLKmode operands.
3887 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3890 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3891 Clean up attributes.
3893 2018-03-23 Richard Biener <rguenther@suse.de>
3896 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3897 we are going to emit early debug for LTO.
3899 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3902 * function.c (match_asm_constraints_1): Don't optimize if input
3903 doesn't satisfy general_operand predicate for output's mode.
3906 * alias.c (write_dependence_p): Don't require for x_canonicalized
3907 non-VOIDmode if x has VOIDmode.
3910 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3911 just don't try to optimize it rather than assert it never happens.
3913 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3915 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3916 macro expansions for definition of ST_INTERNAL_<mode> and
3917 LD_INTERNAL_<mode> builtins.
3918 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3920 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3922 (altivec_expand_st_builtin): Likewise.
3923 (altivec_expand_builtin): Remove calls to deleted functions.
3924 (rs6000_address_for_altivec): Delete this function.
3925 * config/rs6000/vector.md: Remove expands for
3926 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3928 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3931 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3932 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3933 re-computing once computed.
3934 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3935 (arm_init_machine_status): Initialize
3936 machine->static_chain_stack_bytes.
3938 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3941 * doc/extend.texi: Add four new prototypes for vec_ld.
3942 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3943 definitions for more logical presentation.
3944 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3945 entries for V1TI variants of __builtin_altivec_ld builtin.
3946 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3947 handling of V1TI variant of LVX icode pattern.
3948 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3949 (rs6000_gimple_fold_builtin): Likewise.
3950 (altivec_init_builtins): Add code to define
3951 __builtin_altivec_lvx_v1ti function.
3953 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3956 * function.c (match_asm_constraints_1): Don't do the optimization
3957 if input isn't a REG, SUBREG, MEM or constant.
3959 2018-03-22 Tom de Vries <tom@codesourcery.com>
3961 PR tree-optimization/84956
3962 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3963 bb_has_abnormal_pred.
3965 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3968 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3969 DECL_INITIAL (decl) to decl at the end.
3970 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3973 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3975 * doc/extend.texi (__builtin_tgmath): Document when complex
3976 integer types are treated as _Complex _Float64.
3978 2018-03-21 Tom de Vries <tom@codesourcery.com>
3980 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3982 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3984 PR tree-optimization/84960
3985 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3986 if it is ENTRY block, move them into single succ of ENTRY in that case.
3988 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3990 PR tree-optimization/84811
3991 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3992 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3993 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3994 (known_subrange_p): Update accordingly. Cast each value involved
3995 in the size comparison, rather than casting the result of the
3998 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4000 PR tree-optimization/84982
4001 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4002 by flipping the least significant bit rather than all bits from
4003 bitpos to bitpos + bitsize - 1.
4005 2018-03-21 Nathan Sidwell <nathan@acm.org>
4007 * doc/extend.texi (Deprecated Features): Remove mention of
4008 long-deleted deprecations.
4010 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4013 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4014 * configure: Regenerate.
4016 2018-03-21 Tom de Vries <tom@codesourcery.com>
4018 PR tree-optimization/83126
4019 * tree-parloops.c (num_phis): New function.
4020 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4022 2018-03-21 Nathan Sidwell <nathan@acm.org>
4024 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4025 mention anon-struct/union members and trailing attributes.
4027 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4029 PR tree-optimization/84969
4030 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4031 builtin memset partitions if they set different rhs values.
4033 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4035 PR rtl-optimization/84989
4036 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4037 VEC_DUPLICATE with scalar result mode.
4039 2018-03-21 Martin Liska <mliska@suse.cz>
4042 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4043 not intended return statement.
4045 2018-03-21 Martin Liska <mliska@suse.cz>
4048 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4049 (chkp_find_bound_slots_1): Limit number of iterations.
4051 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4054 * Minor grammar fixes for x86 options.
4056 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4059 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4060 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4062 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4065 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4066 (altivec_lvx_<mode>_1op): Likewise.
4067 (altivec_stvx_<mode>_2op): Likewise.
4068 (altivec_stvx_<mode>_1op): Likewise.
4069 (altivec_lvx_<VM2:mode>): New define_expand.
4070 (altivec_stvx_<VM2:mode>): Likewise.
4071 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4072 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4073 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4074 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4075 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4076 (rs6000_gen_lvx): Likewise.
4077 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4078 (altivec_expand_stv_builtin): Likewise.
4079 (altivec_expand_builtin): Likewise.
4080 * config/rs6000/vector.md: Likewise.
4082 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4085 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4088 2018-03-20 Richard Biener <rguenther@suse.de>
4091 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4092 sign-conversions as zero, fall back to standard scalar_stmt
4095 2018-03-20 Martin Liska <mliska@suse.cz>
4098 * predict.c (rebuild_frequencies): Handle case when we have
4099 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4101 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4104 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4105 flag_section_anchors.
4106 * varasm.c (use_blocks_for_decl_p): Remove hack for
4107 dw2_force_const_mem.
4110 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4112 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4113 be created, use lowpart_subreg of operands[0] rather than operands[0]
4115 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4116 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4117 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4118 and n constraint instead of aarch64_shift_imm_di and Usd.
4119 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4120 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4122 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4125 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4126 to favor GPR over NEON registers.
4127 (<shift>di3_neon): Likewise.
4129 2018-03-20 Tom de Vries <tom@codesourcery.com>
4132 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4133 (nvptx_process_pars): Emit bar.sync asap and alap.
4135 2018-03-20 Tom de Vries <tom@codesourcery.com>
4138 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4139 seen_label if seen_label is already set.
4141 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4144 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4145 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4146 Use 1U instead of 1. Formatting fixes.
4149 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4150 instead of TREE_TYPE (s1) for the return value.
4152 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4154 PR tree-optimization/84946
4155 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4156 bitsize + bitsize in poly_uint64 rather than poly_int64.
4159 * dwarf2asm.c: Include fold-const.c.
4160 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4161 of decl rather than decl itself.
4163 PR rtl-optimization/84643
4164 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4166 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4169 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4170 calling assemble_variable.
4172 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4175 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4176 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4178 2018-03-19 Jim Wilson <jimw@sifive.com>
4181 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4182 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4183 (riscv_first_stack_step): Likewise.
4184 (riscv_option_override): Use STACK_BOUNDARY instead of
4186 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4188 (BIGGEST_ALIGNMENT): Set to 128.
4189 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4190 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4193 2018-03-19 Richard Biener <rguenther@suse.de>
4195 PR tree-optimization/84933
4196 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4197 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4199 2018-03-19 Richard Biener <rguenther@suse.de>
4201 PR tree-optimization/84859
4202 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4203 (cond_if_else_store_replacement): Perform sinking operation on
4204 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4205 Generalize what a BB with a single eligible store is.
4207 2018-03-19 Richard Biener <rguenther@suse.de>
4209 PR tree-optimization/84929
4210 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4211 chrec_is_positive against non-chrec arg.
4213 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4216 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4218 2018-03-18 Martin Liska <mliska@suse.cz>
4220 PR rtl-optimization/84635
4221 * regrename.c (build_def_use): Use matches_mode only when
4224 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4226 PR tree-optimization/84913
4227 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4228 vectorize chains of COND_EXPRs.
4230 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4232 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4234 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4236 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4238 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4240 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4242 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4243 Kito Cheng <kito.cheng@gmail.com>
4245 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4246 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4247 (nds32_adjust_reg_alloc_order): New function.
4248 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4250 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4252 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4253 nds32_print_operand, nds32_print_operand_address): Use
4254 HOST_WIDE_INT_PRINT_DEC instead.
4256 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4258 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4260 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4263 * config/i386/i386.c (initial_ix86_tune_features,
4264 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4266 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4267 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4268 rather than 1u << ix86_tune. Formatting fix.
4269 (ix86_option_override_internal): Change ix86_arch_mask from
4270 unsigned int to unsigned HOST_WIDE_INT, initialize to
4271 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4272 (ix86_function_specific_restore): Likewise.
4274 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4277 * postreload.c (reload_combine_recognize_pattern): Perform
4278 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4279 truncate_int_for_mode the result for the destination's mode.
4282 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4283 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4285 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4288 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4291 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4292 regno_allocno_class_array and sorted_pseudos.
4293 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4294 insns where regno is used.
4296 2018-03-16 Martin Liska <mliska@suse.cz>
4299 * multiple_target.c (create_dispatcher_calls): Redirect
4300 reference in the symbol table.
4302 2018-03-16 Martin Liska <mliska@suse.cz>
4305 * multiple_target.c (create_dispatcher_calls): Redirect also
4308 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4312 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4313 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4314 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4316 2018-03-16 Julia Koval <julia.koval@intel.com>
4318 * doc/invoke.texi (Skylake Server): Add CLWB.
4319 Cannonlake): Remove CLWB.
4321 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4323 PR tree-optimization/84841
4324 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4326 (FLOAT_ONE_CONST_TYPE): Define.
4327 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4328 (sort_by_operand_rank): Put entries with higher constant_type last
4329 rather than first to match comments.
4331 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4333 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4336 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4339 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4340 check and use address of target always.
4342 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4345 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4346 (indirect_thunk_bnd_needed): Likewise.
4347 (indirect_thunks_used): Likewise.
4348 (indirect_thunks_bnd_used): Likewise.
4349 (indirect_return_needed): New.
4350 (indirect_return_bnd_needed): Likewise.
4351 (output_indirect_thunk_function): Add a bool argument for
4353 (output_indirect_thunk_function): Don't generate alias for
4354 function return thunk.
4355 (ix86_code_end): Call output_indirect_thunk_function to generate
4356 function return thunks.
4357 (ix86_output_function_return): Set indirect_return_bnd_needed
4358 and indirect_return_needed instead of indirect_thunk_bnd_needed
4359 and indirect_thunk_needed.
4361 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4363 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4364 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4365 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4367 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4368 Paul Hua <paul.hua.gm@gmail.com>
4371 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4373 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4375 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4376 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4379 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4382 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4383 instead of GET_MODE_SIZE when comparing Units.
4385 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4388 * varasm.c (hash_section): Return an unchangeble hash value
4389 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4390 Return !aarch64_can_use_per_function_literal_pools_p ().
4392 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4395 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4396 rather than address of cmode as last argument to prepare_cmp_insn.
4398 2018-03-15 Julia Koval <julia.koval@intel.com>
4400 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4401 F_AVX512VNNI, F_AVX512BITALG): New.
4403 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4406 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4407 insn for floating-point loads and stores.
4409 2018-03-14 Carl Love <cel@us.ibm.com>
4411 * config/rs6000/rs6000-c.c: Add macro definitions for
4412 ALTIVEC_BUILTIN_VEC_PERMXOR.
4413 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4414 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4415 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4416 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4418 * config/doc/extend.texi: Add prototypes for vec_permxor.
4420 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4423 * diagnostic-show-locus.c (class layout_point): Convert m_line
4424 from int to linenum_type.
4425 (line_span::comparator): Use linenum "compare" function when
4426 comparing line numbers.
4427 (test_line_span): New function.
4428 (layout_range::contains_point): Convert param "row" from int to
4430 (layout_range::intersects_line_p): Likewise.
4431 (layout::will_show_line_p): Likewise.
4432 (layout::print_source_line): Likewise.
4433 (layout::should_print_annotation_line_p): Likewise.
4434 (layout::print_annotation_line): Likewise.
4435 (layout::print_leading_fixits): Likewise.
4436 (layout::annotation_line_showed_range_p): Likewise.
4437 (struct line_corrections): Likewise for field m_row.
4438 (line_corrections::line_corrections): Likewise for param "row".
4439 (layout::print_trailing_fixits): Likewise.
4440 (layout::get_state_at_point): Likewise.
4441 (layout::get_x_bound_for_row): Likewise.
4442 (layout::print_line): Likewise.
4443 (diagnostic_show_locus): Likewise for locals "last_line" and
4445 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4446 * input.c (selftest::test_linenum_comparisons): New function.
4447 (selftest::input_c_tests): Call it.
4448 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4449 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4450 * selftest.h (ASSERT_GT): New macro.
4451 (ASSERT_GT_AT): New macro.
4452 (ASSERT_LT): New macro.
4453 (ASSERT_LT_AT): New macro.
4455 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4457 PR rtl-optimization/84780
4458 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4460 2018-03-14 Martin Liska <mliska@suse.cz>
4462 * tree.c (record_node_allocation_statistics): Use
4463 get_stats_node_kind.
4464 (get_stats_node_kind): New function extracted from
4465 record_node_allocation_statistics.
4466 (free_node): Use get_stats_node_kind.
4468 2018-03-14 Richard Biener <rguenther@suse.de>
4470 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4471 that the value-set of ANTIC_IN doesn't grow.
4474 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4476 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4477 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4479 2018-03-14 Julia Koval <julia.koval@intel.com>
4481 * config.gcc (icelake-client, icelake-server): New.
4483 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4484 (initial_ix86_arch_features): Ditto.
4485 (PTA_SKYLAKE): Add SGX.
4486 (PTA_ICELAKE): Remove.
4487 (PTA_ICELAKE_CLIENT): New.
4488 (PTA_ICELAKE_SERVER): New.
4489 (ix86_option_override_internal): Split up icelake on icelake client and
4491 (get_builtin_code_for_version): Ditto.
4492 (fold_builtin_cpu): Ditto.
4493 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4494 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4495 * config/i386/i386.h (processor_type): Ditto.
4496 * doc/invoke.texi: Ditto.
4498 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4501 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4502 INTEGER_CST offset, add it together with bitpos / 8 and
4503 sign extend based on POINTER_SIZE.
4507 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4510 * config/i386/constraints.md (Yc): New register constraint.
4511 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4512 Use Yc constraint for alternative 2 of operand 0. Remove
4513 preferred_for_speed attribute.
4515 2018-03-14 Richard Biener <rguenther@suse.de>
4517 PR tree-optimization/84830
4518 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4519 with the old one to avoid oscillations.
4521 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4524 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4526 (assign_by_spills): Return a flag of reload assignment failure.
4527 Do not process the reload assignment failures. Do not spill other
4528 reload pseudos if they has the same reg class. Update n if
4530 (lra_assign): Add a return arg. Set up from the result of
4531 assign_by_spills call.
4532 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4533 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4534 usage_insns if it is not NULL.
4535 (spill_hard_reg_in_range): New function.
4536 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4537 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4538 function prototypes.
4539 (lra_assign): Change prototype.
4540 * lra.c (lra): Add code to deal with fails by splitting hard reg
4543 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4545 * config/riscv/riscv.opt (mrelax): New option.
4546 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4547 "norelax" when riscv_mrelax is disabled.
4548 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4550 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4553 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4554 reassociation for int modes.
4556 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4558 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4559 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4561 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4562 * config/aarch64/aarch64-sve.md
4563 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4564 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4565 (*extend<mode><Vwide>2): Rename to...
4566 (aarch64_sve_extend<mode><Vwide>2): ...this.
4567 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4568 renaming the old pattern to...
4569 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4571 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4572 define_expand, renaming the old pattern to...
4573 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4574 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4575 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4576 account when deciding which SVE instruction the optab should use.
4577 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4579 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4581 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4582 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4583 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4584 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4585 (tlsdesc_small_<mode>): Turn a define_expand and use
4586 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4587 (tlsdesc_small_advsimd_<mode>): ...this.
4588 (tlsdesc_small_sve_<mode>): New pattern.
4590 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4592 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4593 (UNSPEC_UMUL_HIGHPART): New constants.
4594 (MUL_HIGHPART): New int iteraor.
4595 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4596 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4598 (*<su>mul<mode>3_highpart): New define_insn.
4600 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4603 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4606 2018-03-13 Martin Liska <mliska@suse.cz>
4609 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4611 (sem_item_optimizer::~sem_item_optimizer): Release it.
4612 (sem_item_optimizer::merge_classes): Register variable aliases.
4613 (sem_item_optimizer::fixup_pt_set): New function.
4614 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4615 * ipa-icf.h: Declare new variables and functions.
4617 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4620 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4621 integer_pow2p@2 and test integer_pow2p in condition.
4622 (A < 0 ? C : 0): Similarly for @1.
4625 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4626 characters starting at p contain '\0' character, don't look beyond
4630 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4631 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4634 * reg-stack.c (change_stack): Change update_end var from int to
4635 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4636 also call set_block_for_insn on the newly added insns and rescan.
4639 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4640 on the last operand.
4643 * tree.c (stabilize_reference_1): Return save_expr (e) for
4644 STATEMENT_LIST even if it doesn't have side-effects.
4646 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4648 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4650 2018-03-12 Renlin Li <renlin.li@arm.com>
4652 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4653 aarch64_output_scalar_simd_mov_immediate.
4655 2018-03-12 Martin Sebor <msebor@redhat.com>
4657 PR tree-optimization/83456
4658 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4659 for perfectly overlapping calls to memcpy.
4660 (gimple_fold_builtin_memory_chk): Same.
4661 (gimple_fold_builtin_strcpy): Handle no-warning.
4662 (gimple_fold_builtin_stxcpy_chk): Same.
4663 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4665 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4667 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4668 parameter. Use it for SFmode.
4669 (rs6000_function_arg_advance_1): Adjust.
4670 (rs6000_function_arg): Adjust.
4671 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4673 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4675 PR rtl-optimization/84169
4676 PR rtl-optimization/84780
4677 * combine.c (can_combine_p): Check for a 2-insn combination whether
4678 the destination register is used between the two insns, too.
4680 2018-03-12 Richard Biener <rguenther@suse.de>
4682 PR tree-optimization/84803
4683 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4684 for refs DR analysis didn't process.
4686 2018-03-12 Richard Biener <rguenther@suse.de>
4688 PR tree-optimization/84777
4689 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4690 force-vectorize loops ignore whether we are optimizing for size.
4692 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4694 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4695 (TARGET_MD_ASM_ADJUST): Define.
4697 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4698 Kito Cheng <kito.cheng@gmail.com>
4699 Chung-Ju Wu <jasonwucj@gmail.com>
4701 * config/nds32/nds32.c (nds32_compute_stack_frame,
4702 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4703 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4704 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4705 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4706 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4707 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4708 * config/nds32/nds32.md (prologue, epilogue): Use macro
4709 NDS32_V3PUSH_AVAILABLE_P to do checking.
4711 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4714 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4715 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4716 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4717 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4718 addition of most attributes on !orig_type_die or the attribute not
4719 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4721 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4722 Chung-Ju Wu <jasonwucj@gmail.com>
4724 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4726 * config/nds32/nds32.opt (mvh): New option.
4728 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4729 Chung-Ju Wu <jasonwucj@gmail.com>
4731 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4733 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4734 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4737 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4738 Chung-Ju Wu <jasonwucj@gmail.com>
4740 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4742 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4743 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4745 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4746 Kito Cheng <kito.cheng@gmail.com>
4747 Chung-Ju Wu <jasonwucj@gmail.com>
4749 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4750 UNSPEC_FFMISM and UNSPEC_FLMISM.
4751 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4752 for ffb, ffmism and flmism.
4753 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4754 (unspec_ffmism): Ditto.
4755 (unspec_flmism): Ditto.
4756 (nds32_expand_builtin_impl): Check if string extension is available.
4757 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4758 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4760 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4763 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4766 * lra-assigns.c (assign_by_spills): Return a flag of reload
4767 assignment failure. Do not process the reload assignment
4768 failures. Do not spill other reload pseudos if they has the same
4770 (lra_assign): Add a return arg. Set up from the result of
4771 assign_by_spills call.
4772 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4773 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4774 usage_insns if it is not NULL.
4775 (spill_hard_reg_in_range): New function.
4776 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4777 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4778 function prototypes.
4779 (lra_assign): Change prototype.
4780 * lra.c (lra): Add code to deal with fails by splitting hard reg
4783 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4786 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4788 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4791 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4792 (dw_val_node): Add val_symbolic_view.
4793 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4794 (symview_upper_bound): New.
4795 (new_line_info_table): Initialize symviews_since_reset.
4796 (dwarf2out_source_line): Count symviews_since_reset and set
4797 symview_upper_bound.
4798 (dw_val_equal_p): Handle symview.
4799 (add_AT_symview): New.
4800 (print_dw_val): Handle symview.
4801 (attr_checksum, attr_checksum_ordered): Likewise.
4802 (same_dw_val_p, size_of_die): Likewise.
4803 (value_format, output_die): Likewise.
4804 (add_high_low_attributes): Use add_AT_symview for entry_view.
4805 (dwarf2out_finish): Reset symview_upper_bound, clear
4808 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4811 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4812 Add strict argument and use it.
4813 (rs6000_split_multireg_move): Update for new strict argument.
4814 (mem_operand_gpr): Disallow all non-offsettable addresses.
4815 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4817 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4820 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4821 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4822 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4825 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4826 decl, use remap_type if we want to use the type.
4828 2018-03-09 Martin Sebor <msebor@redhat.com>
4830 PR tree-optimization/84526
4831 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4833 (builtin_access::generic_overlap): Be prepared to handle non-array
4836 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4838 PR rtl-optimization/84682
4839 * lra-constraints.c (process_address_1): Check is_address flag
4840 for address constraints.
4841 (process_alt_operands): Likewise.
4842 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4843 preprocess_constraints.
4844 * recog.h (preprocess_constraints): Add oploc parameter.
4846 * recog.c (preprocess_constraints): Test address_operand for
4847 CT_ADDRESS constraints.
4849 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4852 * lra-assigns.c (assign_by_spills): Return a flag of reload
4853 assignment failure. Do not process the reload assignment
4854 failures. Do not spill other reload pseudos if they has the same
4856 (lra_assign): Add a return arg. Set up from the result of
4857 assign_by_spills call.
4858 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4859 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4860 usage_insns if it is not NULL.
4861 (spill_hard_reg_in_range): New function.
4862 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4863 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4864 function prototypes.
4865 (lra_assign): Change prototype.
4866 * lra.c (lra): Add code to deal with fails by splitting hard reg
4869 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4872 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4873 Accept complain bool parameter. Only emit errors if it is true.
4874 (arm_parse_cpu_option_name): Likewise.
4875 (arm_target_thumb_only): Adjust callers of the above.
4876 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4877 prototype to take a default true bool parameter.
4878 (arm_parse_arch_option_name): Likewise.
4880 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4881 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4885 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4886 * configure: Regenerate.
4887 * configure.ac ("linker --version-script option"): New.
4888 ("linker soname option"): New.
4890 2018-03-09 Richard Biener <rguenther@suse.de>
4892 PR tree-optimization/84775
4893 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4894 immediate uses of predicate stmts and mark them modified.
4897 PR tree-optimization/84178
4898 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4900 (version_loop_for_if_conversion): Delay update_ssa call.
4901 (tree_if_conversion): Delay update_ssa until after predicate
4904 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4907 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4908 when the function accesses prior frames.
4910 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4913 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4914 gen_llsym, otherwise call maybe_gen_llsym.
4917 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4918 has ',' character inside of it.
4920 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4923 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4924 as clobbering CC_REGNUM.
4926 2018-03-08 Richard Biener <rguenther@suse.de>
4929 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4930 (follow_copies_to_constant): Do not follow SSA names registered
4933 2018-03-08 Richard Biener <rguenther@suse.de>
4935 PR tree-optimization/84178
4936 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4938 (version_loop_for_if_conversion): Delay update_ssa call.
4939 (tree_if_conversion): Delay update_ssa until after predicate
4942 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4944 PR tree-optimization/84178
4945 * tree-if-conv.c (release_bb_predicate): Remove the
4946 the assertion that the stmts have NULL use_ops.
4947 Discard the statements, asserting that they haven't
4948 yet been added to a BB.
4950 2018-03-08 Richard Biener <rguenther@suse.de>
4952 PR tree-optimization/84746
4953 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4954 (phi_translate): Pass in destination ANTIC_OUT set.
4955 (phi_translate_1): Likewise. For a simplified result lookup
4956 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4957 (phi_translate_set): Adjust.
4958 (do_pre_regular_insertion): Likewise.
4959 (do_pre_partial_partial_insertion): Likewise.
4961 2018-03-08 Martin Liska <mliska@suse.cz>
4963 PR gcov-profile/84735
4964 * doc/gcov.texi: Document usage of profile files.
4965 * gcov-io.h: Document changes in the format.
4967 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4971 * dwarf2out.c (struct dw_line_info_table): Update comments for
4973 (FORCE_RESET_NEXT_VIEW): New.
4974 (FORCE_RESETTING_VIEW_P): New.
4975 (RESETTING_VIEW_P): Check for -1 too.
4976 (ZERO_VIEW_P): Likewise.
4977 (new_line_info_table): Force-reset next view.
4978 (dwarf2out_begin_function): Likewise.
4979 (dwarf2out_source_line): Simplify zero_view_p initialization.
4980 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4981 view directly. Omit view when omitting .loc at line 0.
4983 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4985 PR tree-optimization/84740
4986 * tree-switch-conversion.c (process_switch): Call build_constructors
4987 only if info.phi_count is non-zero.
4989 PR tree-optimization/84739
4990 * tree-tailcall.c (find_tail_calls): Check call arguments against
4991 DECL_ARGUMENTS (current_function_decl) rather than
4992 DECL_ARGUMENTS (func) when checking for tail recursion.
4994 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4996 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4997 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4998 Volker Reichelt's entry and add entries for people that perform
4999 GCC fuzzy testing and report numerous bugs.
5001 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5004 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5005 readonly data in sdata, if that is disabled.
5006 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5007 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5008 -mreadonly-in-sdata option.
5010 2018-03-07 Martin Sebor <msebor@redhat.com>
5012 PR tree-optimization/84468
5013 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5014 basic block when looking for nul assignment.
5016 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5019 * except.h (output_function_exception_table): Adjust prototype.
5020 * except.c (output_function_exception_table): Remove FNNAME parameter
5021 and add SECTION parameter. Ouput one part of the table at a time.
5022 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5023 the first part of the exception table and emit unwind directives.
5024 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5025 (i386_pe_seh_cold_init): Likewise.
5026 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5027 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5028 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5029 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5030 * config/i386/winnt.c: Include except.h.
5031 (struct seh_frame_state): Add reg_offset, after_prologue and
5032 in_cold_section fields.
5033 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5034 (i386_pe_seh_cold_init): New function.
5035 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5036 to seh->in_cold_section.
5037 (seh_emit_push): Record the offset of the push.
5038 (seh_emit_save): Record the offet of the save.
5039 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5040 Test seh->after_prologue to disregard the epilogue.
5041 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5042 (i386_pe_end_cold_function): New function.
5044 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5047 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5048 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5051 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5053 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5054 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5057 * multiple_target.c: Include tree-inline.h and intl.h.
5058 (expand_target_clones): Diagnose and fail if node->definition and
5059 !tree_versionable_function_p (node->decl).
5061 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5063 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5065 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5066 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5067 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5069 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5072 * combine.c (try_combine): Use reg_or_subregno instead of handling
5073 just paradoxical SUBREGs and REGs.
5075 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5077 * config/arc/arc.c (arc_finalize_pic): Remove function.
5078 (arc_must_save_register): We use single base PIC register, remove
5079 checks to save/restore the PIC register.
5080 (arc_expand_prologue): Likewise.
5081 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5083 (arc_verify_short): Likewise.
5084 (arc_attr_type): Likewise.
5085 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5086 (walk_stores): Likewise.
5087 (arc_address_cost): Make it static.
5088 (arc_verify_short): Likewise.
5089 (branch_dest): Likewise.
5090 (arc_attr_type): Likewise.
5091 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5092 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5093 (arc_final_prescan_insn): Remove inserting the nops due to
5094 hardware hazards. It is done in reorg step.
5095 (insn_length_variant_t): Remove.
5096 (insn_length_parameters_t): Likewise.
5097 (arc_insn_length_parameters): Likewise.
5098 (arc_get_insn_variants): Likewise.
5099 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5101 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5104 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5107 PR tree-optimization/84687
5108 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5110 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5112 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5114 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5115 Rename to ppc_speculation_barrier.
5116 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5117 __builtin_ppc_speculation_barrier.
5119 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5122 * combine.c (combine_simplify_rtx): Don't try to simplify if
5123 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5126 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5128 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5129 to 32 bytes when compiling for POWER9.
5131 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5134 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5135 regparm >= 3 with no arg reg available also for calls with
5136 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5139 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5141 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5143 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5146 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5148 2018-03-05 Richard Biener <rguenther@suse.de>
5150 PR tree-optimization/84486
5151 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5152 When inserting a __builtin_assume_aligned call set the LHS
5153 SSA name alignment info accordingly.
5155 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5157 PR tree-optimization/84114
5158 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5159 Avoid reassociation of FLOAT_MODE addition.
5161 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5163 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5164 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5165 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5166 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5167 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5168 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5169 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5171 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5172 __builtin_ia32_wbinvd): New builtins.
5173 (SPECIAL_ARGS2): New.
5174 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5175 (SPECIAL_ARGS2): New.
5176 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5177 (ix86_valid_target_attribute_inner_p): Ditto.
5178 (ix86_init_mmx_sse_builtins): Add special_args2.
5179 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5180 TARGET_WBNOINVD_P): New.
5181 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5182 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5183 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5184 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5185 * config/i386/pconfigintrin.h: New file.
5186 * config/i386/wbnoinvdintrin.h: Ditto.
5187 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5188 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5190 2018-03-05 Richard Biener <rguenther@suse.de>
5192 PR tree-optimization/84670
5193 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5195 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5196 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5197 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5198 doesn't grow if all successors (recursively) were visited at least
5201 2018-03-05 Richard Biener <rguenther@suse.de>
5203 PR tree-optimization/84650
5204 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5205 if executed in the loop pipeline.
5207 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5209 * doc/configfiles.texi (Configuration Files): Move info about
5210 conditionalizing $target-protos.h to...
5211 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5212 differs from $target-protos.h.
5214 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5215 Chung-Ju Wu <jasonwucj@gmail.com>
5217 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5218 * config/nds32/nds32-multiple.md (setmemsi): Define.
5219 * config/nds32/nds32-memory-manipulation.c
5220 (nds32_gen_dup_4_byte_to_word_value): New.
5221 (emit_setmem_word_loop): New.
5222 (emit_setmem_byte_loop): New.
5223 (nds32_expand_setmem_loop): New.
5224 (nds32_expand_setmem_loop_v3m): New.
5225 (nds32_expand_setmem_unroll): New.
5226 (nds32_expand_setmem): New.
5228 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5229 Chung-Ju Wu <jasonwucj@gmail.com>
5231 * config/nds32/nds32-memory-manipulation.c
5232 (nds32_emit_load_store): New.
5233 (nds32_emit_post_inc_load_store): New.
5234 (nds32_emit_mem_move): New.
5235 (nds32_emit_mem_move_block): New.
5236 (nds32_expand_movmemsi_loop_unknown_size): New.
5237 (nds32_expand_movmemsi_loop_known_size): New.
5238 (nds32_expand_movmemsi_loop): New.
5239 (nds32_expand_movmemsi_unroll): New.
5240 (nds32_expand_movmemqi): Rename ...
5241 (nds32_expand_movmemsi): ... to this.
5242 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5243 (movmemsi): ... to this.
5244 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5245 (nds32_expand_movmemsi): ... to this.
5247 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5248 Monk Chiang <sh.chiang04@gmail.com>
5249 Chung-Ju Wu <jasonwucj@gmail.com>
5251 * config/nds32/nds32-protos.h
5252 (nds32_expand_load_multiple): New arguments.
5253 (nds32_expand_store_multiple): Ditto.
5254 (nds32_valid_multiple_load_store): Rename ...
5255 (nds32_valid_multiple_load_store_p): ... to this.
5256 * config/nds32/nds32-memory-manipulation.c
5257 (nds32_expand_load_multiple): Refine implementation.
5258 (nds32_expand_store_multiple): Ditto.
5259 * config/nds32/nds32-multiple.md
5260 (load_multiple): Update nds32_expand_load_multiple interface.
5261 (store_multiple): Update nds32_expand_store_multiple interface.
5262 * config/nds32/nds32-predicates.c
5263 (nds32_valid_multiple_load_store): Rename ...
5264 (nds32_valid_multiple_load_store_p): ... to this and refine
5266 * config/nds32/predicates.md
5267 (nds32_load_multiple_and_update_address_operation): New predicate.
5268 (nds32_store_multiple_and_update_address_operation): New predicate.
5270 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5271 Chung-Ju Wu <jasonwucj@gmail.com>
5273 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5274 (combo): New attribute.
5275 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5277 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5279 * config/nds32/nds32.opt: Change -mcmodel= default value.
5281 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5282 Monk Chiang <sh.chiang04@gmail.com>
5283 Chung-Ju Wu <jasonwucj@gmail.com>
5285 * config/nds32/constants.md (unspec_element): New enum.
5286 * config/nds32/constraints.md (Umw): New constraint.
5287 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5288 * config/nds32/nds32-intrinsic.md: Likewise.
5289 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5290 (nds32_valid_smw_lwm_base_p): New.
5291 (nds32_output_smw_single_word): New.
5292 (nds32_output_lmw_single_word): New.
5293 (nds32_expand_unaligned_load): New.
5294 (nds32_expand_unaligned_store): New.
5295 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5296 (nds32_output_smw_single_word): Declare.
5297 (nds32_output_lmw_single_word): Declare.
5298 (nds32_expand_unaligned_load): Declare.
5299 (nds32_expand_unaligned_store): Declare.
5300 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5301 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5302 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5303 NDS32_BUILTIN_UASTORE_DW.
5304 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5307 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5308 Kito Cheng <kito.cheng@gmail.com>
5309 Chung-Ju Wu <jasonwucj@gmail.com>
5311 * config/nds32/nds32-intrinsic.c
5312 (nds32_expand_builtin_null_ftype_reg): Delete.
5313 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5314 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5315 (nds32_read_argument): New.
5316 (nds32_legitimize_target): Ditto.
5317 (nds32_legitimize_argument): Ditto.
5318 (nds32_check_constant_argument): Ditto.
5319 (nds32_expand_unop_builtin): Ditto.
5320 (nds32_expand_unopimm_builtin): Ditto.
5321 (nds32_expand_binop_builtin): Ditto.
5322 (nds32_builtin_decl_impl): Ditto.
5323 (builtin_description): Ditto.
5324 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5325 (nds32_init_builtins_impl): Ditto.
5326 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5327 (nds32_builtin_decl): New.
5328 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5329 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5331 2018-03-02 Jeff Law <law@redhat.com>
5333 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5334 (redundant_insn, fill_simple_delay_slots): Likewise.
5335 (fill_slots_from_thread): Likewise.
5336 * resource.c (mark_referenced_resources): Likewise.
5337 (mark_set_resources, find_dead_or_set_registers): Likewise.
5339 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5341 * substring-locations.h (format_warning_va): Formatting fix for
5343 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5345 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5346 * substring-locations.c: Include intl.h.
5347 (format_warning_va): Turned into small wrapper around
5348 format_warning_n_va, renamed to ...
5349 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5350 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5352 (format_warning_at_substring_n): New function.
5353 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5354 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5355 format_warning_at_substring with just a shorter name instead of
5356 const function pointer.
5357 (fmtwarn_n): New function.
5358 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5359 appropriate, get rid of all the fmtstr temporaries, move conditionals
5360 with G_() wrapped string literals directly into fmtwarn arguments,
5361 cast dir.len to (int), formatting fixes.
5363 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5365 * doc/invoke.texi: Remove "Cilk Plus" references.
5367 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5368 Richard Biener <rguenther@suse.de>
5371 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5372 for error or warning attributes if CALL_FROM_THUNK_P is set.
5375 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5378 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5379 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5382 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5383 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5385 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5386 instead of -1U in last predictors element's probability member.
5388 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5391 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5392 arguments if they are comparable.
5394 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5396 PR tree-optimization/84634
5397 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5398 masks and masked_loop_p with a single loop_masks, making sure it's
5399 null for bb vectorization.
5401 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5403 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5404 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5405 loop->force_vectorize to check whether there is no alias.
5407 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5410 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5412 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5414 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5415 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5416 instead of a loop around prev_real_insn.
5417 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5421 * config/i386/i386.c (ix86_print_operand): Use conditional
5422 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5425 2018-03-02 Richard Biener <rguenther@suse.de>
5427 PR tree-optimization/84427
5428 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5429 (bitmap_set_subtract_values): Rewrite to handle multiple
5432 (prune_clobbered_mems): Likewise.
5433 (phi_translate): Take edge instead of pred/phiblock.
5434 (phi_translate_1): Likewise.
5435 (phi_translate_set): Likewise. Insert all translated
5436 exprs for a value into the set, keeping possibly multiple
5437 expressions per value.
5438 (compute_antic_aux): Adjust for phi_translate changes.
5439 When intersecting union the expressions and prune those
5440 not in the final value set, keeping possibly multiple
5441 expressions per value. Do not use value-insertion
5442 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5443 all expressions. Add verification that the value-sets
5444 only shrink during iteration.
5445 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5446 (do_pre_regular_insertion): Likewise.
5447 (do_pre_partial_partial_insertion): Likewise.
5449 2018-03-02 Richard Biener <rguenther@suse.de>
5452 * config/darwin.c (saved_debug_info_level): New static global.
5453 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5454 (darwin_asm_lto_end): Restore debug info generation settings.
5456 2018-03-01 Martin Liska <mliska@suse.cz>
5459 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5462 2018-03-01 Richard Biener <rguenther@suse.de>
5465 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5466 type annotation with LTO when debug was disabled at compile-time.
5468 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5470 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5472 (mips_final_postscan_insn): Likewise.
5474 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5476 PR rtl-optimization/84528
5477 * alias.c (init_alias_target): Add commentary.
5478 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5479 a unique base value if the frame pointer is not eliminated
5480 to the stack pointer.
5482 2018-03-01 Tom de Vries <tom@codesourcery.com>
5484 PR rtl-optimization/83327
5485 * lra-int.h (hard_regs_spilled_into): Declare.
5486 * lra.c (hard_regs_spilled_into): Define.
5487 (init_reg_info): Init hard_regs_spilled_into.
5488 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5489 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5490 (process_bb_lives): Handle hard_regs_spilled_into.
5491 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5492 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5494 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5496 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5497 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5498 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5499 * config/rs6000/aix72.h: New file.
5501 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5503 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5504 instead of warning_at with conditional singular and plural messages
5508 * stor-layout.c (update_alignment_for_field): For
5509 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5510 && !DECL_PACKED (field), do the alignment update, just use
5511 only desired_align instead of MAX (type_align, desired_align)
5513 (place_field): Don't do known_align < desired_align handling
5514 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5515 is non-NULL, instead do it after rli->prev_field handling and
5516 only if not within a bitfield word. For DECL_PACKED (field)
5517 use type_align of BITS_PER_UNIT.
5519 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5521 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5522 superfluous parentheses and trailing spaces.
5524 2018-02-28 Richard Biener <rguenther@suse.de>
5526 PR tree-optimization/84584
5527 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5528 SCoPs with fake exit edge.
5530 2018-02-28 Martin Liska <mliska@suse.cz>
5533 * timevar.c (timer::print): Fix format to properly print 100%
5536 2018-02-28 Richard Biener <rguenther@suse.de>
5539 * genmatch.c (capture_info::walk_match): Do not mark
5540 captured expressions without operands as expr_p given
5541 they act more like predicates and should be subject to
5542 "lost tail" side-effect preserving.
5544 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5546 PR rtl-optimization/81611
5547 * auto-inc-dec.c (attempt_change): Move dead note from
5548 mem_insn if it's the next use of regno
5549 (find_address): Take address use of reg holding
5550 non-incremented value. Add parm to limit search to the named
5552 (merge_in_block): Attempt to use a mem insn that is the next
5553 use of the original regno.
5555 2018-02-27 Martin Sebor <msebor@redhat.com>
5558 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5559 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5561 2018-02-27 Martin Sebor <msebor@redhat.com>
5563 PR translation/84207
5564 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5565 n argument to unsigned HOST_WIDE_INT.
5566 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5567 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5568 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5569 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5571 2018-02-27 Richard Biener <rguenther@suse.de>
5573 PR tree-optimization/84512
5574 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5575 Do not use the estimate returned from record_stmt_cost for
5576 the scalar iteration cost but sum properly using add_stmt_cost.
5578 2018-02-27 Richard Biener <rguenther@suse.de>
5580 PR tree-optimization/84466
5581 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5582 Adjust last change to less strictly validate use operands.
5584 2018-02-27 Martin Liska <mliska@suse.cz>
5586 PR gcov-profile/84548
5587 * gcov.c (process_file): Allow partial overlap and consider it
5588 also as group functions.
5589 (output_lines): Properly calculate range of lines for a group.
5591 2018-02-27 Martin Liska <mliska@suse.cz>
5593 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5594 'ggc' suffixes. Change first column width.
5595 (timer::print): Fix formatting of the column.
5597 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5599 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5600 preserve inline entry blocks for the sake of debug inline
5601 entry point markers alone.
5602 (remove_unused_locals): Suggest in comments a better place to
5603 force the preservation of inline entry blocks that are
5604 otherwise unused, but do not preserve them.
5606 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5608 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5610 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5613 * config/i386/constraints.md (Bs): Replace
5614 ix86_indirect_branch_register with
5615 TARGET_INDIRECT_BRANCH_REGISTER.
5617 * config/i386/i386.md (indirect_jump): Likewise.
5618 (tablejump): Likewise.
5619 (*sibcall_memory): Likewise.
5620 (*sibcall_value_memory): Likewise.
5621 Peepholes of indirect call and jump via memory: Likewise.
5622 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5623 (*sibcall_value_GOT_32): Likewise.
5624 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5625 (GOT_memory_operand): Likewise.
5626 (call_insn_operand): Likewise.
5627 (sibcall_insn_operand): Likewise.
5628 (GOT32_symbol_operand): Likewise.
5629 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5631 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5633 PR rtl-optimization/83496
5634 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5635 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5636 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5637 redundant insn, if any.
5638 (relax_delay_slots): Likewise.
5639 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5641 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5643 PR tree-optimization/83965
5644 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5645 that grouped statements are part of a reduction chain. Return
5646 true if the statement is not marked as a reduction itself but
5648 (vect_recog_dot_prod_pattern): Don't check whether the statement
5649 is part of a group here.
5650 (vect_recog_sad_pattern): Likewise.
5651 (vect_recog_widen_sum_pattern): Likewise.
5653 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5656 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5658 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5661 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5663 (ix86_output_indirect_function_return): New prototype.
5664 (ix86_split_simple_return_pop_internal): Likewise.
5665 * config/i386/i386.c (indirect_return_via_cx): New.
5666 (indirect_return_via_cx_bnd): Likewise.
5667 (indirect_thunk_name): Handle return va CX_REG.
5668 (output_indirect_thunk_function): Create alias for
5669 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5670 (ix86_output_indirect_jmp): Remove the bool argument.
5671 (ix86_output_indirect_function_return): New function.
5672 (ix86_split_simple_return_pop_internal): Likewise.
5673 * config/i386/i386.md (*indirect_jump): Don't pass false
5674 to ix86_output_indirect_jmp.
5675 (*tablejump_1): Likewise.
5676 (simple_return_pop_internal): Change it to define_insn_and_split.
5677 Call ix86_split_simple_return_pop_internal to split it for
5679 (simple_return_indirect_internal): Call
5680 ix86_output_indirect_function_return instead of
5681 ix86_output_indirect_jmp.
5683 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5686 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5687 memset and value initialization afterwards.
5689 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5691 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5693 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5696 * common/config/aarch64/aarch64-common.c
5697 (aarch_option_optimization_table[]): Switch
5698 off fomit-frame-pointer
5700 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5701 Chung-Ju Wu <jasonwucj@gmail.com>
5703 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5705 (store_multiple): Ditto.
5707 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5709 * config.gcc: Add --with-cpu support for nds32 target.
5710 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5711 * config/nds32/nds32.opt: Add -mcpu= option.
5713 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5715 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5716 isel=yes): Warn for these deprecated options.
5718 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5720 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5721 ISA_2_5_MASKS_EMBEDDED.
5723 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5725 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5726 p->max as pointers rather than using iterative_hash_expr.
5728 2018-02-23 Carl Love <cel@us.ibm.com>
5730 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5731 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5733 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5734 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5735 P8V_BUILTIN_VEC_VUNSIGNED2.
5737 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5740 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5741 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5743 * lra-constraints.c (curr_insn_transform): Set up
5744 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5746 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5747 * lra-eliminations.c (spill_pseudos): Ditto.
5748 (process_insn_for_elimination): Ditto.
5749 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5750 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5751 LRA_NON_CLOBBERED_ALT.
5753 2018-02-22 Martin Sebor <msebor@redhat.com>
5755 PR tree-optimization/84480
5756 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5757 to maybe_diag_stxncpy_trunc. Call it.
5758 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5759 from gimple_fold_builtin_strcpy. Print inlining stack.
5760 (handle_builtin_stxncpy): Print inlining stack.
5761 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5763 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5766 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5767 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5768 and -fcheck-pointer-bounds are used together.
5769 (indirect_thunk_prefix): New enum.
5770 (indirect_thunk_need_prefix): New function.
5771 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5772 "_nt" instead of "_bnd" for NOTRACK prefix.
5773 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5774 (output_indirect_thunk_function): Likewise.
5776 (ix86_code_end): Update output_indirect_thunk_function calls.
5777 (ix86_output_indirect_branch_via_reg): Replace
5778 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5779 (ix86_output_indirect_branch_via_push): Likewise.
5780 (ix86_output_function_return): Likewise.
5781 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5782 incompatible with -fcf-protection=branch and
5783 -fcheck-pointer-bounds.
5785 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5788 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5789 Change gcc_assert call to output_operand_lossage.
5791 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5793 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5795 2018-02-22 DJ Delorie <dj@redhat.com>
5796 Sebastian Perta <sebastian.perta@renesas.com>
5797 Oleg Endo <olegendo@gcc.gnu.org>
5799 * config/rx/rx.c (rx_rtx_costs): New function.
5800 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5802 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5804 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5806 2018-02-22 Martin Liska <mliska@suse.cz>
5809 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5810 Add "native" as a possible value.
5812 2018-02-22 Martin Liska <mliska@suse.cz>
5815 * config/i386/i386.c (ix86_option_override_internal):
5816 Add "native" as a possible value for -march and -mtune.
5818 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5821 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5822 to all type variants.
5824 PR tree-optimization/84503
5825 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5826 width as info->bitpos + info->bitsize - start.
5827 (merged_store_group::merge_overlapping): Simplify width computation.
5828 (check_no_overlap): New function.
5829 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5830 start + width and last_order of the group, fail if check_no_overlap
5832 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5833 to group if check_no_overlap fails.
5835 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5837 * config/rs6000/altivec.md: Delete contraint arguments to
5838 define_expand, define_split, and define_peephole2, and in
5839 define_insn_and_split if always unused.
5840 * config/rs6000/darwin.md: Ditto.
5841 * config/rs6000/dfp.md: Ditto.
5842 * config/rs6000/rs6000.md: Ditto.
5843 * config/rs6000/sync.md: Ditto.
5844 * config/rs6000/vector.md: Ditto.
5845 * config/rs6000/vsx.md: Ditto.
5847 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5849 * config/rs6000/altivec.md: Write output control strings as braced
5850 blocks instead of double-quoted strings.
5851 * config/rs6000/darwin.md: Ditto.
5852 * config/rs6000/rs6000.md: Ditto.
5853 * config/rs6000/vector.md: Ditto.
5854 * config/rs6000/vsx.md: Ditto.
5856 2018-02-21 Jason Merrill <jason@redhat.com>
5858 PR c++/84314 - ICE with templates and fastcall attribute.
5859 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5861 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5863 * ipa-cp.c (determine_versionability): Fix comment typos.
5865 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5868 * ipa-cp.c (determine_versionability): Do not version functions caling
5871 2018-02-21 Martin Liska <mliska@suse.cz>
5874 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5875 Add "native" as a possible value.
5876 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5877 the macro when native cpu detection is available.
5879 2018-02-21 Martin Liska <mliska@suse.cz>
5882 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5883 Add "native" as a possible value.
5884 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5885 when native cpu detection is available.
5887 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5888 Martin Sebor <msebor@redhat.com>
5890 PR tree-optimization/84478
5891 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5893 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5894 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5895 support which is conservatively correct, for 2 only stay conservative
5896 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5897 argument to the 2 argument get_range_strlen, adjust 6 arg
5898 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5900 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5901 (gimple_fold_builtin_strlen): Pass true as last argument to
5904 2018-02-20 Martin Sebor <msebor@redhat.com>
5907 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5908 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5909 (builtin_memref::builtin_memref): Factor out parts into
5910 set_base_and_offset and call it.
5912 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5915 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5916 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5917 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5918 search at the associated MODE_INT.
5920 2018-02-20 Jeff Law <law@redhat.com>
5923 PR tree-optimization/81592
5925 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5926 for range data rather than using global data.
5927 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5928 range data rather than using global data.
5929 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5930 pass it to children as needed.
5931 (struct directive::fmtresult): Similarly.
5932 (struct directive::set_width): Similarly.
5933 (struct directive::set_precision): Similarly.
5934 (format_integer, format_directive, parse_directive): Similarly.
5935 (format_none): Accept unnamed vr_values parameter.
5936 (format_percent, format_floating, format_character): Similarly.
5937 (format_string, format_plain): Similarly.
5938 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5939 the EVRP range analyzer for range data rather than using global data.
5940 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5941 gimple-ssa-evrp-analyze.h
5942 (class sprintf_dom_walker): Add after_dom_children member function.
5943 Add evrp_range_analyzer member.
5944 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5945 range analyzer as needed.
5946 (sprintf_dom_walker::after_dom_children): New member function.
5947 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5949 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5950 (evrp_range_analyzer::pop_to_marker): Likewise.
5952 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5954 PR tree-optimization/84419
5955 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5956 with the required type if its current type is compatible but
5959 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5962 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5963 after vectorization.
5965 2018-02-20 Martin Liska <mliska@suse.cz>
5968 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5969 possible values if we don't have a hint.
5971 2018-02-20 Martin Liska <mliska@suse.cz>
5975 * final.c (shorten_branches): Build align_tab array with one
5977 * opts.c (finish_options): Add alignment option limit check.
5978 (MAX_CODE_ALIGN): Likewise.
5979 (MAX_CODE_ALIGN_VALUE): Likewise.
5980 * doc/invoke.texi: Document maximum allowed option value for
5981 all -falign-* options.
5983 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5986 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5987 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5988 * var-tracking.c (emit_note_insn_var_location): Remove all references
5989 to NOTE_INSN_CALL_ARG_LOCATION.
5990 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5991 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5992 Use copy_rtx_if_shared.
5993 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5994 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5995 (dwarf2out_var_location): Remove handling of
5996 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5998 * final.c (final_scan_insn): Remove all references to
5999 NOTE_INSN_CALL_ARG_LOCATION.
6000 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6001 before dumping final insns.
6002 * except.c (emit_note_eh_region_end): Remove all references to
6003 NOTE_INSN_CALL_ARG_LOCATION.
6004 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6005 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6006 * config/arc/arc.c (hwloop_optimize): Likewise.
6007 * config/arm/arm.c (create_fix_barrier): Likewise.
6008 * config/s390/s390.c (s390_chunkify_start): Likewise.
6009 * config/sh/sh.c (find_barrier): Likewise.
6010 * config/i386/i386.c (rest_of_insert_endbranch,
6011 ix86_seh_fixup_eh_fallthru): Likewise.
6012 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6013 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6014 * config/frv/frv.c (frv_function_prologue): Likewise.
6015 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6017 (note_outside_basic_block_p): Remove all references to
6018 NOTE_INSN_CALL_ARG_LOCATION.
6019 * gengtype.c (adjust_field_rtx_def): Likewise.
6020 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6022 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6023 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6026 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6029 PR tree-optimization/84452
6030 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6031 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6034 2018-02-19 Martin Liska <mliska@suse.cz>
6037 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6039 2018-02-19 Martin Liska <mliska@suse.cz>
6040 Richard Sandiford <richard.sandiford@linaro.org>
6042 PR tree-optimization/82491
6043 * gimple-fold.c (get_base_constructor): Make earlier bail out
6046 2018-02-19 Carl Love <cel@us.ibm.com>
6048 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6049 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6051 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6052 P8V_BUILTIN_VEC_NEG.
6054 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6056 * config/rl78/rl78.md (movdf): New define expand.
6058 2018-02-19 Martin Liska <mliska@suse.cz>
6061 * doc/invoke.texi: Fix typo.
6062 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6064 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6066 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6067 handle rs6000_single_float and rs6000_double_float specially for
6070 2018-02-16 Jeff Law <law@redhat.com>
6072 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6075 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6078 * ipa-inline.c (inline_small_functions): Fix a typo.
6080 2018-02-16 Nathan Sidwell <nathan@acm.org>
6082 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6084 2018-02-16 Carl Love <cel@us.ibm.com>
6086 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6087 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6088 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6089 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6090 expansion to P8V_BUILTIN_VEC_FLOAT2.
6092 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6094 PR rtl-optimization/70023
6095 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6096 src_regno into account.
6098 2018-02-16 Carl Love <cel@us.ibm.com>
6100 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6101 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6102 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6103 * config/rs6000/rs6000.c: Remove case statements for
6104 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6105 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6106 and P9V_BUILTIN_VEC_VINSERT4B.
6107 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6108 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6109 * config/rs6000/vsx.md:
6110 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6113 2018-02-16 Carl Love <cel@us.ibm.com>
6115 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6117 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6119 * config/rs6000/rs6000-c.c: Add the definitions for
6120 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6121 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6122 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6123 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6124 definition for insert4b and define insn *insert3b_internal.
6125 * doc/extend.texi: Add documentation for vec_extract4b.
6127 2018-02-16 Nathan Sidwell <nathan@acm.org>
6129 * doc/extend.texi (Backwards Compatibility): Mention friend
6130 injection. Note for-scope is deprecated.
6131 * doc/invoke.texi (-ffriend-injection): Deprecate.
6133 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6135 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6136 that moved to I2, also allow destinations that are a paradoxical
6137 subreg (instead of a normal reg).
6139 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6142 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6145 2018-02-16 Richard Biener <rguenther@suse.de>
6147 PR tree-optimization/84037
6148 PR tree-optimization/84016
6150 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6151 Adjust vec_construct for the fact we need additional higher latency
6152 128bit inserts for AVX256 and AVX512 vector builds.
6153 (ix86_add_stmt_cost): Scale vector construction cost for
6156 2018-02-16 Richard Biener <rguenther@suse.de>
6158 PR tree-optimization/84417
6159 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6160 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6161 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6163 2018-02-16 Martin Liska <mliska@suse.cz>
6166 * internal-fn.def (ASAN_CHECK): Set proper flags.
6167 (ASAN_MARK): Likewise.
6169 2018-02-16 Julia Koval <julia.koval@intel.com>
6171 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6172 from PTA_CANNONLAKE.
6174 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6177 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6178 Use ++iter rather than iter++ for std::list iterators.
6179 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6180 defer deleting them until all nodes in the forest are processed. Do
6181 free even leaf nodes. Change to_process into auto_vec.
6184 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6185 * vec.h (vec_default_construct): Use memset instead of placement new
6186 if BROKEN_VALUE_INITIALIZATION is defined.
6187 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6188 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6191 PR rtl-optimization/83723
6192 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6193 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6194 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6196 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6198 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6200 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6202 PR rtl-optimization/81443
6203 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6204 from inner REGs to paradoxical SUBREGs.
6206 2018-02-16 Richard Biener <rguenther@suse.de>
6208 PR tree-optimization/84399
6209 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6210 For operands we can analyze at their definition make sure we can
6211 analyze them at each use as well.
6213 2018-02-16 Richard Biener <rguenther@suse.de>
6215 PR tree-optimization/84190
6216 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6217 volatile accesses if the decl isn't volatile.
6219 2018-02-15 Jason Merrill <jason@redhat.com>
6221 PR c++/84314 - ICE with templates and fastcall attribute.
6222 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6223 TYPE_CANONICAL on an existing type.
6225 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6227 PR tree-optimization/84383
6228 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6229 dstoff nor call operand_equal_p if dstbase is NULL.
6231 PR tree-optimization/84334
6232 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6233 also a CONSTANT_CLASS_P, punt.
6235 2018-02-14 Jim Wilson <jimw@sifive.com>
6237 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6238 first SMALL_OPERAND check. New local min_second_step. Move assert
6239 to where locals are set. Add TARGET_RVC support.
6240 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6242 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6244 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6246 2018-02-14 Martin Sebor <msebor@redhat.com>
6248 PR tree-optimization/83698
6249 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6250 arrays constrain the offset range to their bounds.
6251 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6252 (builtin_access::overlap): Avoid setting the size of overlap if it's
6254 (maybe_diag_overlap): Also consider arrays when deciding what values
6255 of offsets to include in diagnostics.
6257 2018-02-14 Martin Sebor <msebor@redhat.com>
6260 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6261 that correspond to the kind of a declaration.
6263 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6266 * config/pa/pa.md: Load address of PIC label using the linkage table
6267 if the label is nonlocal.
6269 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6271 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6272 warning message if user requests -maltivec=be.
6273 * doc/invoke.texi: Document deprecation of -maltivec=be.
6275 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6278 * config/rs6000/rs6000-c.c: Update definitions for
6279 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6280 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6282 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6285 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6286 add _get_ssp intrinsics. Remove argument from
6287 __builtin_ia32_rdssp[d|q].
6288 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6289 * config/i386/i386-builtin.def: Remove argument from
6290 __builtin_ia32_rdssp[d|q].
6291 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6292 ix86_expand_special_args_builtin for _rdssp[d|q].
6293 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6294 Clear register before usage.
6295 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6296 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6298 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6300 PR tree-optimization/84357
6301 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6302 operand 1 of an ARRAY_REF too.
6304 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6307 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6308 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6310 (set_of_reg): New struct.
6311 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6312 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6313 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6315 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6316 Split into bitclr, bitset, bitinvert patterns if appropriate.
6317 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6318 use rx_fuse_in_memory_bitop.
6319 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6320 to named insn, correct maximum insn length.
6322 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6325 * machmode.def: Define a complex mode for PARTIAL_INT.
6326 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6328 * doc/rtl.texi: Document CSPImode.
6329 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6331 (msp430_hard_regno_nregs_with_padding): Likewise.
6333 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6336 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6338 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6340 PR rtl-optimization/84169
6341 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6342 we generated a parallel as new i3 and we split that to new i2 and i3
6343 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6344 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6345 those to i2, not i1. Partially rewrite this scan code.
6347 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6350 * stor-layout.c (place_field): For variable length fields, adjust
6351 offset_align afterwards not just based on the field's alignment,
6352 but also on the size.
6355 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6356 of exps and logs in the use_exp2 case.
6358 2018-02-13 Jeff Law <law@redhat.com>
6360 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6363 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6366 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6370 * common.opt (gas-loc-support, gas-locview-support): New.
6371 (ginline-points, ginternal-reset-location-views): New.
6372 * doc/invoke.texi: Document them. Use @itemx where intended.
6373 (gvariable-location-views): Adjust.
6374 * target.def (reset_location_view): New.
6375 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6376 (TARGET_RESET_LOCATION_VIEW): New.
6377 * doc/tm.texi: Rebuilt.
6378 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6379 (dwarf2out_default_as_locview_support): New.
6380 (output_asm_line_debug_info): Use option variables.
6381 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6382 (output_loc_list): Likewise.
6383 (add_high_low_attributes): Check option variables.
6384 Don't output entry view attribute in strict mode.
6385 (gen_inlined_subroutine_die): Check option variables.
6386 (dwarf2out_inline_entry): Likewise.
6387 (init_sections_and_labels): Likewise.
6388 (dwarf2out_early_finish): Likewise.
6389 (maybe_reset_location_view): New, from...
6390 (dwarf2out_var_location): ... here. Call it.
6391 * debug.h (dwarf2out_default_as_loc_support): Declare.
6392 (dwarf2out_default_as_locview_support): Declare.
6393 * hooks.c (hook_int_rtx_insn_0): New.
6394 * hooks.h (hook_int_rtx_insn_0): Declare.
6395 * toplev.c (process_options): Take -gas-loc-support and
6396 -gas-locview-support from dwarf2out. Enable
6397 -gvariable-location-views by default only with locview
6398 assembler support. Enable -ginternal-reset-location-views by
6399 default only if the target defines the corresponding hook.
6400 Enable -ginline-points by default if location views are
6401 enabled; force it disabled if statement frontiers are
6403 * tree-inline.c (expand_call_inline): Check option variables.
6404 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6406 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6408 PR tree-optimization/84321
6409 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6410 handling. Also check whether the anti-range contains any values
6411 that satisfy the mask; switch to a VR_RANGE if not.
6413 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6416 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6418 2018-02-13 Martin Jambor <mjambor@suse.cz>
6421 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6422 of call statements, also set location of a load to a temporary.
6424 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6426 * config/rl78/rl78.c (add_vector_labels): New function.
6427 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6428 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6429 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6430 which checks that no arguments are passed.
6431 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6432 * doc/extend.texi: Documentation for the new attribute.
6434 2018-02-13 Andreas Schwab <schwab@suse.de>
6436 * config/riscv/linux.h (CPP_SPEC): Define.
6438 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6441 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6442 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6443 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6444 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6445 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6446 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6447 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6448 temporarily for AES and PCLMUL builtins.
6450 PR tree-optimization/84339
6451 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6452 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6456 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6457 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6458 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6460 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6462 * omp-simd-clone.h: New file.
6463 * omp-simd-clone.c: Include omp-simd-clone.h.
6464 (expand_simd_clones): No longer static.
6465 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6466 cgraph.h and omp-simd-clone.h.
6467 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6468 (vect_recog_widen_shift_pattern): Formatting fix.
6469 (vect_pattern_recog_1): Don't check optab for calls.
6472 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6473 operands[2] into a REG before using gen_lowpart on it.
6475 2018-02-12 Jeff Law <law@redhat.com>
6478 * config/sh/sh.c (find_barrier): Consider a sibling call
6481 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6482 successfully back substituting a reg.
6484 2018-02-12 Richard Biener <rguenther@suse.de>
6486 PR tree-optimization/84037
6487 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6488 parameter, move visited init to caller.
6489 (vect_slp_analyze_operations): Separate cost from validity
6490 check, initialize visited once for all instances.
6491 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6493 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6495 (vect_model_promotion_demotion_cost): Likewise.
6496 (vectorizable_bswap): Guard cost modeling with !slp_node
6497 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6499 (vectorizable_call): Likewise.
6500 (vectorizable_conversion): Likewise.
6501 (vectorizable_assignment): Likewise.
6502 (vectorizable_shift): Likewise.
6503 (vectorizable_operation): Likewise.
6504 (vectorizable_store): Likewise.
6505 (vectorizable_load): Likewise.
6506 (vectorizable_condition): Likewise.
6507 (vectorizable_comparison): Likewise.
6509 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6512 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6513 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6514 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6516 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6519 * params.def (inline-min-speedup): Increase from 8 to 15.
6520 (max-inline-insns-auto): Decrease from 40 to 30.
6521 * ipa-split.c (consider_split): Add some buffer for function to
6522 be considered inlining candidate.
6523 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6526 2018-02-12 Richard Biener <rguenther@suse.de>
6528 PR tree-optimization/84037
6529 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6530 matched stmts if we cannot swap the non-matched ones.
6532 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6534 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6535 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6536 _mm_maskz_scalef_round_ss): New intrinsics.
6537 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6538 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6539 __builtin_ia32_scalefss_round): Remove.
6540 (__builtin_ia32_scalefsd_mask_round,
6541 __builtin_ia32_scalefss_mask_round): New intrinsics.
6542 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6543 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6544 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6545 "<round_constraint>")): Changed to ...
6546 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6547 "<round_scalar_constraint>")): ... this.
6548 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6549 %0, %1, %2<round_op3>}"): Changed to ...
6550 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6551 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6552 %2<round_scalar_mask_op3>}"): ... this.
6553 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6555 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6557 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6558 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6559 (_mm_maskz_sqrt_round_ss): New intrinsics.
6560 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6561 (__builtin_ia32_sqrtsd_mask_round)
6562 (__builtin_ia32_sqrtss_mask_round): New builtins.
6563 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6564 (__builtin_ia32_sqrtss_round): Remove.
6565 (__builtin_ia32_sqrtsd_mask_round)
6566 (__builtin_ia32_sqrtss_mask_round): New builtins.
6567 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6568 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6569 ((match_operand:VF_128 1 "vector_operand"
6570 "xBm,<round_constraint>")): Changed to ...
6571 ((match_operand:VF_128 1 "vector_operand"
6572 "xBm,<round_scalar_constraint>")): ... this.
6573 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6574 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6575 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6576 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6577 %<iptr>1<round_scalar_mask_op3>}): ... this.
6578 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6579 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6581 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6584 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6585 Cast vec_cmpeq result to correct type.
6586 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6587 Cast vec_cmpgt result to correct type.
6589 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6591 * final.c (final_scan_insn_1): Renamed from...
6592 (final_scan_insn): ... this. New wrapper, to recover
6593 seen from the outermost call in recursive ones.
6594 * config/sparc/sparc.c (output_return): Drop seen from call.
6595 (output_sibcall): Likewise.
6596 * config/visium/visium.c (output_branch): Likewise.
6598 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6600 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6603 2018-02-10 Alan Modra <amodra@gmail.com>
6606 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6607 Specify LR as an input.
6609 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6612 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6613 remove_member_access_dummy_vars): New functions.
6614 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6615 lower_omp_1, execute_lower_omp): Use them.
6617 PR rtl-optimization/84308
6618 * shrink-wrap.c (spread_components): Release todo vector.
6620 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6622 PR rtl-optimization/57193
6623 * ira-color.c (struct allocno_color_data): Add member
6624 conflict_allocno_hard_prefs.
6625 (update_conflict_allocno_hard_prefs): New.
6626 (bucket_allocno_compare_func): Add a preference based on
6627 conflict_allocno_hard_prefs.
6628 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6629 (color_allocnos): Remove a dead code. Initiate
6630 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6632 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6635 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6636 constraint from =wa to wa. Avoid a subreg on the output operand,
6637 instead use a pseudo and subreg it in a move.
6638 (p9_xxbrd_<mode>): Changed to ...
6639 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6640 (p9_xxbrd_v2df): New expander.
6641 (p9_xxbrw_<mode>): Changed to ...
6642 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6643 (p9_xxbrw_v4sf): New expander.
6645 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6647 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6649 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6652 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6653 multiply in 32-bit mode.
6654 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6655 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6658 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6660 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6661 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6662 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6663 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6665 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6668 * dwarf2out.c (is_trivial_indirect_ref): New function.
6669 (dwarf2out_late_global_decl): Do not generate a location
6670 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6671 and that are not defined in the current unit.
6673 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6675 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6676 instead of a libcall for UNORDERED.
6678 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6681 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6682 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6684 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6687 * config/s390/s390.c (s390_set_current_function): Invoke
6688 s390_indirect_branch_settings also if fndecl didn't change.
6690 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6692 * config/rs6000/rs6000.md (blockage): Set length to zero.
6694 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6696 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6698 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6701 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6702 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6706 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6707 PARALLEL incoming that failed vt_get_decl_and_offset check.
6710 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6711 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6713 (get_variable_section): For decls in named .bss* sections pass true as
6714 second argument to bss_initializer_p.
6716 2018-02-09 Marek Polacek <polacek@redhat.com>
6717 Jakub Jelinek <jakub@redhat.com>
6720 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6721 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6722 Sync some changes from cxx_fold_indirect_ref.
6724 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6726 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6728 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6729 (BLOCK_INLINE_ENTRY_LABEL): New.
6730 (dwarf2out_var_location): Disregard inline entry markers.
6731 (inline_entry_data): New struct.
6732 (inline_entry_data_hasher): New hashtable type.
6733 (inline_entry_data_hasher::hash): New.
6734 (inline_entry_data_hasher::equal): New.
6735 (inline_entry_data_table): New variable.
6736 (add_high_low_attributes): Add DW_AT_entry_pc and
6737 DW_AT_GNU_entry_view attributes if a pending entry is found
6738 in inline_entry_data_table. Add old entry_pc attribute only
6739 if debug nonbinding markers are disabled.
6740 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6741 markers are enabled.
6742 (block_within_block_p, dwarf2out_inline_entry): New.
6743 (dwarf2out_finish): Check that no entries remained in
6744 inline_entry_data_table.
6745 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6746 (final_scan_insn, notice_source_line): Likewise.
6747 (rest_of_clean_state): Skip inline entry markers.
6748 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6750 * gimple.c (gimple_build_debug_inline_entry): New.
6751 * gimple.h (enum gimple_debug_subcode): Add
6752 GIMPLE_DEBUG_INLINE_ENTRY.
6753 (gimple_build_debug_inline_entry): Declare.
6754 (gimple_debug_inline_entry_p): New.
6755 (gimple_debug_nonbind_marker_p): Adjust.
6756 * insn-notes.def (INLINE_ENTRY): New.
6757 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6758 inline entry marker notes.
6759 (print_insn): Likewise.
6760 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6761 (INSN_DEBUG_MARKER_KIND): Likewise.
6762 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6763 * tree-inline.c (expand_call_inline): Build and insert
6764 debug_inline_entry stmt.
6765 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6766 inline entry blocks early, if nonbind markers are enabled.
6767 (dump_scope_block): Dump fragment info.
6768 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6769 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6770 (gimple_build_debug_inline_entry): New.
6771 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6772 Enable/disable inline entry points too.
6773 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6774 (DEBUG_INSN): Describe inline entry markers.
6776 * common.opt (gvariable-location-views): New.
6777 (gvariable-location-views=incompat5): New.
6778 * config.in: Rebuilt.
6779 * configure: Rebuilt.
6780 * configure.ac: Test assembler for view support.
6781 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6782 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6783 * dwarf2out.c (var_loc_view): New typedef.
6784 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6785 (dwarf2out_locviews_in_attribute): New.
6786 (dwarf2out_locviews_in_loclist): New.
6787 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6788 (enum dw_line_info_opcode): Add LI_adv_address.
6789 (struct dw_line_info_table): Add view.
6790 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6791 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6792 (zero_view_p): New variable.
6793 (ZERO_VIEW_P): New macro.
6794 (output_asm_line_debug_info): New.
6795 (struct var_loc_node): Add view.
6796 (add_AT_view_list, AT_loc_list): New.
6797 (add_var_loc_to_decl): Add view param. Test it against last.
6798 (new_loc_list): Add view params. Record them.
6799 (AT_loc_list_ptr): Handle loc and view lists.
6800 (view_list_to_loc_list_val_node): New.
6801 (print_dw_val): Handle dw_val_class_view_list.
6802 (size_of_die): Likewise.
6803 (value_format): Likewise.
6804 (loc_list_has_views): New.
6805 (gen_llsym): Set vl_symbol too.
6806 (maybe_gen_llsym, skip_loc_list_entry): New.
6807 (dwarf2out_maybe_output_loclist_view_pair): New.
6808 (output_loc_list): Output view list or entries too.
6809 (output_view_list_offset): New.
6810 (output_die): Handle dw_val_class_view_list.
6811 (output_dwarf_version): New.
6812 (output_compilation_unit_header): Use it.
6813 (output_skeleton_debug_sections): Likewise.
6814 (output_rnglists, output_line_info): Likewise.
6815 (output_pubnames, output_aranges): Update version comments.
6816 (output_one_line_info_table): Output view numbers in asm comments.
6817 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6818 Call maybe_gen_llsym.
6819 (loc_list_from_tree_1): Adjust.
6820 (add_AT_location_description): Create view list attribute if
6821 needed, check it's absent otherwise.
6822 (convert_cfa_to_fb_loc_list): Adjust.
6823 (maybe_emit_file): Call output_asm_line_debug_info for test.
6824 (dwarf2out_var_location): Reset views as needed. Precompute
6825 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6826 attribute. Set view.
6827 (new_line_info_table): Reset next view.
6828 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6829 (dwarf2out_source_line): Likewise. Output view resets and labels to
6830 the assembler, or select appropriate line info opcodes.
6831 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6832 (optimize_string_length): Catch it. Adjust.
6833 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6834 dw_val_class_view_list, and remove it if no longer needed.
6835 (hash_loc_list): Hash view numbers.
6836 (loc_list_hasher::equal): Compare them.
6837 (optimize_location_lists): Check whether a view list symbol is
6838 needed, and whether the locview attribute is present, and
6839 whether they match. Remove the locview attribute if no longer
6841 (index_location_lists): Call skip_loc_list_entry for test.
6842 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6843 Use output_dwarf_version.
6844 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6845 (struct dw_val_node): Add val_view_list.
6846 * final.c (SEEN_NEXT_VIEW): New.
6847 (set_next_view_needed): New.
6848 (clear_next_view_needed): New.
6849 (maybe_output_next_view): New.
6850 (final_start_function): Rename to...
6851 (final_start_function_1): ... this. Take pointer to FIRST,
6852 add SEEN parameter. Emit param bindings in the initial view.
6853 (final_start_function): Reintroduce SEEN-less interface.
6854 (final): Rename to...
6855 (final_1): ... this. Take SEEN parameter. Output final pending
6856 next view at the end.
6857 (final): Reintroduce seen-less interface.
6858 (final_scan_insn): Output pending next view before switching
6859 sections or ending a block. Mark the next view as needed when
6860 outputting variable locations. Notify debug backend of section
6861 changes, and of location view changes.
6862 (rest_of_handle_final): Adjust.
6863 * toplev.c (process_options): Autodetect value for debug variable
6864 location views option. Warn on incompat5 without -gdwarf-5.
6865 * doc/invoke.texi (gvariable-location-views): New.
6866 (gvariable-location-views=incompat5): New.
6867 (gno-variable-location-views): New.
6869 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6871 PR tree-optimization/84136
6872 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6873 that the result of find_edge is non-NULL.
6875 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6878 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6879 storing integer register in SImode. Fix cost of 256 and 512
6880 byte aligned SSE register store.
6882 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6884 * config/i386/i386.c (ix86_multiplication_cost): Fix
6885 multiplication cost for TARGET_AVX512DQ.
6887 2018-02-08 Marek Polacek <polacek@redhat.com>
6889 PR tree-optimization/84238
6890 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6893 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6895 PR tree-optimization/84265
6896 * tree-vect-stmts.c (vectorizable_store): Don't treat
6897 VMAT_CONTIGUOUS accesses as grouped.
6898 (vectorizable_load): Likewise.
6900 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6902 PR tree-optimization/81635
6903 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6904 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6905 (test_round_for_mask): New functions.
6906 (wide_int_cc_tests): Call test_round_for_mask.
6907 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6908 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6909 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6910 range returned by get_range_info.
6912 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6915 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6916 * symtab.c: Include builtins.h
6917 (symtab_node::output_to_lto_symbol_table_p): Move here
6918 from lto-streamer-out.c:output_symbol_p.
6919 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6920 (output_symbol_p): Move all logic to symtab.c
6921 (produce_symtab): Update.
6923 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6925 * config/s390/s390-opts.h (enum indirect_branch): Define.
6926 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6927 (s390_indirect_branch_via_thunk)
6928 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6929 (enum s390_indirect_branch_type): Define.
6930 * config/s390/s390.c (struct s390_frame_layout, struct
6931 machine_function): Remove.
6932 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6933 (indirect_branch_table_label_no, indirect_branch_table_name):
6935 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6936 (enum s390_indirect_branch_option): Define.
6937 (s390_return_addr_from_memory): New function.
6938 (s390_handle_string_attribute): New function.
6939 (s390_attribute_table): Add new attribute handler.
6940 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6941 (s390_indirect_branch_via_thunk): New function.
6942 (s390_indirect_branch_via_inline_thunk): New function.
6943 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6944 sibling call optimization for non z10 compiles.
6945 (s390_emit_call): Force indirect branch target to be a single
6946 register. Add r1 clobber for non-z10 compiles.
6947 (s390_emit_epilogue): Emit return jump via return_use expander.
6948 (s390_reorg): Handle JUMP_INSNs as execute targets.
6949 (s390_option_override_internal): Perform validity checks for the
6950 new command line options.
6951 (s390_indirect_branch_attrvalue): New function.
6952 (s390_indirect_branch_settings): New function.
6953 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6954 (s390_output_indirect_thunk_function): New function.
6955 (s390_code_end): Implement target hook.
6956 (s390_case_values_threshold): Implement target hook.
6957 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6959 * config/s390/s390.h (struct s390_frame_layout)
6960 (struct machine_function): Move here from s390.c.
6961 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6962 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6963 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6964 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6965 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6966 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6967 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6968 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6969 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6970 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6971 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6972 (mnemonic attribute): Add values which aren't recognized
6974 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6975 pattern for branch conversion. Fix mnemonic attribute.
6976 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6977 indirect branch via thunk if requested.
6978 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6979 ("*indirect_jump"): Disable for branch conversion using out of
6981 ("indirect_jump_via_thunk<mode>_z10")
6982 ("indirect_jump_via_thunk<mode>")
6983 ("indirect_jump_via_inlinethunk<mode>_z10")
6984 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6985 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6986 ("casesi_jump_via_inlinethunk<mode>_z10")
6987 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6988 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6989 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6990 ("*indirect2_jump"): Disable for branch conversion.
6991 ("casesi_jump"): Turn into expander and expand patterns for branch
6993 ("return_use"): New expander.
6994 ("*return"): Emit return via thunk and rename it to ...
6995 ("*return<mode>"): ... this one.
6996 * config/s390/s390.opt: Add new options and and enum for the
6999 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7001 * lra-constraints.c (match_reload): Unconditionally use
7002 gen_lowpart_SUBREG, rather than selecting between that
7003 and equivalent gen_rtx_SUBREG code.
7005 2018-02-08 Richard Biener <rguenther@suse.de>
7007 PR tree-optimization/84233
7008 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7009 changed flag instead of boguously re-using phi_inserted.
7011 2018-02-08 Martin Jambor <mjambor@suse.cz>
7013 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7014 static local variables.
7016 2018-02-08 Richard Biener <rguenther@suse.de>
7018 PR tree-optimization/84278
7019 * tree-vect-stmts.c (vectorizable_store): When looking for
7020 smaller vector types to perform grouped strided loads/stores
7021 make sure the mode is supported by the target.
7022 (vectorizable_load): Likewise.
7024 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7026 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7027 Increase LDP/STP opportunities by adding adjacent callee-saves.
7029 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7031 PR rtl-optimization/84068
7032 PR rtl-optimization/83459
7033 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7035 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7037 PR tree-optimization/84224
7038 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7039 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7042 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7045 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7046 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7047 count, remove one USE.
7049 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7051 * doc/install.texi (Configuration): Document the
7052 --with-long-double-format={ibm,ieee} PowerPC configuration
7056 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7057 Convert from define_expand to be define_insn_and_split. Rework
7058 float/double/_Float128 conversions to QI/HI/SImode to work with
7059 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7060 conversions to QI/HImode types did a store and then a load to
7061 truncate the value. For conversions to VSX registers, don't split
7062 the insn, instead emit the code directly. Use the code iterator
7063 any_fix to combine signed and unsigned conversions.
7064 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7065 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7066 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7067 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7068 (fix_<mode>di2_hw): Likewise.
7069 (fixuns_<mode>di2_hw): Likewise.
7070 (fix_<mode>si2_hw): Likewise.
7071 (fixuns_<mode>si2_hw): Likewise.
7072 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7073 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7074 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7075 fix<uns>_trunc<SFDF:mode>si2_p8.
7076 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7078 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7079 (fix<uns>_<mode>_mem): Likewise.
7080 (fctiw<u>z_<mode>_mem): Likewise.
7081 (fix<uns>_<mode>_mem): Likewise.
7082 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7083 the register allocator from doing a direct move to the GPRs to do
7084 a store, and instead use the ISA 3.0 store byte/half-word from
7085 vector register instruction. For IEEE 128-bit floating point,
7086 also optimize stores of 32-bit ints.
7087 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7089 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7091 * genextract.c (push_pathstr_operand): New function to support
7093 (walk_rtx): Call push_pathstr_operand.
7094 (print_path): Support [a-zA-Z].
7096 2018-02-07 Richard Biener <rguenther@suse.de>
7098 PR tree-optimization/84037
7099 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7100 (cse_and_gimplify_to_preheader): Declare.
7101 (vect_get_place_in_interleaving_chain): Likewise.
7102 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7104 (_loop_vec_info::~_loop_vec_info): Delete it.
7105 (cse_and_gimplify_to_preheader): New function.
7106 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7107 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7108 (vectorizable_load): Likewise. For grouped stores always base
7109 the IV on the first element.
7110 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7111 condition before gimplifying.
7113 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7115 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7116 *DIV_EXPR and *MOD_EXPR.
7118 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7121 * config/i386/i386.c (ix86_option_override_internal): Mask out
7122 the CF_SET bit when checking -fcf-protection.
7124 2018-02-07 Tom de Vries <tom@codesourcery.com>
7127 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7130 2018-02-07 Richard Biener <rguenther@suse.de>
7132 PR tree-optimization/84204
7133 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7136 PR tree-optimization/84205
7137 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7138 special-case isl_ast_op_zdiv_r.
7140 PR tree-optimization/84223
7141 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7142 Only add conditions from within the region.
7143 (gather_bbs::after_dom_children): Adjust.
7145 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7148 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7149 * config/avr/avr.md: Only post-reload split REG-REG moves if
7150 either register is GENERAL_REG_P.
7152 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7154 PR tree-optimization/84235
7155 * tree-ssa-scopedtables.c
7156 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7157 if the subtraction is performed in floating point type where NaNs are
7158 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7159 build 1. Formatting fix.
7161 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7164 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7165 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7166 and skip it regardless of bb boundaries. Use CALL_P macro,
7167 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7168 unnecessarily, formatting fix.
7170 2018-02-06 Michael Collison <michael.collison@arm.com>
7172 * config/arm/thumb2.md:
7173 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7174 (*thumb_mov_notscc): Ditto.
7176 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7179 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7180 unsigned_fix, not "s".
7182 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7184 * configure.ac (gcc_fn_eh_frame_ro): New function.
7185 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7186 correct .eh_frame permissions.
7187 * configure: Regenerate.
7189 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7191 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7194 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7196 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7197 Display warning message for -mno-speculate-indirect-jumps.
7199 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7201 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7203 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7205 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7207 PR tree-optimization/84225
7208 * tree-eh.c (find_trapping_overflow): Only call
7209 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7211 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7214 * config/i386/i386.c: Reimplement the check of possible options
7215 -mibt/-mshstk conbination. Change error messages.
7216 * doc/invoke.texi: Fix a typo: remove extra '='.
7218 2018-02-06 Marek Polacek <polacek@redhat.com>
7220 PR tree-optimization/84228
7221 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7223 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7226 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7227 emitted arch directives.
7228 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7229 __ARM_FEATURE_COPROC before changing architectures.
7231 2018-02-06 Richard Biener <rguenther@suse.de>
7233 * config/i386/i386.c (print_reg): Fix typo.
7234 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7236 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7238 * configure: Regenerate.
7240 2018-02-05 Martin Sebor <msebor@redhat.com>
7242 PR tree-optimization/83369
7243 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7246 2018-02-05 Martin Liska <mliska@suse.cz>
7248 * doc/invoke.texi: Cherry-pick upstream r323995.
7250 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7252 * ira.c (ira_init_register_move_cost): Adjust comment.
7254 2018-02-05 Martin Liska <mliska@suse.cz>
7256 PR gcov-profile/84137
7257 * doc/gcov.texi: Fix typo in documentation.
7259 2018-02-05 Martin Liska <mliska@suse.cz>
7261 PR gcov-profile/83879
7262 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7263 using dlopen functionality.
7265 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7267 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7268 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7269 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7270 _mm_maskz_range_round_ss): New intrinsics.
7271 (__builtin_ia32_rangesd128_round)
7272 (__builtin_ia32_rangess128_round): Remove.
7273 (__builtin_ia32_rangesd128_mask_round,
7274 __builtin_ia32_rangess128_mask_round): New builtins.
7275 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7276 __builtin_ia32_rangess128_round): Remove.
7277 (__builtin_ia32_rangesd128_mask_round,
7278 __builtin_ia32_rangess128_mask_round): New builtins.
7279 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7280 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7281 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7282 "<round_saeonly_constraint>")): Changed to ...
7283 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7284 "<round_saeonly_scalar_constraint>")): ... this.
7285 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7286 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7287 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7288 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7289 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7291 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7293 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7295 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7296 Remove all values except native, 8540 and 8548.
7298 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7300 * config/i386/i386.c (ix86_output_function_return): Pass
7301 INVALID_REGNUM, instead of -1, as invalid register number to
7302 indirect_thunk_name and output_indirect_thunk.
7304 2018-02-02 Julia Koval <julia.koval@intel.com>
7306 * config.gcc: Add -march=icelake.
7307 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7308 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7309 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7310 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7311 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7312 (processor_target_table): Add icelake.
7313 (ix86_option_override_internal): Handle new PTAs.
7314 (get_builtin_code_for_version): Handle icelake.
7315 (M_INTEL_COREI7_ICELAKE): New.
7316 (fold_builtin_cpu): Handle icelake.
7317 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7318 * doc/invoke.texi: Add -march=icelake.
7320 2018-02-02 Julia Koval <julia.koval@intel.com>
7322 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7323 to wide_int_bitmask.
7324 * wide-int-bitmask.h: New.
7326 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7329 * config/i386/i386.md: Replace Pmode with word_mode in
7330 builtin_setjmp_setup and builtin_longjmp to support x32.
7332 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7336 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7338 (rs6000_supported_cpu_names): New static variable.
7339 (linux_cpu_translation_table): Likewise.
7340 (elf_platform) <cpu>: Define new static variable and use it.
7341 Translate kernel AT_PLATFORM name to canonical name if needed.
7342 Error if platform name is unknown.
7344 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7347 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7349 2018-02-01 Jeff Law <law@redhat.com>
7352 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7353 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7354 the scratch if RELEASE_VIA_POP is false.
7355 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7356 If we have to save a temporary register, decrement SIZE appropriately.
7357 Pass new arguments to release_scratch_register_on_entry.
7358 (ix86_adjust_stack_and_probe): Likewise.
7359 (ix86_emit_probe_stack_range): Pass new arguments to
7360 release_scratch_register_on_entry.
7362 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7364 PR rtl-optimization/84157
7365 * combine.c (change_zero_ext): Use REG_P predicate in
7366 front of HARD_REGISTER_P predicate.
7368 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7370 * config/avr/avr.c (avr_option_override): Move disabling of
7371 -fdelete-null-pointer-checks to...
7372 * common/config/avr/avr-common.c (avr_option_optimization_table):
7375 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7377 PR tree-optimization/81635
7378 * tree-data-ref.c (split_constant_offset_1): For types that
7379 wrap on overflow, try to use range info to prove that wrapping
7382 2018-02-01 Renlin Li <renlin.li@arm.com>
7385 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7387 (aarch64_register_move_cost): Likewise.
7388 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7390 (REG_CLASS_NAMES): Likewise.
7391 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7392 TAILCALL_ADDR_REGS. Remove IP registers.
7393 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7395 2018-02-01 Richard Biener <rguenther@suse.de>
7397 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7398 for specifying RPO order and allow NULL for that.
7399 * domwalk.c (dom_walker::dom_walker): Likewise.
7400 (dom_walker::walk): Handle NULL RPO order.
7401 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7403 (rewrite_update_dom_walker): Likewise.
7404 (mark_def_dom_walker): Likewise.
7406 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7408 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7409 (aarch64_maybe_expand_sve_subreg_move): Declare.
7410 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7411 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7413 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7414 that are semantically a reverse operation.
7415 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7416 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7417 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7419 (aarch64_can_change_mode_class): For big-endian, forbid changes
7420 between two SVE modes if they have different element sizes.
7422 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7424 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7425 the TImode handling for big-endian targets.
7427 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7429 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7430 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7432 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7433 Remove BSWAP handing for big-endian targets and use the form of
7434 LD1RQ appropariate for the mode.
7436 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7438 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7439 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7442 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7445 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7446 check for operands that need to go through aarch64_sve_reload_be.
7448 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7450 PR tree-optimization/81661
7451 PR tree-optimization/84117
7452 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7453 * tree-eh.c: Include gimplify.h.
7454 (find_trapping_overflow, replace_trapping_overflow,
7455 rewrite_to_non_trapping_overflow): New functions.
7456 * tree-vect-loop.c: Include tree-eh.h.
7457 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7458 * tree-data-ref.c: Include tree-eh.h.
7459 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7461 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7463 PR rtl-optimization/84123
7464 * combine.c (change_zero_ext): Check if hard register satisfies
7465 can_change_dest_mode before calling gen_lowpart_SUBREG.
7467 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7470 * ira.c (ira_init_register_move_cost): Remove assert.
7472 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7474 PR rtl-optimization/84071
7475 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7476 * doc/tm.texi: Regenerate.
7478 2018-01-31 Richard Biener <rguenther@suse.de>
7480 PR tree-optimization/84132
7481 * tree-data-ref.c (analyze_miv_subscript): Properly
7482 check whether evolution_function_is_affine_multivariate_p
7483 before calling gcd_of_steps_may_divide_p.
7485 2018-01-31 Julia Koval <julia.koval@intel.com>
7488 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7489 * config/i386/i386.md (rdpid_rex64) New.
7490 (rdpid): Make 32bit only.
7492 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7495 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7496 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7498 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7501 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7503 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7505 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7507 PR rtl-optimization/84071
7508 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7509 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7511 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7513 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7514 (arc_attribute_table): Add 'aux' attribute.
7515 (arc_in_small_data_p): Consider aux like variables.
7516 (arc_is_aux_reg_p): New function.
7517 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7518 (arc_get_aux_arg): New function.
7519 (prepare_move_operands): Handle aux-register access.
7520 (arc_handle_aux_attribute): New function.
7521 * doc/extend.texi (ARC Variable attributes): Add subsection.
7523 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7525 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7526 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7527 (arc_attribute_table): Add 'uncached' attribute.
7528 (arc_print_operand): Print '.di' flag for uncached memory
7530 (arc_in_small_data_p): Do not consider for small data the uncached
7532 (arc_is_uncached_mem_p): New function.
7533 * config/arc/predicates.md (compact_store_memory_operand): Check
7534 for uncached memory accesses.
7535 (nonvol_nonimm_operand): Likewise.
7536 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7538 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7541 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7542 falign-loops=): Add Optimization flag.
7544 2018-01-30 Jeff Law <law@redhat.com>
7547 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7548 INT_REGISTERS_SAVED. Check it prior to calling
7549 get_scratch_register_on_entry.
7550 (ix86_adjust_stack_and_probe): Similarly.
7551 (ix86_emit_probe_stack_range): Similarly.
7552 (ix86_expand_prologue): Corresponding changes.
7554 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7557 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7558 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7560 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7563 * lra-constraints.c (curr_insn_transform): Process AND in the
7566 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7568 PR rtl-optimization/83986
7569 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7570 dependence against last_pending_memory_flush in addition to
7573 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7575 PR tree-optimization/81611
7576 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7579 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7582 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7585 2018-01-30 Richard Biener <rguenther@suse.de>
7586 Jakub Jelinek <jakub@redhat.com>
7588 PR tree-optimization/84111
7589 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7590 inner loops added during recursion, as they don't have up-to-date
7593 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7596 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7597 (can_inline_edge_by_limits_p): ... here.
7598 (can_early_inline_edge_p, check_callers,
7599 update_caller_keys, update_callee_keys, recursive_inlining,
7600 add_new_edges_to_heap, speculation_useful_p,
7601 inline_small_functions,
7602 inline_small_functions, flatten_function,
7603 inline_to_all_callers_1): Update.
7605 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7607 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7610 2018-01-30 Richard Biener <rguenther@suse.de>
7612 PR tree-optimization/83008
7613 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7614 invariant and constant vector uses in stmts when they need
7617 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7620 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7621 * configure: Regenerate.
7623 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7625 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7627 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7628 Use gen_rtx_REG rather than gen_lowpart.
7630 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7632 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7633 rather than 0 when creating partial subregs.
7635 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7637 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7640 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7643 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7644 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7645 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7646 flags. This restores the settings used before the 2017-07-24.
7647 Turning off pre increment/decrement/modify allows IVOPTS to
7648 optimize DF/SF loops where the index is an int.
7650 2018-01-29 Richard Biener <rguenther@suse.de>
7651 Kelvin Nilsen <kelvin@gcc.gnu.org>
7654 * tree-vect-stmts.c (vectorizable_call): Don't call
7655 targetm.vectorize_builtin_md_vectorized_function if callee is
7658 2018-01-22 Carl Love <cel@us.ibm.com>
7660 * doc/extend.tex: Fix typo in second arg in
7661 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7663 2018-01-29 Richard Biener <rguenther@suse.de>
7665 PR tree-optimization/84086
7666 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7667 (flush_ssaname_freelist): When SSA names were released reset
7668 the SCEV hash table.
7670 2018-01-29 Richard Biener <rguenther@suse.de>
7672 PR tree-optimization/84057
7673 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7674 removed paths when removing edges.
7676 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7678 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7679 -mfunction-return=@var{choice}.
7681 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7684 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7686 (layout::print_source_line): Likewise.
7687 (test_get_line_width_without_trailing_whitespace): Add test cases.
7689 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7692 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7695 2018-01-26 Jim Wilson <jimw@sifive.com>
7697 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7699 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7702 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7704 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7705 and CMP + SUB-immediate -> SUBS.
7707 2018-01-26 Martin Sebor <msebor@redhat.com>
7709 PR tree-optimization/83896
7710 * tree-ssa-strlen.c (get_string_len): Rename...
7711 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7712 Avoid assuming length is constant.
7713 (handle_char_store): Use HOST_WIDE_INT for string length.
7715 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7718 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7719 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7721 2018-01-26 Richard Biener <rguenther@suse.de>
7723 PR rtl-optimization/84003
7724 * dse.c (record_store): Only record redundant stores when
7725 the earlier store aliases at least all accesses the later one does.
7727 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7729 PR rtl-optimization/83985
7730 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7731 REG_CFA_RESTORE insns.
7732 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7733 REG_CFA_RESTORE insns here.
7736 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7737 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7739 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7741 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7742 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7743 (arc_init): Likewise.
7744 (arc_override_options): Likewise.
7745 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7747 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7749 * config/arc/arc.h (TARGET_DBNZ): Define.
7750 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7751 properly set the tune attribute.
7752 (dbnz): Use TARGET_DBNZ guard.
7753 * config/arc/arc.opt (mtune): Add core3 option.
7755 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7757 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7758 recognize new pic like addresses.
7759 (arc_delegitimize_address): Clean up.
7761 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7763 * config/arc/arc-arches.def: Option mrf16 valid for all
7765 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7766 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7767 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7768 * config/arc/arc-tables.opt: Regenerate.
7769 * config/arc/arc.c (arc_conditional_register_usage): Handle
7770 reduced register file case.
7771 (arc_file_start): Set must have build attributes.
7772 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7774 * config/arc/arc.opt (mrf16): Add new option.
7775 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7776 * config/arc/genmultilib.awk: Handle new mrf16 option.
7777 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7778 * config/arc/t-multilib: Regenerate.
7779 * doc/invoke.texi (ARC Options): Document mrf16 option.
7781 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7783 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7784 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7785 (arc_attribute_table): Add 'secure_call' attribute.
7786 (arc_print_operand): Print secure call operand.
7787 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7789 (arc_is_secure_call_p): New function. * config/arc/arc.md
7790 (call_i): Add support for sjli instruction.
7791 (call_value_i): Likewise.
7792 * config/arc/constraints.md (Csc): New constraint.
7794 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7795 John Eric Martin <John.Martin@emmicro-us.com>
7797 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7798 * config/arc/arc.c (_arc_jli_section): New struct.
7799 (arc_jli_section): New type.
7800 (rc_jli_sections): New static variable.
7801 (arc_handle_jli_attribute): New function.
7802 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7803 (arc_file_end): New function.
7804 (TARGET_ASM_FILE_END): Define.
7805 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7806 (arc_add_jli_section): New function.
7807 (jli_call_scan): Likewise.
7808 (arc_reorg): Call jli_call_scan.
7809 (arc_output_addsi): Remove 'S' from printing asm operand.
7810 (arc_is_jli_call_p): New function.
7811 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7813 (movhi_insn): Likewise.
7814 (movsi_insn): Likewise.
7815 (movsi_set_cc_insn): Likewise.
7816 (loadqi_update): Likewise.
7817 (load_zeroextendqisi_update): Likewise.
7818 (load_signextendqisi_update): Likewise.
7819 (loadhi_update): Likewise.
7820 (load_zeroextendhisi_update): Likewise.
7821 (load_signextendhisi_update): Likewise.
7822 (loadsi_update): Likewise.
7823 (loadsf_update): Likewise.
7824 (movsicc_insn): Likewise.
7825 (bset_insn): Likewise.
7826 (bxor_insn): Likewise.
7827 (bclr_insn): Likewise.
7828 (bmsk_insn): Likewise.
7829 (bicsi3_insn): Likewise.
7830 (cmpsi_cc_c_insn): Likewise.
7831 (movsi_ne): Likewise.
7832 (movsi_cond_exec): Likewise.
7833 (clrsbsi2): Likewise.
7844 (call_i): Remove 'S' asm letter, add jli instruction.
7845 (call_value_i): Likewise.
7846 * config/arc/arc.op (mjli-always): New option.
7847 * config/arc/constraints.md (Cji): New constraint.
7848 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7850 (subsf3_fpx): Likewise.
7851 (mulsf3_fpx): Likewise.
7852 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7854 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7856 * doc/invoke.texi (ARC): Document mjli-always option.
7858 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7860 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7861 avoid addition with 0 and use incw and decw where possible.
7863 2018-01-26 Richard Biener <rguenther@suse.de>
7865 PR tree-optimization/81082
7866 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7867 association if it requires casting to unsigned.
7868 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7869 from fold_plusminus_mult_expr to catch important cases late when
7870 range info is available.
7872 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7874 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7875 * configure.ac (hidden_linkonce): New test.
7876 * configure: Regenerate.
7877 * config.in: Regenerate.
7879 2018-01-26 Julia Koval <julia.koval@intel.com>
7881 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7882 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7883 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7884 _mm_mask_bitshuffle_epi64_mask): Fix type.
7885 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7886 USI_FTYPE_V4DI_V4DI_USI): Remove.
7887 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7888 __builtin_ia32_vpshufbitqmb256_mask,
7889 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7890 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7891 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7893 2018-01-26 Alan Modra <amodra@gmail.com>
7896 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7897 UNSPEC_VBPERMQ. Sort other unspecs.
7899 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7901 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7903 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7906 * predict.c (drop_profile): Do not push/pop cfun; update also
7908 (handle_missing_profiles): Fix logic looking for zero profiles.
7910 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7913 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7914 on functions with #pragma omp declare simd or functions with simd
7916 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7917 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7918 Remove trailing \n from warning_at calls.
7920 2018-01-25 Tom de Vries <tom@codesourcery.com>
7923 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7924 for neutered workers.
7926 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7929 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7930 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7932 2018-01-24 Jeff Law <law@redhat.com>
7935 * i386.c (get_probe_interval): Move to earlier point.
7936 (ix86_compute_frame_layout): If -fstack-clash-protection and
7937 the frame is larger than the probe interval, then use pushes
7938 to save registers rather than reg->mem moves.
7939 (ix86_expand_prologue): Remove conditional for int_registers_saved
7942 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7945 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7946 min/max for never referenced object.
7948 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7951 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7953 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7954 attributes from DECL_ATTRIBUTES (decl) without affecting
7955 DECL_ATTRIBUTES (current_function_decl).
7956 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7957 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7959 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7961 PR tree-optimization/83979
7962 * fold-const.c (fold_comparison): Use constant_boolean_node
7963 instead of boolean_{true,false}_node.
7965 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7967 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7970 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7972 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7973 Simplify the clause that sets the length attribute.
7974 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7975 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7976 clause that sets the length attribute.
7977 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7979 2018-01-24 Tom de Vries <tom@codesourcery.com>
7982 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7983 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7984 Add strict parameter.
7985 (prevent_branch_around_nothing): Insert dummy insn between branch to
7986 label and label with no ptx insn inbetween.
7987 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7989 2018-01-24 Tom de Vries <tom@codesourcery.com>
7992 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7993 for neutered threads in warp.
7994 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7996 2018-01-24 Richard Biener <rguenther@suse.de>
7998 PR tree-optimization/83176
7999 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8002 2018-01-24 Richard Biener <rguenther@suse.de>
8004 PR tree-optimization/82819
8005 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8006 code generating pluses that are no-ops in the target precision.
8008 2018-01-24 Richard Biener <rguenther@suse.de>
8011 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8013 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8015 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8016 to merge probabilities.
8017 * predict.c (probably_never_executed): Also mark as cold functions
8018 with global 0 profile and guessed local profile.
8019 * profile-count.c (profile_probability::combine_with_count): New
8021 * profile-count.h (profile_probability::operator*,
8022 profile_probability::operator*=, profile_probability::operator/,
8023 profile_probability::operator/=): Reduce precision to adjusted
8024 and set value to guessed on contradictory divisions.
8025 (profile_probability::combine_with_freq): Remove.
8026 (profile_probability::combine_wiht_count): Declare.
8027 (profile_count::force_nonzero):: Set to adjusted.
8028 (profile_count::probability_in):: Set quality to adjusted.
8029 * tree-ssa-tail-merge.c (replace_block_by): Use
8032 2018-01-23 Andrew Waterman <andrew@sifive.com>
8033 Jim Wilson <jimw@sifive.com>
8035 * config/riscv/riscv.c (riscv_stack_boundary): New.
8036 (riscv_option_override): Set riscv_stack_boundary. Handle
8037 riscv_preferred_stack_boundary_arg.
8038 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8039 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8040 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8041 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8042 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8043 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8045 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8048 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8049 of struct ix86_frame.
8050 (ix86_expand_epilogue): Likewise. Add a local variable for
8051 the reg_save_offset field in struct ix86_frame.
8053 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8055 PR tree-optimization/82604
8056 * tree-loop-distribution.c (enum partition_kind): New enum item
8057 PKIND_PARTIAL_MEMSET.
8058 (partition_builtin_p): Support above new enum item.
8059 (generate_code_for_partition): Ditto.
8060 (compute_access_range): Differentiate cases that equality can be
8061 proven at all loops, the innermost loops or no loops.
8062 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8063 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8064 (finalize_partitions, distribute_loop): Don't fuse partition of
8065 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8066 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8069 2018-01-23 Martin Liska <mliska@suse.cz>
8071 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8072 order to ignore the predictor.
8073 (PRED_POLYMORPHIC_CALL): Likewise.
8074 (PRED_RECURSIVE_CALL): Likewise.
8076 2018-01-23 Martin Liska <mliska@suse.cz>
8078 * tree-profile.c (tree_profiling): Print function header to
8079 aware reader which function we are working on.
8080 * value-prof.c (gimple_find_values_to_profile): Do not print
8081 not interesting value histograms.
8083 2018-01-23 Martin Liska <mliska@suse.cz>
8085 * profile-count.h (enum profile_quality): Add
8086 profile_uninitialized as the first value. Do not number values
8087 as they are zero based.
8088 (profile_count::verify): Update sanity check.
8089 (profile_probability::verify): Likewise.
8091 2018-01-23 Nathan Sidwell <nathan@acm.org>
8093 * doc/invoke.texi (ffor-scope): Deprecate.
8095 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8097 PR tree-optimization/83510
8098 * domwalk.c (set_all_edges_as_executable): New function.
8099 (dom_walker::dom_walker): Convert bool param
8100 "skip_unreachable_blocks" to enum reachability. Move setup of
8101 edge flags to set_all_edges_as_executable and only do it when
8102 reachability is REACHABLE_BLOCKS.
8103 * domwalk.h (enum dom_walker::reachability): New enum.
8104 (dom_walker::dom_walker): Convert bool param
8105 "skip_unreachable_blocks" to enum reachability.
8106 (set_all_edges_as_executable): New decl.
8107 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8108 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8110 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8111 but converting true to REACHABLE_BLOCKS.
8112 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8114 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8115 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8116 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8118 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8119 if check_all_array_refs will be called.
8121 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8123 * tree.c (selftest::test_location_wrappers): Add more test
8126 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8128 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8129 (selftest::test_bit_in_range): Likewise.
8131 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8134 * doc/sourcebuild.texi (vect_float): Say that the selector
8135 only describes the situation when -funsafe-math-optimizations is on.
8136 (vect_float_strict): Document.
8138 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8140 PR tree-optimization/83965
8141 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8142 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8143 instead of checking only for a reduction.
8144 (vect_recog_widen_sum_pattern): Likewise.
8146 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8148 * predict.c (probably_never_executed): Only use precise profile info.
8149 (compute_function_frequency): Skip after inlining hack since we now
8150 have quality checking.
8152 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8154 * profile-count.h (profile_probability::very_unlikely,
8155 profile_probability::unlikely, profile_probability::even): Set
8156 precision to guessed.
8158 2018-01-23 Richard Biener <rguenther@suse.de>
8160 PR tree-optimization/83963
8161 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8162 Properly terminate dominator walk when crossing the exit edge not
8163 when visiting its source block.
8165 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8168 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8169 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8171 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8173 PR tree-optimization/83957
8174 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8175 semicolon after for body surrounded by braces.
8177 PR tree-optimization/83081
8178 * profile-count.h (profile_probability::split): New method.
8179 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8180 Use profile_probability::split.
8181 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8182 when splitting a single conditional jump into 2.
8184 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8186 PR tree-optimization/69452
8187 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8190 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8192 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8193 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8194 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8196 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8198 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8199 * config/rl78/rl78.md (movdi): New define_expand.
8200 * config/rl78/rl78.c (rl78_split_movdi): New function.
8202 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8205 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8207 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8208 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8209 128-bit to produce an UNSPEC move to get the double word with the
8210 signbit and then a shift directly to do signbit.
8211 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8212 implementation with a new version that just does either a direct
8213 move or a regular move. Move memory interface to separate insns.
8214 Move insns so they are next to the expander.
8215 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8216 with signbit move. Split big and little endian case.
8217 (signbit<mode>2_dm_mem_le): Likewise.
8218 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8219 (signbit<mode>2_dm2): Likewise.
8221 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8223 * config/rl78/rl78.md (anddi3): New define_expand.
8225 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8227 * config/rl78/rl78.md (umindi3): New define_expand.
8229 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8231 * config/rl78/rl78.md (smindi3): New define_expand.
8233 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8235 * config/rl78/rl78.md (smaxdi3): New define_expand.
8237 2018-01-22 Carl Love <cel@us.ibm.com>
8239 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8240 LVX_V1TI): Add macro expansion.
8241 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8242 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8243 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8244 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8245 Change check to determine if the instruction is a byte reversing
8246 entry. Fix typo in comment.
8247 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8248 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8249 Add def_builtin calls for new builtins.
8250 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8251 Add define_insn expansion.
8253 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8255 * config/rl78/rl78.md (umaxdi3): New define_expand.
8257 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8259 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8260 for non-QImode registers.
8262 2018-01-22 Richard Biener <rguenther@suse.de>
8264 PR tree-optimization/83963
8265 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8266 including the loop exit block.
8267 (scop_detection::merge_sese): Likewise.
8268 (scop_detection::add_scop): Do it here instead.
8270 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8272 * doc/sourcebuild.texi (arm_softfloat): Document.
8274 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8277 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8278 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8279 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8281 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8282 David Edelsohn <dje.gcc@gmail.com>
8285 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8286 Change "crset eq" to "crset 2".
8287 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8288 (*call_indirect_aix<mode>_nospec): Likewise.
8289 (*call_value_indirect_aix<mode>_nospec): Likewise.
8290 (*call_indirect_elfv2<mode>_nospec): Likewise.
8291 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8292 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8293 change assembly output from . to $.
8294 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8295 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8296 (*tablejump<mode>_internal1_nospec): Likewise.
8298 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8301 * config/sh/sh_optimize_sett_clrt.cc:
8302 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8304 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8306 PR tree-optimization/83940
8307 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8308 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8309 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8310 use it to pass back the definition type.
8311 (vect_check_store_rhs): Likewise rhs_dt_out.
8312 (vect_build_gather_load_calls): Add a mask_dt argument and use
8313 it instead of a call to vect_is_simple_use.
8314 (vectorizable_store): Update calls to vect_check_load_store_mask
8315 and vect_check_store_rhs. Use the dt returned by the latter instead
8316 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8317 instead of calls to vect_is_simple_use. Pass the scalar rather
8318 than the vector operand to vect_is_simple_use when handling
8319 second and subsequent copies of an rhs value.
8320 (vectorizable_load): Update calls to vect_check_load_store_mask
8321 and vect_build_gather_load_calls. Use the cached mask_dt and
8322 gs_info.offset_dt instead of calls to vect_is_simple_use.
8324 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8327 * tree-emutls.c: Include gimplify.h.
8328 (lower_emutls_2): New function.
8329 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8330 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8331 it before further processing.
8334 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8335 UINTVAL (trueop1) instead of INTVAL (op1).
8337 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8341 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8342 INCOMING_FRAME_SP_OFFSET if not defined.
8343 (scan_trace): Add ENTRY argument. If true and
8344 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8345 emit a note to adjust the CFA offset.
8346 (create_cfi_notes): Adjust scan_trace callers.
8347 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8348 INCOMING_FRAME_SP_OFFSET in the CIE.
8349 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8350 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8352 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8353 * doc/tm.texi: Regenerated.
8355 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8357 PR rtl-optimization/83147
8358 * lra-constraints.c (remove_inheritance_pseudos): Use
8359 lra_substitute_pseudo_within_insn.
8361 2018-01-19 Tom de Vries <tom@codesourcery.com>
8362 Cesar Philippidis <cesar@codesourcery.com>
8365 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8367 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8370 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8371 spaces for function labels.
8373 2018-01-19 Martin Liska <mliska@suse.cz>
8375 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8376 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8377 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8378 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8379 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8380 (PRED_CONST_RETURN): Change from 69 to 65.
8381 (PRED_NULL_RETURN): Change from 91 to 71.
8382 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8383 (PRED_LOOP_GUARD): Change from 66 to 73.
8385 2018-01-19 Martin Liska <mliska@suse.cz>
8387 * predict.c (predict_insn_def): Add new assert.
8388 (struct branch_predictor): Change type to signed integer.
8389 (test_prediction_value_range): Amend test to cover
8391 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8392 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8393 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8394 (PRED_LOOP_IV_COMPARE): Likewise.
8395 * predict.h (PROB_UNINITIALIZED): Define new constant.
8397 2018-01-19 Martin Liska <mliska@suse.cz>
8399 * predict.c (dump_prediction): Add new format for
8400 analyze_brprob.py script which is enabled with -details
8402 * profile-count.h (precise_p): New function.
8404 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8406 PR tree-optimization/83922
8407 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8408 there are no statements that need masking.
8409 (vect_active_double_reduction_p): New function.
8410 (vect_analyze_loop_operations): Use it when handling phis that
8411 are not in the loop header.
8413 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8415 PR tree-optimization/83914
8416 * tree-vect-loop.c (vectorizable_induction): Don't convert
8417 init_expr or apply the peeling adjustment for inductions
8418 that are nested within the vectorized loop.
8420 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8422 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8425 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8429 * function.h (gimplify_parameters): Add gimple_seq * argument.
8430 * function.c: Include gimple.h and options.h.
8431 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8432 for the added local temporaries if needed.
8433 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8434 if there are any parameter cleanups, wrap whole body into a
8435 try/finally with the cleanups.
8437 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8440 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8441 Use GET_MODE_CLASS for scalar floating point.
8443 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8447 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8448 Fix call of call_cgraph_insertion_hooks.
8450 2018-01-18 Martin Sebor <msebor@redhat.com>
8452 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8454 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8457 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8460 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8463 * common.opt: (-ffile-prefix-map): New option.
8464 * opts.c (common_handle_option): Defer it.
8465 * opts-global.c (handle_common_deferred_options): Handle it.
8466 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8467 * file-prefix-map.h: New file.
8468 (remap_debug_filename, add_debug_prefix_map): ...here.
8469 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8470 * final.c (debug_prefix_map, add_debug_prefix_map
8471 remap_debug_filename): Move to...
8472 * file-prefix-map.c: New file.
8473 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8474 generalize, get rid of alloca(), use strrchr() instead of strchr().
8475 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8476 Implement in terms of add_prefix_map().
8477 (remap_macro_filename, remap_debug_filename): Implement in term of
8479 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8480 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8481 * dbxout.c: Include file-prefix-map.h.
8482 * varasm.c: Likewise.
8483 * vmsdbgout.c: Likewise.
8484 * xcoffout.c: Likewise.
8485 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8486 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8487 * doc/invoke.texi (-ffile-prefix-map): Document.
8488 (-fdebug-prefix-map): Update description.
8490 2018-01-18 Martin Liska <mliska@suse.cz>
8492 * config/i386/i386.c (indirect_thunk_name): Document that also
8494 (output_indirect_thunk): Document why both instructions
8495 (pause and lfence) are generated.
8497 2018-01-18 Richard Biener <rguenther@suse.de>
8499 PR tree-optimization/83887
8500 * graphite-scop-detection.c
8501 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8502 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8503 (scop_detection::merge_sese): Re-implement with a flood-fill
8504 algorithm that properly finds a SESE region if it exists.
8506 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8509 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8510 pointer_diff optimizations use view_convert instead of convert.
8512 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8514 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8515 Generate different code for -mno-speculate-indirect-jumps.
8516 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8517 (*call_indirect_aix<mode>): Disable for
8518 -mno-speculate-indirect-jumps.
8519 (*call_indirect_aix<mode>_nospec): New define_insn.
8520 (*call_value_indirect_aix<mode>): Disable for
8521 -mno-speculate-indirect-jumps.
8522 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8523 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8524 -mno-speculate-indirect-jumps.
8525 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8527 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8529 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8530 long double type, set the flags for noting the default long double
8531 type, even if we don't pass or return a long double type.
8533 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8536 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8539 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8541 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8542 support for merge[hl].
8543 (fold_mergehl_helper): New helper function.
8544 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8545 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8546 (altivec_vmrglw_direct): Add xxmrglw insn.
8548 2018-01-17 Andrew Waterman <andrew@sifive.com>
8550 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8551 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8553 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8556 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8557 call the lto_location_cache before reading the
8558 DECL_SOURCE_LOCATION of the types.
8560 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8561 Richard Sandiford <richard.sandiford@linaro.org>
8563 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8564 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8565 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8566 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8567 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8569 * config/aarch64/constraints.md (aarch64_movti_operand):
8571 * config/aarch64/predicates.md (Uti): Add new constraint.
8573 2018-01-17 Carl Love <cel@us.ibm.com>
8575 * config/rs6000/vsx.md (define_expand xl_len_r,
8576 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8577 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8579 (define_expand, define_insn): Move the shift left from the
8580 define_insn to the define_expand for lxvl and stxvl instructions.
8581 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8582 and XL_LEN_R definitions to PURE.
8584 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8586 * config/i386/i386.c (indirect_thunk_name): Declare regno
8587 as unsigned int. Compare regno with INVALID_REGNUM.
8588 (output_indirect_thunk): Ditto.
8589 (output_indirect_thunk_function): Ditto.
8590 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8591 in the call to output_indirect_thunk_function.
8593 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8596 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8597 rather than the size of inner_type to determine the stack slot size
8598 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8600 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8603 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8606 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8608 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8609 endian Linux systems to optionally enable multilibs for selecting
8610 the long double type if the user configured an explicit type.
8611 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8612 have no long double multilibs if not defined.
8613 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8614 warn if the user used -mabi={ieee,ibm}longdouble and we built
8615 multilibs for long double.
8616 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8617 appropriate multilib option.
8618 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8620 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8621 for building long double multilibs.
8622 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8624 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8626 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8629 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8631 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8634 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8637 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8640 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8642 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8643 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8645 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8647 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8648 different rtl trees depending on TARGET_64BIT.
8649 (rs6000_gen_lvx): Likewise.
8651 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8653 * config/visium/visium.md (nop): Tweak comment.
8654 (hazard_nop): Likewise.
8656 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8658 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8659 -mspeculate-indirect-jumps.
8660 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8661 for -mno-speculate-indirect-jumps.
8662 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8663 (*call_value_indirect_elfv2<mode>): Disable for
8664 -mno-speculate-indirect-jumps.
8665 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8666 (indirect_jump): Emit different RTL for
8667 -mno-speculate-indirect-jumps.
8668 (*indirect_jump<mode>): Disable for
8669 -mno-speculate-indirect-jumps.
8670 (*indirect_jump<mode>_nospec): New define_insn.
8671 (tablejump): Emit different RTL for
8672 -mno-speculate-indirect-jumps.
8673 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8674 (tablejumpsi_nospec): New define_expand.
8675 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8676 (tablejumpdi_nospec): New define_expand.
8677 (*tablejump<mode>_internal1): Disable for
8678 -mno-speculate-indirect-jumps.
8679 (*tablejump<mode>_internal1_nospec): New define_insn.
8680 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8683 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8685 * caller-save.c (insert_save): Drop unnecessary parameter. All
8688 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8689 Richard Biener <rguenth@suse.de>
8692 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8693 return early, inline manually is_gimple_sizepos. Make sure if we
8694 call gimplify_expr we don't end up with a gimple constant.
8695 * tree.c (variably_modified_type_p): Don't return true for
8696 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8697 * gimplify.h (is_gimple_sizepos): Remove.
8699 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8701 PR tree-optimization/83857
8702 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8703 vectorizable_live_operation for pure SLP statements.
8704 (vectorizable_live_operation): Handle PHIs.
8706 2018-01-16 Richard Biener <rguenther@suse.de>
8708 PR tree-optimization/83867
8709 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8710 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8712 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8715 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8716 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8717 If off is not INTEGER_CST, issue a may not be aligned warning
8718 rather than isn't aligned. Use isn%'t rather than isn't.
8719 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8721 <case MULT_EXPR>: Improve the case when bottom and one of the
8722 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8723 operand, in that case check if the other operand is multiple of
8724 bottom divided by the INTEGER_CST operand.
8726 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8729 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8730 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8731 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8732 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8733 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8734 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8735 (pa_function_arg_size): New function.
8737 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8739 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8740 in a separate statement.
8742 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8744 PR tree-optimization/83847
8745 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8746 group gathers and scatters.
8748 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8750 PR rtl-optimization/86620
8751 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8753 PR rtl-optimization/83213
8754 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8755 to last if both are JUMP_INSNs.
8757 PR tree-optimization/83843
8758 * gimple-ssa-store-merging.c
8759 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8760 store_immediate_info for bswap/nop orig_stores.
8762 2018-01-15 Andrew Waterman <andrew@sifive.com>
8764 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8766 <UDIV>: Increase cost if !TARGET_DIV.
8768 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8770 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8771 (define_attr "cr_logical_3op"): New.
8772 (cceq_ior_compare): Adjust.
8773 (cceq_ior_compare_complement): Adjust.
8774 (*cceq_rev_compare): Adjust.
8775 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8776 (is_cracked_insn): Adjust.
8777 (insn_must_be_first_in_group): Adjust.
8778 * config/rs6000/40x.md: Adjust.
8779 * config/rs6000/440.md: Adjust.
8780 * config/rs6000/476.md: Adjust.
8781 * config/rs6000/601.md: Adjust.
8782 * config/rs6000/603.md: Adjust.
8783 * config/rs6000/6xx.md: Adjust.
8784 * config/rs6000/7450.md: Adjust.
8785 * config/rs6000/7xx.md: Adjust.
8786 * config/rs6000/8540.md: Adjust.
8787 * config/rs6000/cell.md: Adjust.
8788 * config/rs6000/e300c2c3.md: Adjust.
8789 * config/rs6000/e500mc.md: Adjust.
8790 * config/rs6000/e500mc64.md: Adjust.
8791 * config/rs6000/e5500.md: Adjust.
8792 * config/rs6000/e6500.md: Adjust.
8793 * config/rs6000/mpc.md: Adjust.
8794 * config/rs6000/power4.md: Adjust.
8795 * config/rs6000/power5.md: Adjust.
8796 * config/rs6000/power6.md: Adjust.
8797 * config/rs6000/power7.md: Adjust.
8798 * config/rs6000/power8.md: Adjust.
8799 * config/rs6000/power9.md: Adjust.
8800 * config/rs6000/rs64.md: Adjust.
8801 * config/rs6000/titan.md: Adjust.
8803 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8805 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8806 ix86_indirect_branch_register logic.
8808 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8810 * config/i386/constraints.md (Bs): Update
8811 ix86_indirect_branch_register check. Don't check
8812 ix86_indirect_branch_register with GOT_memory_operand.
8814 * config/i386/predicates.md (GOT_memory_operand): Don't check
8815 ix86_indirect_branch_register here.
8816 (GOT32_symbol_operand): Likewise.
8818 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8820 * config/i386/predicates.md (constant_call_address_operand):
8821 Rewrite ix86_indirect_branch_register logic.
8822 (sibcall_insn_operand): Likewise.
8824 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8826 * config/i386/constraints.md (Bs): Replace
8827 ix86_indirect_branch_thunk_register with
8828 ix86_indirect_branch_register.
8830 * config/i386/i386.md (indirect_jump): Likewise.
8831 (tablejump): Likewise.
8832 (*sibcall_memory): Likewise.
8833 (*sibcall_value_memory): Likewise.
8834 Peepholes of indirect call and jump via memory: Likewise.
8835 * config/i386/i386.opt: Likewise.
8836 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8837 (GOT_memory_operand): Likewise.
8838 (call_insn_operand): Likewise.
8839 (sibcall_insn_operand): Likewise.
8840 (GOT32_symbol_operand): Likewise.
8842 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8845 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8846 type rather than type addr's type points to.
8847 (expand_omp_atomic_mutex): Likewise.
8848 (expand_omp_atomic): Likewise.
8850 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8853 * config/i386/i386.c (output_indirect_thunk_function): Use
8854 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8855 for __x86_return_thunk.
8857 2018-01-15 Richard Biener <rguenther@suse.de>
8860 * expmed.c (extract_bit_field_1): Fix typo.
8862 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8865 * config/arm/iterators.md (VF): New mode iterator.
8866 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8867 Remove integer-related logic from pattern.
8868 (neon_vabd<mode>_3): Likewise.
8870 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8873 * common.opt (fstrict-overflow): No longer an alias.
8874 (fwrapv-pointer): New option.
8875 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8876 also for pointer types based on flag_wrapv_pointer.
8877 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8878 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8879 opts->x_flag_wrapv got set.
8880 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8881 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8882 POINTER_TYPE_OVERFLOW_UNDEFINED.
8883 * match.pd: Likewise in address comparison pattern.
8884 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8886 2018-01-15 Richard Biener <rguenther@suse.de>
8889 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8890 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8891 Reset type names to their identifier if their TYPE_DECL doesn't
8892 have linkage (and thus is used for ODR and devirt).
8893 (save_debug_info_for_decl): Remove.
8894 (save_debug_info_for_type): Likewise.
8895 (add_tree_to_fld_list): Adjust.
8896 * tree-pretty-print.c (dump_generic_node): Make dumping of
8897 type names more robust.
8899 2018-01-15 Richard Biener <rguenther@suse.de>
8901 * BASE-VER: Bump to 8.0.1.
8903 2018-01-14 Martin Sebor <msebor@redhat.com>
8906 * builtins.c (check_access): Avoid warning when the no-warning bit
8909 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8911 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8912 * ira-color (allocno_hard_regs_compare): Likewise.
8914 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8917 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8918 Use .pushsection/.popsection.
8920 2018-01-14 Martin Sebor <msebor@redhat.com>
8923 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8925 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8927 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8928 entry from extra_headers.
8929 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8930 extra_headers, make the list bitwise identical to the i?86-*-* one.
8932 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8934 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8935 -mcmodel=large with -mindirect-branch=thunk,
8936 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8937 -mfunction-return=thunk-extern.
8938 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8939 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8940 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8942 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8944 * config/i386/i386.c (print_reg): Print the name of the full
8945 integer register without '%'.
8946 (ix86_print_operand): Handle 'V'.
8947 * doc/extend.texi: Document 'V' modifier.
8949 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8951 * config/i386/constraints.md (Bs): Disallow memory operand for
8952 -mindirect-branch-register.
8954 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8955 (GOT_memory_operand): Likewise.
8956 (call_insn_operand): Likewise.
8957 (sibcall_insn_operand): Likewise.
8958 (GOT32_symbol_operand): Likewise.
8959 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8960 for -mindirect-branch-register.
8961 (tablejump): Likewise.
8962 (*sibcall_memory): Likewise.
8963 (*sibcall_value_memory): Likewise.
8964 Disallow peepholes of indirect call and jump via memory for
8965 -mindirect-branch-register.
8966 (*call_pop): Replace m with Bw.
8967 (*call_value_pop): Likewise.
8968 (*sibcall_pop_memory): Replace m with Bs.
8969 * config/i386/i386.opt (mindirect-branch-register): New option.
8970 * doc/invoke.texi: Document -mindirect-branch-register option.
8972 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8974 * config/i386/i386-protos.h (ix86_output_function_return): New.
8975 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8976 set function_return_type.
8977 (indirect_thunk_name): Add ret_p to indicate thunk for function
8979 (output_indirect_thunk_function): Pass false to
8980 indirect_thunk_name.
8981 (ix86_output_indirect_branch_via_reg): Likewise.
8982 (ix86_output_indirect_branch_via_push): Likewise.
8983 (output_indirect_thunk_function): Create alias for function
8984 return thunk if regno < 0.
8985 (ix86_output_function_return): New function.
8986 (ix86_handle_fndecl_attribute): Handle function_return.
8987 (ix86_attribute_table): Add function_return.
8988 * config/i386/i386.h (machine_function): Add
8989 function_return_type.
8990 * config/i386/i386.md (simple_return_internal): Use
8991 ix86_output_function_return.
8992 (simple_return_internal_long): Likewise.
8993 * config/i386/i386.opt (mfunction-return=): New option.
8994 (indirect_branch): Mention -mfunction-return=.
8995 * doc/extend.texi: Document function_return function attribute.
8996 * doc/invoke.texi: Document -mfunction-return= option.
8998 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9000 * config/i386/i386-opts.h (indirect_branch): New.
9001 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9002 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9003 with local indirect jump when converting indirect call and jump.
9004 (ix86_set_indirect_branch_type): New.
9005 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9006 (indirectlabelno): New.
9007 (indirect_thunk_needed): Likewise.
9008 (indirect_thunk_bnd_needed): Likewise.
9009 (indirect_thunks_used): Likewise.
9010 (indirect_thunks_bnd_used): Likewise.
9011 (INDIRECT_LABEL): Likewise.
9012 (indirect_thunk_name): Likewise.
9013 (output_indirect_thunk): Likewise.
9014 (output_indirect_thunk_function): Likewise.
9015 (ix86_output_indirect_branch_via_reg): Likewise.
9016 (ix86_output_indirect_branch_via_push): Likewise.
9017 (ix86_output_indirect_branch): Likewise.
9018 (ix86_output_indirect_jmp): Likewise.
9019 (ix86_code_end): Call output_indirect_thunk_function if needed.
9020 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9022 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9023 (ix86_attribute_table): Add indirect_branch.
9024 * config/i386/i386.h (machine_function): Add indirect_branch_type
9025 and has_local_indirect_jump.
9026 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9028 (tablejump): Likewise.
9029 (*indirect_jump): Use ix86_output_indirect_jmp.
9030 (*tablejump_1): Likewise.
9031 (simple_return_indirect_internal): Likewise.
9032 * config/i386/i386.opt (mindirect-branch=): New option.
9033 (indirect_branch): New.
9036 (thunk-inline): Likewise.
9037 (thunk-extern): Likewise.
9038 * doc/extend.texi: Document indirect_branch function attribute.
9039 * doc/invoke.texi: Document -mindirect-branch= option.
9041 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9044 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9046 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9048 * ipa-inline.c (want_inline_small_function_p): Return false if
9049 inlining has already failed with CIF_FINAL_ERROR.
9050 (update_caller_keys): Call want_inline_small_function_p before
9052 (update_callee_keys): Likewise.
9054 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9056 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9058 (rs6000_quadword_masked_address_p): Likewise.
9059 (quad_aligned_load_p): Likewise.
9060 (quad_aligned_store_p): Likewise.
9061 (const_load_sequence_p): Add comment to describe the outer-most loop.
9062 (mimic_memory_attributes_and_flags): New function.
9063 (rs6000_gen_stvx): Likewise.
9064 (replace_swapped_aligned_store): Likewise.
9065 (rs6000_gen_lvx): Likewise.
9066 (replace_swapped_aligned_load): Likewise.
9067 (replace_swapped_load_constant): Capitalize argument name in
9068 comment describing this function.
9069 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9070 and stores that access quad-word aligned addresses and replace
9071 with stvx or lvx instructions when appropriate.
9072 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9073 New function prototype.
9074 (rs6000_quadword_masked_address_p): Likewise.
9075 (rs6000_gen_lvx): Likewise.
9076 (rs6000_gen_stvx): Likewise.
9077 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9078 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9079 when memory address is aligned.
9080 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9081 this split to select lvx instruction when memory address is aligned.
9082 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9083 instruction when memory address is aligned.
9084 (*vsx_le_perm_load_v16qi): Likewise.
9085 (four unnamed splitters): Modify to select the stvx instruction
9086 when memory is aligned.
9088 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9090 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9091 which appears in the queue multiple times.
9093 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9094 Alan Hayward <alan.hayward@arm.com>
9095 David Sherwood <david.sherwood@arm.com>
9097 * tree-vectorizer.h (vec_lower_bound): New structure.
9098 (_loop_vec_info): Add check_nonzero and lower_bounds.
9099 (LOOP_VINFO_CHECK_NONZERO): New macro.
9100 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9101 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9102 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9103 fields. Make seg_len the distance travelled, not including the
9105 (dr_direction_indicator): Declare.
9106 (dr_zero_step_indicator): Likewise.
9107 (dr_known_forward_stride_p): Likewise.
9108 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9110 (runtime_alias_check_p): Allow runtime alias checks with
9112 (operator ==): Compare access_size and align.
9113 (prune_runtime_alias_test_list): Rework for new distinction between
9114 the access_size and seg_len.
9115 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9117 (get_segment_min_max): New function.
9118 (create_intersect_range_checks): Use it.
9119 (dr_step_indicator): New function.
9120 (dr_direction_indicator): Likewise.
9121 (dr_zero_step_indicator): Likewise.
9122 (dr_known_forward_stride_p): Likewise.
9123 * tree-loop-distribution.c (data_ref_segment_size): Return
9124 DR_STEP * (niters - 1).
9125 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9127 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9128 (vect_preserves_scalar_order_p): New function, split out from...
9129 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9130 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9131 (vect_vfa_access_size): New function.
9132 (vect_vfa_align): Likewise.
9133 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9134 (dump_lower_bound): New function.
9135 (vect_check_lower_bound): Likewise.
9136 (vect_small_gap_p): Likewise.
9137 (vectorizable_with_step_bound_p): Likewise.
9138 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9139 depencies if the vectorization factor is 1. Convert the checks
9140 for nonzero steps into checks on the bounds of DR_STEP. Try using
9141 a bunds check for variable steps if the minimum required step is
9142 relatively small. Update calls to the dr_with_seg_len
9143 constructor and to vect_compile_time_alias.
9144 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9146 (vect_loop_versioning): Call it.
9147 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9149 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9151 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9152 Alan Hayward <alan.hayward@arm.com>
9153 David Sherwood <david.sherwood@arm.com>
9155 * doc/sourcebuild.texi (vect_scatter_store): Document.
9156 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9158 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9160 * genopinit.c (main): Add supports_vec_scatter_store and
9161 supports_vec_scatter_store_cached to target_optabs.
9162 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9163 IFN_MASK_SCATTER_STORE.
9164 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9166 * internal-fn.h (internal_store_fn_p): Declare.
9167 (internal_fn_stored_value_index): Likewise.
9168 * internal-fn.c (scatter_store_direct): New macro.
9169 (expand_scatter_store_optab_fn): New function.
9170 (direct_scatter_store_optab_supported_p): New macro.
9171 (internal_store_fn_p): New function.
9172 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9173 IFN_MASK_SCATTER_STORE.
9174 (internal_fn_mask_index): Likewise.
9175 (internal_fn_stored_value_index): New function.
9176 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9178 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9179 * optabs-query.c (supports_vec_scatter_store_p): New function.
9180 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9181 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9182 true for scatter stores.
9183 (vect_gather_scatter_fn_p): Handle scatter stores too.
9184 (vect_check_gather_scatter): Consider using scatter stores if
9185 supports_vec_scatter_store_p.
9186 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9188 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9189 internal_fn_stored_value_index.
9190 (check_load_store_masking): Handle scatter stores too.
9191 (vect_get_store_rhs): Make public.
9192 (vectorizable_call): Use internal_store_fn_p.
9193 (vectorizable_store): Handle scatter store internal functions.
9194 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9195 when deciding whether the end of the group has been reached.
9196 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9197 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9198 (mask_scatter_store<mode>): New insns.
9200 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9201 Alan Hayward <alan.hayward@arm.com>
9202 David Sherwood <david.sherwood@arm.com>
9204 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9205 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9206 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9208 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9209 Use vect_truncate_gather_scatter_offset if we can't treat the
9210 operation as a normal gather load or scatter store.
9211 (get_group_load_store_type): Take the gather_scatter_info
9212 as argument. Try using a gather load or scatter store for
9213 single-element groups.
9214 (get_load_store_type): Update calls to get_group_load_store_type
9215 and vect_use_strided_gather_scatters_p.
9217 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9218 Alan Hayward <alan.hayward@arm.com>
9219 David Sherwood <david.sherwood@arm.com>
9221 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9222 optional tree argument.
9223 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9225 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9226 but continue to use the current value as a fallback.
9227 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9228 to compare the updates.
9229 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9230 (get_load_store_type): Use it when handling a strided access.
9231 (vect_get_strided_load_store_ops): New function.
9232 (vect_get_data_ptr_increment): Likewise.
9233 (vectorizable_load): Handle strided gather loads. Always pass
9234 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9236 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9237 Alan Hayward <alan.hayward@arm.com>
9238 David Sherwood <david.sherwood@arm.com>
9240 * doc/md.texi (gather_load@var{m}): Document.
9241 (mask_gather_load@var{m}): Likewise.
9242 * genopinit.c (main): Add supports_vec_gather_load and
9243 supports_vec_gather_load_cached to target_optabs.
9244 * optabs-tree.c (init_tree_optimization_optabs): Use
9245 ggc_cleared_alloc to allocate target_optabs.
9246 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9247 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9249 * internal-fn.h (internal_load_fn_p): Declare.
9250 (internal_gather_scatter_fn_p): Likewise.
9251 (internal_fn_mask_index): Likewise.
9252 (internal_gather_scatter_fn_supported_p): Likewise.
9253 * internal-fn.c (gather_load_direct): New macro.
9254 (expand_gather_load_optab_fn): New function.
9255 (direct_gather_load_optab_supported_p): New macro.
9256 (direct_internal_fn_optab): New function.
9257 (internal_load_fn_p): Likewise.
9258 (internal_gather_scatter_fn_p): Likewise.
9259 (internal_fn_mask_index): Likewise.
9260 (internal_gather_scatter_fn_supported_p): Likewise.
9261 * optabs-query.c (supports_at_least_one_mode_p): New function.
9262 (supports_vec_gather_load_p): Likewise.
9263 * optabs-query.h (supports_vec_gather_load_p): Declare.
9264 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9265 and memory_type field.
9266 (NUM_PATTERNS): Bump to 15.
9267 * tree-vect-data-refs.c: Include internal-fn.h.
9268 (vect_gather_scatter_fn_p): New function.
9269 (vect_describe_gather_scatter_call): Likewise.
9270 (vect_check_gather_scatter): Try using internal functions for
9271 gather loads. Recognize existing calls to a gather load function.
9272 (vect_analyze_data_refs): Consider using gather loads if
9273 supports_vec_gather_load_p.
9274 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9275 (vect_get_gather_scatter_offset_type): Likewise.
9276 (vect_convert_mask_for_vectype): Likewise.
9277 (vect_add_conversion_to_patterm): Likewise.
9278 (vect_try_gather_scatter_pattern): Likewise.
9279 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9280 (vect_vect_recog_func_ptrs): Add it.
9281 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9282 internal_fn_mask_index and internal_gather_scatter_fn_p.
9283 (check_load_store_masking): Take the gather_scatter_info as an
9284 argument and handle gather loads.
9285 (vect_get_gather_scatter_ops): New function.
9286 (vectorizable_call): Check internal_load_fn_p.
9287 (vectorizable_load): Likewise. Handle gather load internal
9289 (vectorizable_store): Update call to check_load_store_masking.
9290 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9291 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9292 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9293 (aarch64_gather_scale_operand_d): New predicates.
9294 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9295 (mask_gather_load<mode>): New insns.
9297 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9298 Alan Hayward <alan.hayward@arm.com>
9299 David Sherwood <david.sherwood@arm.com>
9301 * optabs.def (fold_left_plus_optab): New optab.
9302 * doc/md.texi (fold_left_plus_@var{m}): Document.
9303 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9304 * internal-fn.c (fold_left_direct): Define.
9305 (expand_fold_left_optab_fn): Likewise.
9306 (direct_fold_left_optab_supported_p): Likewise.
9307 * fold-const-call.c (fold_const_fold_left): New function.
9308 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9309 * tree-parloops.c (valid_reduction_p): New function.
9310 (gather_scalar_reductions): Use it.
9311 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9312 (vect_finish_replace_stmt): Declare.
9313 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9314 (needs_fold_left_reduction_p): New function, split out from...
9315 (vect_is_simple_reduction): ...here. Accept reductions that
9316 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9317 (vect_force_simple_reduction): Also store the reduction type in
9318 the assignment's STMT_VINFO_REDUC_TYPE.
9319 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9320 (merge_with_identity): New function.
9321 (vect_expand_fold_left): Likewise.
9322 (vectorize_fold_left_reduction): Likewise.
9323 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9324 scalar phi in place for it. Check for target support and reject
9325 cases that would reassociate the operation. Defer the transform
9326 phase to vectorize_fold_left_reduction.
9327 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9328 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9329 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9331 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9333 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9334 call to ifc_temp_var.
9336 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9337 Alan Hayward <alan.hayward@arm.com>
9338 David Sherwood <david.sherwood@arm.com>
9340 * target.def (legitimize_address_displacement): Take the original
9341 offset as a poly_int.
9342 * targhooks.h (default_legitimize_address_displacement): Update
9344 * targhooks.c (default_legitimize_address_displacement): Likewise.
9345 * doc/tm.texi: Regenerate.
9346 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9347 as an argument, moving assert of ad->disp == ad->disp_term to...
9348 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9349 Try calling targetm.legitimize_address_displacement before expanding
9350 the address rather than afterwards, and adjust for the new interface.
9351 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9352 Match the new hook interface. Handle SVE addresses.
9353 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9356 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9358 * Makefile.in (OBJS): Add early-remat.o.
9359 * target.def (select_early_remat_modes): New hook.
9360 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9361 * doc/tm.texi: Regenerate.
9362 * targhooks.h (default_select_early_remat_modes): Declare.
9363 * targhooks.c (default_select_early_remat_modes): New function.
9364 * timevar.def (TV_EARLY_REMAT): New timevar.
9365 * passes.def (pass_early_remat): New pass.
9366 * tree-pass.h (make_pass_early_remat): Declare.
9367 * early-remat.c: New file.
9368 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9370 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9372 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9373 Alan Hayward <alan.hayward@arm.com>
9374 David Sherwood <david.sherwood@arm.com>
9376 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9377 vfm1 with a bound_epilog parameter.
9378 (vect_do_peeling): Update calls accordingly, and move the prologue
9379 call earlier in the function. Treat the base bound_epilog as 0 for
9380 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9381 this base when peeling for gaps.
9382 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9383 with fully-masked loops.
9384 (vect_estimate_min_profitable_iters): Handle the single peeled
9385 iteration in that case.
9387 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9388 Alan Hayward <alan.hayward@arm.com>
9389 David Sherwood <david.sherwood@arm.com>
9391 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9392 single-element interleaving even if the size is not a power of 2.
9393 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9394 accesses for single-element interleaving if the group size is
9397 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9398 Alan Hayward <alan.hayward@arm.com>
9399 David Sherwood <david.sherwood@arm.com>
9401 * doc/md.texi (fold_extract_last_@var{m}): Document.
9402 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9403 * optabs.def (fold_extract_last_optab): New optab.
9404 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9405 * internal-fn.c (fold_extract_direct): New macro.
9406 (expand_fold_extract_optab_fn): Likewise.
9407 (direct_fold_extract_optab_supported_p): Likewise.
9408 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9409 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9410 EXTRACT_LAST_REDUCTION.
9411 (get_initial_def_for_reduction): Do not create an initial vector
9412 for EXTRACT_LAST_REDUCTION reductions.
9413 (vectorizable_reduction): Leave the scalar phi in place for
9414 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9415 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9416 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9417 transform phase to vectorizable_condition.
9418 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9420 (vect_finish_stmt_generation): ...here.
9421 (vect_finish_replace_stmt): New function.
9422 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9423 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9425 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9427 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9428 Alan Hayward <alan.hayward@arm.com>
9429 David Sherwood <david.sherwood@arm.com>
9431 * doc/md.texi (extract_last_@var{m}): Document.
9432 * optabs.def (extract_last_optab): New optab.
9433 * internal-fn.def (EXTRACT_LAST): New internal function.
9434 * internal-fn.c (cond_unary_direct): New macro.
9435 (expand_cond_unary_optab_fn): Likewise.
9436 (direct_cond_unary_optab_supported_p): Likewise.
9437 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9438 loops using EXTRACT_LAST.
9439 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9440 (extract_last_<mode>): ...this optab.
9441 (vec_extract<mode><Vel>): Update accordingly.
9443 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9444 Alan Hayward <alan.hayward@arm.com>
9445 David Sherwood <david.sherwood@arm.com>
9447 * target.def (empty_mask_is_expensive): New hook.
9448 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9449 * doc/tm.texi: Regenerate.
9450 * targhooks.h (default_empty_mask_is_expensive): Declare.
9451 * targhooks.c (default_empty_mask_is_expensive): New function.
9452 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9453 if the target says that empty masks are expensive.
9454 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9456 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9458 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9459 Alan Hayward <alan.hayward@arm.com>
9460 David Sherwood <david.sherwood@arm.com>
9462 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9463 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9464 (vect_use_loop_mask_for_alignment_p): New function.
9465 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9466 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9467 niters_skip argument. Make sure that the first niters_skip elements
9468 of the first iteration are inactive.
9469 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9470 Update call to vect_set_loop_masks_directly.
9471 (get_misalign_in_elems): New function, split out from...
9472 (vect_gen_prolog_loop_niters): ...here.
9473 (vect_update_init_of_dr): Take a code argument that specifies whether
9474 the adjustment should be added or subtracted.
9475 (vect_update_init_of_drs): Likewise.
9476 (vect_prepare_for_masked_peels): New function.
9477 (vect_do_peeling): Skip prologue peeling if we're using a mask
9478 instead. Update call to vect_update_inits_of_drs.
9479 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9481 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9482 alignment. Do not include the number of peeled iterations in
9483 the minimum threshold in that case.
9484 (vectorizable_induction): Adjust the start value down by
9485 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9486 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9487 Take the number of skipped iterations into account when calculating
9489 * tree-vect-stmts.c (vect_gen_while_not): New function.
9491 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9492 Alan Hayward <alan.hayward@arm.com>
9493 David Sherwood <david.sherwood@arm.com>
9495 * doc/sourcebuild.texi (vect_fully_masked): Document.
9496 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9498 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9500 (vect_analyze_loop_2): ...here. Don't check the vectorization
9501 factor against the number of loop iterations if the loop is
9504 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9505 Alan Hayward <alan.hayward@arm.com>
9506 David Sherwood <david.sherwood@arm.com>
9508 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9509 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9510 (dump_groups): Update accordingly.
9511 (iv_use::mem_type): New member variable.
9512 (address_p): New function.
9513 (record_use): Add a mem_type argument and initialize the new
9515 (record_group_use): Add a mem_type argument. Use address_p.
9516 Remove obsolete null checks of base_object. Update call to record_use.
9517 (find_interesting_uses_op): Update call to record_group_use.
9518 (find_interesting_uses_cond): Likewise.
9519 (find_interesting_uses_address): Likewise.
9520 (get_mem_type_for_internal_fn): New function.
9521 (find_address_like_use): Likewise.
9522 (find_interesting_uses_stmt): Try find_address_like_use before
9523 calling find_interesting_uses_op.
9524 (addr_offset_valid_p): Use the iv mem_type field as the type
9525 of the addressed memory.
9526 (add_autoinc_candidates): Likewise.
9527 (get_address_cost): Likewise.
9528 (split_small_address_groups_p): Use address_p.
9529 (split_address_groups): Likewise.
9530 (add_iv_candidate_for_use): Likewise.
9531 (autoinc_possible_for_pair): Likewise.
9532 (rewrite_groups): Likewise.
9533 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9534 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9535 (get_alias_ptr_type_for_ptr_address): New function.
9536 (rewrite_use_address): Rewrite address uses in calls that were
9537 identified by find_address_like_use.
9539 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9540 Alan Hayward <alan.hayward@arm.com>
9541 David Sherwood <david.sherwood@arm.com>
9543 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9545 * gimple-expr.h (is_gimple_addressable: Likewise.
9546 * gimple-expr.c (is_gimple_address): Likewise.
9547 * internal-fn.c (expand_call_mem_ref): New function.
9548 (expand_mask_load_optab_fn): Use it.
9549 (expand_mask_store_optab_fn): Likewise.
9551 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9552 Alan Hayward <alan.hayward@arm.com>
9553 David Sherwood <david.sherwood@arm.com>
9555 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9556 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9557 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9558 (cond_umax@var{mode}): Document.
9559 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9560 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9561 (cond_umin_optab, cond_umax_optab): New optabs.
9562 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9563 (COND_IOR, COND_XOR): New internal functions.
9564 * internal-fn.h (get_conditional_internal_fn): Declare.
9565 * internal-fn.c (cond_binary_direct): New macro.
9566 (expand_cond_binary_optab_fn): Likewise.
9567 (direct_cond_binary_optab_supported_p): Likewise.
9568 (get_conditional_internal_fn): New function.
9569 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9570 Cope with reduction statements that are vectorized as calls rather
9572 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9573 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9574 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9575 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9576 (UNSPEC_COND_EOR): New unspecs.
9577 (optab): Add mappings for them.
9578 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9579 (sve_int_op, sve_fp_op): New int attributes.
9581 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9582 Alan Hayward <alan.hayward@arm.com>
9583 David Sherwood <david.sherwood@arm.com>
9585 * optabs.def (while_ult_optab): New optab.
9586 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9587 * internal-fn.def (WHILE_ULT): New internal function.
9588 * internal-fn.h (direct_internal_fn_supported_p): New override
9589 that takes two types as argument.
9590 * internal-fn.c (while_direct): New macro.
9591 (expand_while_optab_fn): New function.
9592 (convert_optab_supported_p): Likewise.
9593 (direct_while_optab_supported_p): New macro.
9594 * wide-int.h (wi::udiv_ceil): New function.
9595 * tree-vectorizer.h (rgroup_masks): New structure.
9596 (vec_loop_masks): New typedef.
9597 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9599 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9600 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9601 (vect_max_vf): New function.
9602 (slpeel_make_loop_iterate_ntimes): Delete.
9603 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9604 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9605 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9606 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9607 internal-fn.h, stor-layout.h and optabs-query.h.
9608 (vect_set_loop_mask): New function.
9609 (add_preheader_seq): Likewise.
9610 (add_header_seq): Likewise.
9611 (interleave_supported_p): Likewise.
9612 (vect_maybe_permute_loop_masks): Likewise.
9613 (vect_set_loop_masks_directly): Likewise.
9614 (vect_set_loop_condition_masked): Likewise.
9615 (vect_set_loop_condition_unmasked): New function, split out from
9616 slpeel_make_loop_iterate_ntimes.
9617 (slpeel_make_loop_iterate_ntimes): Rename to..
9618 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9619 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9620 (vect_do_peeling): Update call accordingly.
9621 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9623 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9624 mask_compare_type, can_fully_mask_p and fully_masked_p.
9625 (release_vec_loop_masks): New function.
9626 (_loop_vec_info): Use it to free the loop masks.
9627 (can_produce_all_loop_masks_p): New function.
9628 (vect_get_max_nscalars_per_iter): Likewise.
9629 (vect_verify_full_masking): Likewise.
9630 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9631 retries, and free the mask rgroups before retrying. Check loop-wide
9632 reasons for disallowing fully-masked loops. Make the final decision
9633 about whether use a fully-masked loop or not.
9634 (vect_estimate_min_profitable_iters): Do not assume that peeling
9635 for the number of iterations will be needed for fully-masked loops.
9636 (vectorizable_reduction): Disable fully-masked loops.
9637 (vectorizable_live_operation): Likewise.
9638 (vect_halve_mask_nunits): New function.
9639 (vect_double_mask_nunits): Likewise.
9640 (vect_record_loop_mask): Likewise.
9641 (vect_get_loop_mask): Likewise.
9642 (vect_transform_loop): Handle the case in which the final loop
9643 iteration might handle a partial vector. Call vect_set_loop_condition
9644 instead of slpeel_make_loop_iterate_ntimes.
9645 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9646 (check_load_store_masking): New function.
9647 (prepare_load_store_mask): Likewise.
9648 (vectorizable_store): Handle fully-masked loops.
9649 (vectorizable_load): Likewise.
9650 (supportable_widening_operation): Use vect_halve_mask_nunits for
9652 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9653 (vect_gen_while): New function.
9654 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9655 (aarch64_uqdec<mode>): New insn.
9657 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9658 Alan Hayward <alan.hayward@arm.com>
9659 David Sherwood <david.sherwood@arm.com>
9661 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9662 (reduc_xor_scal_optab): New optabs.
9663 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9664 (reduc_xor_scal_@var{m}): Document.
9665 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9666 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9668 * fold-const-call.c (fold_const_call): Handle them.
9669 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9670 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9671 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9672 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9673 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9674 (UNSPEC_XORV): New unspecs.
9675 (optab): Add entries for them.
9676 (BITWISEV): New int iterator.
9677 (bit_reduc_op): New int attributes.
9679 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9680 Alan Hayward <alan.hayward@arm.com>
9681 David Sherwood <david.sherwood@arm.com>
9683 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9684 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9685 * optabs.def (vec_shl_insert_optab): New optab.
9686 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9687 (duplicate_and_interleave): Likewise.
9688 * tree-vect-loop.c: Include internal-fn.h.
9689 (neutral_op_for_slp_reduction): New function, split out from
9690 get_initial_defs_for_reduction.
9691 (get_initial_def_for_reduction): Handle option 2 for variable-length
9692 vectors by loading the neutral value into a vector and then shifting
9693 the initial value into element 0.
9694 (get_initial_defs_for_reduction): Replace the code argument with
9695 the neutral value calculated by neutral_op_for_slp_reduction.
9696 Use gimple_build_vector for constant-length vectors.
9697 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9698 but the first group_size elements have a neutral value.
9699 Use duplicate_and_interleave otherwise.
9700 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9701 Update call to get_initial_defs_for_reduction. Handle SLP
9702 reductions for variable-length vectors by creating one vector
9703 result for each scalar result, with the elements associated
9704 with other scalar results stubbed out with the neutral value.
9705 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9706 Require IFN_VEC_SHL_INSERT for double reductions on
9707 variable-length vectors, or SLP reductions that have
9708 a neutral value. Require can_duplicate_and_interleave_p
9709 support for variable-length unchained SLP reductions if there
9710 is no neutral value, such as for MIN/MAX reductions. Also require
9711 the number of vector elements to be a multiple of the number of
9712 SLP statements when doing variable-length unchained SLP reductions.
9713 Update call to vect_create_epilog_for_reduction.
9714 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9715 and remove initial values.
9716 (duplicate_and_interleave): Make public.
9717 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9718 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9720 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9721 Alan Hayward <alan.hayward@arm.com>
9722 David Sherwood <david.sherwood@arm.com>
9724 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9725 (can_duplicate_and_interleave_p): New function.
9726 (vect_get_and_check_slp_defs): Take the vector of statements
9727 rather than just the current one. Remove excess parentheses.
9728 Restriction rejectinon of vect_constant_def and vect_external_def
9729 for variable-length vectors to boolean types, or types for which
9730 can_duplicate_and_interleave_p is false.
9731 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9732 (duplicate_and_interleave): New function.
9733 (vect_get_constant_vectors): Use gimple_build_vector for
9734 constant-length vectors and suitable variable-length constant
9735 vectors. Use duplicate_and_interleave for other variable-length
9736 vectors. Don't defer the update when inserting new statements.
9738 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9739 Alan Hayward <alan.hayward@arm.com>
9740 David Sherwood <david.sherwood@arm.com>
9742 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9743 min_profitable_iters doesn't go negative.
9745 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9746 Alan Hayward <alan.hayward@arm.com>
9747 David Sherwood <david.sherwood@arm.com>
9749 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9750 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9751 * optabs.def (vec_mask_load_lanes_optab): New optab.
9752 (vec_mask_store_lanes_optab): Likewise.
9753 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9754 (MASK_STORE_LANES): Likewise.
9755 * internal-fn.c (mask_load_lanes_direct): New macro.
9756 (mask_store_lanes_direct): Likewise.
9757 (expand_mask_load_optab_fn): Handle masked operations.
9758 (expand_mask_load_lanes_optab_fn): New macro.
9759 (expand_mask_store_optab_fn): Handle masked operations.
9760 (expand_mask_store_lanes_optab_fn): New macro.
9761 (direct_mask_load_lanes_optab_supported_p): Likewise.
9762 (direct_mask_store_lanes_optab_supported_p): Likewise.
9763 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9765 (vect_load_lanes_supported): Likewise.
9766 * tree-vect-data-refs.c (strip_conversion): New function.
9767 (can_group_stmts_p): Likewise.
9768 (vect_analyze_data_ref_accesses): Use it instead of checking
9769 for a pair of assignments.
9770 (vect_store_lanes_supported): Take a masked_p parameter.
9771 (vect_load_lanes_supported): Likewise.
9772 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9773 vect_store_lanes_supported and vect_load_lanes_supported.
9774 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9775 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9776 parameter. Don't allow gaps for masked accesses.
9777 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9778 and vect_load_lanes_supported.
9779 (get_load_store_type): Take a masked_p parameter and update
9780 call to get_group_load_store_type.
9781 (vectorizable_store): Update call to get_load_store_type.
9782 Handle IFN_MASK_STORE_LANES.
9783 (vectorizable_load): Update call to get_load_store_type.
9784 Handle IFN_MASK_LOAD_LANES.
9786 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9787 Alan Hayward <alan.hayward@arm.com>
9788 David Sherwood <david.sherwood@arm.com>
9790 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9792 * config/aarch64/aarch64-protos.h
9793 (aarch64_sve_struct_memory_operand_p): Declare.
9794 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9795 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9796 (VPRED, vpred): Handle SVE structure modes.
9797 * config/aarch64/constraints.md (Utx): New constraint.
9798 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9799 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9800 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9801 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9802 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9803 structure modes. Split into pieces after RA.
9804 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9805 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9807 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9808 SVE structure modes.
9809 (aarch64_classify_address): Likewise.
9810 (sizetochar): Move earlier in file.
9811 (aarch64_print_operand): Handle SVE register lists.
9812 (aarch64_array_mode): New function.
9813 (aarch64_sve_struct_memory_operand_p): Likewise.
9814 (TARGET_ARRAY_MODE): Redefine.
9816 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9817 Alan Hayward <alan.hayward@arm.com>
9818 David Sherwood <david.sherwood@arm.com>
9820 * target.def (array_mode): New target hook.
9821 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9822 * doc/tm.texi: Regenerate.
9823 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9824 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9825 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9827 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9830 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9831 Alan Hayward <alan.hayward@arm.com>
9832 David Sherwood <david.sherwood@arm.com>
9834 * fold-const.c (fold_binary_loc): Check the argument types
9835 rather than the result type when testing for a vector operation.
9837 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9839 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9840 * doc/tm.texi: Regenerate.
9842 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9843 Alan Hayward <alan.hayward@arm.com>
9844 David Sherwood <david.sherwood@arm.com>
9846 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9847 (sve): Document new AArch64 extension.
9848 * doc/md.texi (w): Extend the description of the AArch64
9849 constraint to include SVE vectors.
9850 (Upl, Upa): Document new AArch64 predicate constraints.
9851 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9853 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9854 (msve-vector-bits=): New option.
9855 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9856 SVE when these are disabled.
9857 (sve): New extension.
9858 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9859 modes. Adjust their number of units based on aarch64_sve_vg.
9860 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9861 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9862 aarch64_addr_query_type.
9863 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9864 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9865 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9866 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9867 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9868 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9869 (aarch64_simd_imm_zero_p): Delete.
9870 (aarch64_check_zero_based_sve_index_immediate): Declare.
9871 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9872 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9873 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9874 (aarch64_sve_float_mul_immediate_p): Likewise.
9875 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9877 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9878 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9879 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9880 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9881 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9882 (aarch64_regmode_natural_size): Likewise.
9883 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9884 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9886 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9887 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9888 for VG and the SVE predicate registers.
9889 (V_ALIASES): Add a "z"-prefixed alias.
9890 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9891 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9892 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9893 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9894 (REG_CLASS_NAMES): Add entries for them.
9895 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9896 and the predicate registers.
9897 (aarch64_sve_vg): Declare.
9898 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9899 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9900 (REGMODE_NATURAL_SIZE): Define.
9901 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9903 * config/aarch64/aarch64.c: Include cfgrtl.h.
9904 (simd_immediate_info): Add a constructor for series vectors,
9905 and an associated step field.
9906 (aarch64_sve_vg): New variable.
9907 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9908 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9909 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9910 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9911 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9912 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9913 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9914 (aarch64_get_mask_mode): New functions.
9915 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9916 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9917 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9918 predicate modes and predicate registers. Explicitly restrict
9919 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9920 to store a vector mode if it is recognized by
9921 aarch64_classify_vector_mode.
9922 (aarch64_regmode_natural_size): New function.
9923 (aarch64_hard_regno_caller_save_mode): Return the original mode
9925 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9926 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9927 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9928 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9930 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9931 does not overlap dest if the function is frame-related. Handle
9933 (aarch64_split_add_offset): New function.
9934 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9935 them aarch64_add_offset.
9936 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9937 and update call to aarch64_sub_sp.
9938 (aarch64_add_cfa_expression): New function.
9939 (aarch64_expand_prologue): Pass extra temporary registers to the
9940 functions above. Handle the case in which we need to emit new
9941 DW_CFA_expressions for registers that were originally saved
9942 relative to the stack pointer, but now have to be expressed
9943 relative to the frame pointer.
9944 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9946 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9947 IP0 and IP1 values for SVE frames.
9948 (aarch64_expand_vec_series): New function.
9949 (aarch64_expand_sve_widened_duplicate): Likewise.
9950 (aarch64_expand_sve_const_vector): Likewise.
9951 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9952 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9953 into the register, rather than emitting a SET directly.
9954 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9955 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9956 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9957 (offset_9bit_signed_scaled_p): New functions.
9958 (aarch64_replicate_bitmask_imm): New function.
9959 (aarch64_bitmask_imm): Use it.
9960 (aarch64_cannot_force_const_mem): Reject expressions involving
9961 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9962 (aarch64_classify_index): Handle SVE indices, by requiring
9963 a plain register index with a scale that matches the element size.
9964 (aarch64_classify_address): Handle SVE addresses. Assert that
9965 the mode of the address is VOIDmode or an integer mode.
9966 Update call to aarch64_classify_symbol.
9967 (aarch64_classify_symbolic_expression): Update call to
9968 aarch64_classify_symbol.
9969 (aarch64_const_vec_all_in_range_p): New function.
9970 (aarch64_print_vector_float_operand): Likewise.
9971 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9972 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9973 and the FP immediates 1.0 and 0.5.
9974 (aarch64_print_address_internal): Handle SVE addresses.
9975 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9976 (aarch64_regno_regclass): Handle predicate registers.
9977 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9979 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9980 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9981 (aarch64_convert_sve_vector_bits): New function.
9982 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9983 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9985 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9986 Handle SVE vector and predicate modes. Accept VL-based constants
9987 that need only one temporary register, and VL offsets that require
9988 no temporary registers.
9989 (aarch64_conditional_register_usage): Mark the predicate registers
9990 as fixed if SVE isn't available.
9991 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9992 Return true for SVE vector and predicate modes.
9993 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9994 rather than an unsigned int. Handle SVE modes.
9995 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9997 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9999 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10000 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10001 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10002 (aarch64_sve_float_mul_immediate_p): New functions.
10003 (aarch64_sve_valid_immediate): New function.
10004 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10005 Explicitly reject structure modes. Check for INDEX constants.
10006 Handle PTRUE and PFALSE constants.
10007 (aarch64_check_zero_based_sve_index_immediate): New function.
10008 (aarch64_simd_imm_zero_p): Delete.
10009 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10010 vector modes. Accept constants in the range of CNT[BHWD].
10011 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10012 ask for an Advanced SIMD mode.
10013 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10014 (aarch64_simd_vector_alignment): Handle SVE predicates.
10015 (aarch64_vectorize_preferred_vector_alignment): New function.
10016 (aarch64_simd_vector_alignment_reachable): Use it instead of
10018 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10019 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10021 (MAX_VECT_LEN): Delete.
10022 (expand_vec_perm_d): Add a vec_flags field.
10023 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10024 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10025 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10027 (aarch64_evpc_rev): Rename to...
10028 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10029 (aarch64_evpc_rev_global): New function.
10030 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10031 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10033 (aarch64_evpc_sve_tbl): New function.
10034 (aarch64_expand_vec_perm_const_1): Update after rename of
10035 aarch64_evpc_rev. Handle SVE permutes too, trying
10036 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10037 than aarch64_evpc_tbl.
10038 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10039 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10040 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10041 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10042 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10043 (aarch64_expand_sve_vcond): New functions.
10044 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10045 of aarch64_vector_mode_p.
10046 (aarch64_dwarf_poly_indeterminate_value): New function.
10047 (aarch64_compute_pressure_classes): Likewise.
10048 (aarch64_can_change_mode_class): Likewise.
10049 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10050 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10051 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10052 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10053 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10054 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10055 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10056 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10058 (Dn, Dl, Dr): Accept const as well as const_vector.
10059 (Dz): Likewise. Compare against CONST0_RTX.
10060 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10061 of "vector" where appropriate.
10062 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10063 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10064 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10065 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10066 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10067 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10068 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10069 (v_int_equiv): Extend to SVE modes.
10070 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10072 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10073 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10074 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10075 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10076 (SVE_COND_FP_CMP): New int iterators.
10077 (perm_hilo): Handle the new unpack unspecs.
10078 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10080 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10081 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10082 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10083 (aarch64_equality_operator, aarch64_constant_vector_operand)
10084 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10085 (aarch64_sve_nonimmediate_operand): Likewise.
10086 (aarch64_sve_general_operand): Likewise.
10087 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10088 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10089 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10090 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10091 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10092 (aarch64_sve_float_arith_immediate): Likewise.
10093 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10094 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10095 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10096 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10097 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10098 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10099 (aarch64_sve_float_arith_operand): Likewise.
10100 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10101 (aarch64_sve_float_mul_operand): Likewise.
10102 (aarch64_sve_vec_perm_operand): Likewise.
10103 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10104 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10105 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10106 as well as const_vector.
10107 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10108 in file. Use CONST0_RTX and CONSTM1_RTX.
10109 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10110 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10111 Use aarch64_simd_imm_zero.
10112 * config/aarch64/aarch64-sve.md: New file.
10113 * config/aarch64/aarch64.md: Include it.
10114 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10115 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10116 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10117 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10118 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10119 (sve): New attribute.
10120 (enabled): Disable instructions with the sve attribute unless
10122 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10123 aarch64_expand_mov_immediate.
10124 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10125 CNT[BHSD] immediates.
10126 (movti): Split CONST_POLY_INT moves into two halves.
10127 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10128 Split additions that need a temporary here if the destination
10129 is the stack pointer.
10130 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10131 (*add<mode>3_poly_1): New instruction.
10132 (set_clobber_cc): New expander.
10134 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10136 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10137 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10138 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10139 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10140 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10141 Change innermode from fixed_mode_size to machine_mode.
10142 (simplify_subreg): Update call accordingly. Handle a constant-sized
10143 subreg of a variable-length CONST_VECTOR.
10145 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10146 Alan Hayward <alan.hayward@arm.com>
10147 David Sherwood <david.sherwood@arm.com>
10149 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10150 (add_offset_to_base): New function, split out from...
10151 (create_mem_ref): ...here. When handling a scale other than 1,
10152 check first whether the address is valid without the offset.
10153 Add it into the base if so, leaving the index and scale as-is.
10155 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10158 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10159 fold_for_warn before checking if arg2 is INTEGER_CST.
10161 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10163 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10164 (store_multiple_operation): Delete.
10165 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10166 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10167 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10168 guarded by TARGET_STRING.
10169 (rs6000_output_load_multiple): Delete.
10170 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10171 OPTION_MASK_STRING / TARGET_STRING handling.
10172 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10173 (const rs6000_opt_masks) <"string">: Change mask to 0.
10174 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10175 (MASK_STRING): Delete.
10176 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10178 (load_multiple): Delete.
10185 (store_multiple): Delete.
10192 (movmemsi_8reg): Delete.
10193 (corresponding unnamed define_insn): Delete.
10194 (movmemsi_6reg): Delete.
10195 (corresponding unnamed define_insn): Delete.
10196 (movmemsi_4reg): Delete.
10197 (corresponding unnamed define_insn): Delete.
10198 (movmemsi_2reg): Delete.
10199 (corresponding unnamed define_insn): Delete.
10200 (movmemsi_1reg): Delete.
10201 (corresponding unnamed define_insn): Delete.
10202 * config/rs6000/rs6000.opt (mno-string): New.
10203 (mstring): Replace by deprecation warning stub.
10204 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10206 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10208 * regrename.c (regrename_do_replace): If replacing the same
10209 reg multiple times, try to reuse last created gen_raw_REG.
10212 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10213 main to workaround a bug in GDB.
10215 2018-01-12 Tom de Vries <tom@codesourcery.com>
10218 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10220 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10222 PR rtl-optimization/80481
10223 * ira-color.c (get_cap_member): New function.
10224 (allocnos_conflict_by_live_ranges_p): Use it.
10225 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10226 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10228 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10231 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10232 (*saddl_se_1): Ditto.
10233 (*ssubsi_1): Ditto.
10234 (*ssubl_se_1): Ditto.
10236 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10238 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10239 rather than wi::to_widest for DR_INITs.
10240 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10241 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10242 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10244 (vect_analyze_group_access_1): Note that here.
10246 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10248 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10249 polynomial type sizes.
10251 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10253 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10254 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10255 (gimple_add_tmp_var): Likewise.
10257 2018-01-12 Martin Liska <mliska@suse.cz>
10259 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10260 (gimple_alloc_sizes): Likewise.
10261 (dump_gimple_statistics): Use PRIu64 in printf format.
10262 * gimple.h: Change uint64_t to int.
10264 2018-01-12 Martin Liska <mliska@suse.cz>
10266 * tree-core.h: Use uint64_t instead of int.
10267 * tree.c (tree_node_counts): Likewise.
10268 (tree_node_sizes): Likewise.
10269 (dump_tree_statistics): Use PRIu64 in printf format.
10271 2018-01-12 Martin Liska <mliska@suse.cz>
10273 * Makefile.in: As qsort_chk is implemented in vec.c, add
10274 vec.o to linkage of gencfn-macros.
10275 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10276 passing the info to record_node_allocation_statistics.
10277 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10279 * ggc-common.c (struct ggc_usage): Add operator== and use
10280 it in operator< and compare function.
10281 * mem-stats.h (struct mem_usage): Likewise.
10282 * vec.c (struct vec_usage): Remove operator< and compare
10283 function. Can be simply inherited.
10285 2018-01-12 Martin Jambor <mjambor@suse.cz>
10288 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10289 * tree-ssa-math-opts.c: Include domwalk.h.
10290 (convert_mult_to_fma_1): New function.
10291 (fma_transformation_info): New type.
10292 (fma_deferring_state): Likewise.
10293 (cancel_fma_deferring): New function.
10294 (result_of_phi): Likewise.
10295 (last_fma_candidate_feeds_initial_phi): Likewise.
10296 (convert_mult_to_fma): Added deferring logic, split actual
10297 transformation to convert_mult_to_fma_1.
10298 (math_opts_dom_walker): New type.
10299 (math_opts_dom_walker::after_dom_children): New method, body moved
10300 here from pass_optimize_widening_mul::execute, added deferring logic
10302 (pass_optimize_widening_mul::execute): Moved most of code to
10303 math_opts_dom_walker::after_dom_children.
10304 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10305 * config/i386/i386.c (ix86_option_override_internal): Added
10306 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10308 2018-01-12 Richard Biener <rguenther@suse.de>
10311 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10312 inline instance vars.
10314 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10317 * config/rx/rx.c (rx_is_restricted_memory_address):
10318 Handle SUBREG case.
10320 2018-01-12 Richard Biener <rguenther@suse.de>
10322 PR tree-optimization/80846
10323 * target.def (split_reduction): New target hook.
10324 * targhooks.c (default_split_reduction): New function.
10325 * targhooks.h (default_split_reduction): Declare.
10326 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10327 target requests first reduce vectors by combining low and high
10329 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10330 (get_vectype_for_scalar_type_and_size): Export.
10331 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10332 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10333 * doc/tm.texi: Regenerate.
10334 * config/i386/i386.c (ix86_split_reduction): Implement
10335 TARGET_VECTORIZE_SPLIT_REDUCTION.
10337 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10340 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10341 in PIC mode except for TARGET_VXWORKS_RTP.
10342 * config/sparc/sparc.c: Include cfgrtl.h.
10343 (TARGET_INIT_PIC_REG): Define.
10344 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10345 (sparc_pic_register_p): New predicate.
10346 (sparc_legitimate_address_p): Use it.
10347 (sparc_legitimize_pic_address): Likewise.
10348 (sparc_delegitimize_address): Likewise.
10349 (sparc_mode_dependent_address_p): Likewise.
10350 (gen_load_pcrel_sym): Remove 4th parameter.
10351 (load_got_register): Adjust call to above. Remove obsolete stuff.
10352 (sparc_expand_prologue): Do not call load_got_register here.
10353 (sparc_flat_expand_prologue): Likewise.
10354 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10355 (sparc_use_pseudo_pic_reg): New function.
10356 (sparc_init_pic_reg): Likewise.
10357 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10358 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10360 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10362 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10363 Add item for branch_cost.
10365 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10367 PR rtl-optimization/83565
10368 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10369 not extend the result to a larger mode for rotate operations.
10370 (num_sign_bit_copies1): Likewise.
10372 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10375 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10377 Use values-Xc.o for -pedantic.
10378 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10380 2018-01-12 Martin Liska <mliska@suse.cz>
10383 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10385 (possible_polymorphic_call_targets): Use it.
10386 (ipa_devirt): Likewise.
10388 2018-01-12 Martin Liska <mliska@suse.cz>
10390 * profile-count.h (enum profile_quality): Use 0 as invalid
10391 enum value of profile_quality.
10393 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10395 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10396 -mext-string options.
10398 2018-01-12 Richard Biener <rguenther@suse.de>
10400 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10401 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10402 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10404 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10406 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10408 * configure.ac (--with-long-double-format): Add support for the
10409 configuration option to change the default long double format on
10411 * config.gcc (powerpc*-linux*-*): Likewise.
10412 * configure: Regenerate.
10413 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10414 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10415 used without modification.
10417 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10419 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10420 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10421 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10422 MISC_BUILTIN_SPEC_BARRIER.
10423 (rs6000_init_builtins): Likewise.
10424 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10426 (speculation_barrier): New define_insn.
10427 * doc/extend.texi: Document __builtin_speculation_barrier.
10429 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10432 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10433 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10434 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10436 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10437 integral modes instead of "ss" and "sd".
10438 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10439 vectors with 32-bit and 64-bit elements.
10440 (vecdupssescalarmodesuffix): New mode attribute.
10441 (vec_dup<mode>): Use it.
10443 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10446 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10447 frame if argument is passed on stack.
10449 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10452 * ree.c (combine_reaching_defs): Optimize also
10453 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10454 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10456 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10458 PR middle-end/83189
10459 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10461 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10463 PR middle-end/83718
10464 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10465 after they are computed.
10467 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10469 PR tree-optimization/83695
10470 * gimple-loop-linterchange.cc
10471 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10472 reset cached scev information after interchange.
10473 (pass_linterchange::execute): Remove call to scev_reset_htab.
10475 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10477 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10478 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10479 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10480 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10481 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10482 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10483 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10484 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10485 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10486 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10487 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10488 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10489 (V_lane_reg): Likewise.
10490 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10492 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10493 (vfmal_lane_low<mode>_intrinsic,
10494 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10495 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10496 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10497 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10498 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10499 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10501 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10503 * config/arm/arm-cpus.in (fp16fml): New feature.
10504 (ALL_SIMD): Add fp16fml.
10505 (armv8.2-a): Add fp16fml as an option.
10506 (armv8.3-a): Likewise.
10507 (armv8.4-a): Add fp16fml as part of fp16.
10508 * config/arm/arm.h (TARGET_FP16FML): Define.
10509 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10511 * config/arm/arm-modes.def (V2HF): Define.
10512 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10513 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10514 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10515 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10516 vfmsl_low, vfmsl_high): New set of builtins.
10517 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10518 (vfml_op): New code attribute.
10519 (VFMLHALVES): New int iterator.
10520 (VFML, VFMLSEL): New mode attributes.
10521 (V_reg): Define mapping for V2HF.
10522 (V_hi, V_lo): New mode attributes.
10523 (VF_constraint): Likewise.
10524 (vfml_half, vfml_half_selector): New int attributes.
10525 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10527 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10528 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10530 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10531 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10532 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10533 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10535 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10536 Document new effective target and option set.
10538 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10540 * config/arm/arm-cpus.in (armv8_4): New feature.
10541 (ARMv8_4a): New fgroup.
10542 (armv8.4-a): New arch.
10543 * config/arm/arm-tables.opt: Regenerate.
10544 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10545 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10546 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10547 Add matching rules for -march=armv8.4-a and extensions.
10548 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10550 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10553 * config/rx/rx.md (BW): New mode attribute.
10554 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10556 2018-01-11 Richard Biener <rguenther@suse.de>
10558 PR tree-optimization/83435
10559 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10560 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10561 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10563 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10564 Alan Hayward <alan.hayward@arm.com>
10565 David Sherwood <david.sherwood@arm.com>
10567 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10569 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10570 (aarch64_print_address_internal): Use it to check for a zero offset.
10572 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10573 Alan Hayward <alan.hayward@arm.com>
10574 David Sherwood <david.sherwood@arm.com>
10576 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10577 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10578 Return a poly_int64 rather than a HOST_WIDE_INT.
10579 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10580 rather than a HOST_WIDE_INT.
10581 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10582 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10583 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10584 final_offset from HOST_WIDE_INT to poly_int64.
10585 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10586 to_constant when getting the number of units in an Advanced SIMD
10588 (aarch64_builtin_vectorized_function): Check for a constant number
10590 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10592 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10593 attribute instead of GET_MODE_NUNITS.
10594 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10595 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10596 GET_MODE_SIZE for fixed-size registers.
10597 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10598 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10599 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10600 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10601 (aarch64_print_operand, aarch64_print_address_internal)
10602 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10603 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10604 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10605 Handle polynomial GET_MODE_SIZE.
10606 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10607 wider than SImode without modification.
10608 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10609 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10610 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10611 passing and returning SVE modes.
10612 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10613 rather than GEN_INT.
10614 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10615 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10616 (aarch64_allocate_and_probe_stack_space): Likewise.
10617 (aarch64_layout_frame): Cope with polynomial offsets.
10618 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10619 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10620 polynomial offsets.
10621 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10622 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10623 poly_int64 rather than a HOST_WIDE_INT.
10624 (aarch64_get_separate_components, aarch64_process_components)
10625 (aarch64_expand_prologue, aarch64_expand_epilogue)
10626 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10627 (aarch64_anchor_offset): New function, split out from...
10628 (aarch64_legitimize_address): ...here.
10629 (aarch64_builtin_vectorization_cost): Handle polynomial
10630 TYPE_VECTOR_SUBPARTS.
10631 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10633 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10634 number of elements from the PARALLEL rather than the mode.
10635 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10636 rather than GET_MODE_BITSIZE.
10637 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10638 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10639 (aarch64_expand_vec_perm_const_1): Handle polynomial
10640 d->perm.length () and d->perm elements.
10641 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10642 Apply to_constant to d->perm elements.
10643 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10644 polynomial CONST_VECTOR_NUNITS.
10645 (aarch64_move_pointer): Take amount as a poly_int64 rather
10647 (aarch64_progress_pointer): Avoid temporary variable.
10648 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10649 the mode attribute instead of GET_MODE.
10651 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10652 Alan Hayward <alan.hayward@arm.com>
10653 David Sherwood <david.sherwood@arm.com>
10655 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10656 x exists before using it.
10657 (aarch64_add_constant_internal): Rename to...
10658 (aarch64_add_offset_1): ...this. Replace regnum with separate
10659 src and dest rtxes. Handle the case in which they're different,
10660 including when the offset is zero. Replace scratchreg with an rtx.
10661 Use 2 additions if there is no spare register into which we can
10662 move a 16-bit constant.
10663 (aarch64_add_constant): Delete.
10664 (aarch64_add_offset): Replace reg with separate src and dest
10665 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10666 Use aarch64_add_offset_1.
10667 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10668 an rtx rather than an int. Take the delta as a poly_int64
10669 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10670 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10671 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10672 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10673 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10674 and aarch64_add_sp.
10675 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10676 aarch64_add_constant.
10678 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10680 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10681 Use scalar_float_mode.
10683 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10685 * config/aarch64/aarch64-simd.md
10686 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10687 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10688 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10689 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10690 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10691 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10692 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10693 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10694 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10695 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10697 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10700 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10701 targ_options->x_arm_arch_string is non NULL.
10703 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10705 * config/aarch64/aarch64.h
10706 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10708 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10711 * expmed.c (emit_store_flag_force): Swap if const op0
10712 and change VOIDmode to mode of op0.
10714 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10716 PR rtl-optimization/83761
10717 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10718 than bytes to mode_for_size.
10720 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10722 PR middle-end/83189
10723 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10724 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10727 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10729 PR middle-end/83575
10730 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10731 when in layout mode.
10732 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10733 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10736 2018-01-10 Michael Collison <michael.collison@arm.com>
10738 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10739 * config/aarch64/aarch64-option-extension.def: Add
10740 AARCH64_OPT_EXTENSION of 'fp16fml'.
10741 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10742 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10743 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10744 * config/aarch64/constraints.md (Ui7): New constraint.
10745 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10746 (VFMLA_SEL_W): Ditto.
10749 (VFMLA16_LOW): New int iterator.
10750 (VFMLA16_HIGH): Ditto.
10751 (UNSPEC_FMLAL): New unspec.
10752 (UNSPEC_FMLSL): Ditto.
10753 (UNSPEC_FMLAL2): Ditto.
10754 (UNSPEC_FMLSL2): Ditto.
10755 (f16mac): New code attribute.
10756 * config/aarch64/aarch64-simd-builtins.def
10757 (aarch64_fmlal_lowv2sf): Ditto.
10758 (aarch64_fmlsl_lowv2sf): Ditto.
10759 (aarch64_fmlalq_lowv4sf): Ditto.
10760 (aarch64_fmlslq_lowv4sf): Ditto.
10761 (aarch64_fmlal_highv2sf): Ditto.
10762 (aarch64_fmlsl_highv2sf): Ditto.
10763 (aarch64_fmlalq_highv4sf): Ditto.
10764 (aarch64_fmlslq_highv4sf): Ditto.
10765 (aarch64_fmlal_lane_lowv2sf): Ditto.
10766 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10767 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10768 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10769 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10770 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10771 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10772 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10773 (aarch64_fmlal_lane_highv2sf): Ditto.
10774 (aarch64_fmlsl_lane_highv2sf): Ditto.
10775 (aarch64_fmlal_laneq_highv2sf): Ditto.
10776 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10777 (aarch64_fmlalq_lane_highv4sf): Ditto.
10778 (aarch64_fmlsl_lane_highv4sf): Ditto.
10779 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10780 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10781 * config/aarch64/aarch64-simd.md:
10782 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10783 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10784 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10785 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10786 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10787 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10788 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10789 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10790 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10791 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10792 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10793 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10794 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10795 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10796 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10797 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10798 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10799 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10800 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10801 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10802 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10803 (vfmlsl_low_u32): Ditto.
10804 (vfmlalq_low_u32): Ditto.
10805 (vfmlslq_low_u32): Ditto.
10806 (vfmlal_high_u32): Ditto.
10807 (vfmlsl_high_u32): Ditto.
10808 (vfmlalq_high_u32): Ditto.
10809 (vfmlslq_high_u32): Ditto.
10810 (vfmlal_lane_low_u32): Ditto.
10811 (vfmlsl_lane_low_u32): Ditto.
10812 (vfmlal_laneq_low_u32): Ditto.
10813 (vfmlsl_laneq_low_u32): Ditto.
10814 (vfmlalq_lane_low_u32): Ditto.
10815 (vfmlslq_lane_low_u32): Ditto.
10816 (vfmlalq_laneq_low_u32): Ditto.
10817 (vfmlslq_laneq_low_u32): Ditto.
10818 (vfmlal_lane_high_u32): Ditto.
10819 (vfmlsl_lane_high_u32): Ditto.
10820 (vfmlal_laneq_high_u32): Ditto.
10821 (vfmlsl_laneq_high_u32): Ditto.
10822 (vfmlalq_lane_high_u32): Ditto.
10823 (vfmlslq_lane_high_u32): Ditto.
10824 (vfmlalq_laneq_high_u32): Ditto.
10825 (vfmlslq_laneq_high_u32): Ditto.
10826 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10827 (AARCH64_FL_FOR_ARCH8_4): New.
10828 (AARCH64_ISA_F16FML): New ISA flag.
10829 (TARGET_F16FML): New feature flag for fp16fml.
10830 (doc/invoke.texi): Document new fp16fml option.
10832 2018-01-10 Michael Collison <michael.collison@arm.com>
10834 * config/aarch64/aarch64-builtins.c:
10835 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10836 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10837 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10838 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10839 (AARCH64_ISA_SHA3): New ISA flag.
10840 (TARGET_SHA3): New feature flag for sha3.
10841 * config/aarch64/iterators.md (sha512_op): New int attribute.
10842 (CRYPTO_SHA512): New int iterator.
10843 (UNSPEC_SHA512H): New unspec.
10844 (UNSPEC_SHA512H2): Ditto.
10845 (UNSPEC_SHA512SU0): Ditto.
10846 (UNSPEC_SHA512SU1): Ditto.
10847 * config/aarch64/aarch64-simd-builtins.def
10848 (aarch64_crypto_sha512hqv2di): New builtin.
10849 (aarch64_crypto_sha512h2qv2di): Ditto.
10850 (aarch64_crypto_sha512su0qv2di): Ditto.
10851 (aarch64_crypto_sha512su1qv2di): Ditto.
10852 (aarch64_eor3qv8hi): Ditto.
10853 (aarch64_rax1qv2di): Ditto.
10854 (aarch64_xarqv2di): Ditto.
10855 (aarch64_bcaxqv8hi): Ditto.
10856 * config/aarch64/aarch64-simd.md:
10857 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10858 (aarch64_crypto_sha512su0qv2di): Ditto.
10859 (aarch64_crypto_sha512su1qv2di): Ditto.
10860 (aarch64_eor3qv8hi): Ditto.
10861 (aarch64_rax1qv2di): Ditto.
10862 (aarch64_xarqv2di): Ditto.
10863 (aarch64_bcaxqv8hi): Ditto.
10864 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10865 (vsha512h2q_u64): Ditto.
10866 (vsha512su0q_u64): Ditto.
10867 (vsha512su1q_u64): Ditto.
10868 (veor3q_u16): Ditto.
10869 (vrax1q_u64): Ditto.
10870 (vxarq_u64): Ditto.
10871 (vbcaxq_u16): Ditto.
10872 * config/arm/types.md (crypto_sha512): New type attribute.
10873 (crypto_sha3): Ditto.
10874 (doc/invoke.texi): Document new sha3 option.
10876 2018-01-10 Michael Collison <michael.collison@arm.com>
10878 * config/aarch64/aarch64-builtins.c:
10879 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10880 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10881 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10882 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10883 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10884 (AARCH64_ISA_SM4): New ISA flag.
10885 (TARGET_SM4): New feature flag for sm4.
10886 * config/aarch64/aarch64-simd-builtins.def
10887 (aarch64_sm3ss1qv4si): Ditto.
10888 (aarch64_sm3tt1aq4si): Ditto.
10889 (aarch64_sm3tt1bq4si): Ditto.
10890 (aarch64_sm3tt2aq4si): Ditto.
10891 (aarch64_sm3tt2bq4si): Ditto.
10892 (aarch64_sm3partw1qv4si): Ditto.
10893 (aarch64_sm3partw2qv4si): Ditto.
10894 (aarch64_sm4eqv4si): Ditto.
10895 (aarch64_sm4ekeyqv4si): Ditto.
10896 * config/aarch64/aarch64-simd.md:
10897 (aarch64_sm3ss1qv4si): Ditto.
10898 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10899 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10900 (aarch64_sm4eqv4si): Ditto.
10901 (aarch64_sm4ekeyqv4si): Ditto.
10902 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10903 (sm3part_op): Ditto.
10904 (CRYPTO_SM3TT): Ditto.
10905 (CRYPTO_SM3PART): Ditto.
10906 (UNSPEC_SM3SS1): New unspec.
10907 (UNSPEC_SM3TT1A): Ditto.
10908 (UNSPEC_SM3TT1B): Ditto.
10909 (UNSPEC_SM3TT2A): Ditto.
10910 (UNSPEC_SM3TT2B): Ditto.
10911 (UNSPEC_SM3PARTW1): Ditto.
10912 (UNSPEC_SM3PARTW2): Ditto.
10913 (UNSPEC_SM4E): Ditto.
10914 (UNSPEC_SM4EKEY): Ditto.
10915 * config/aarch64/constraints.md (Ui2): New constraint.
10916 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10917 * config/arm/types.md (crypto_sm3): New type attribute.
10918 (crypto_sm4): Ditto.
10919 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10920 (vsm3tt1aq_u32): Ditto.
10921 (vsm3tt1bq_u32): Ditto.
10922 (vsm3tt2aq_u32): Ditto.
10923 (vsm3tt2bq_u32): Ditto.
10924 (vsm3partw1q_u32): Ditto.
10925 (vsm3partw2q_u32): Ditto.
10926 (vsm4eq_u32): Ditto.
10927 (vsm4ekeyq_u32): Ditto.
10928 (doc/invoke.texi): Document new sm4 option.
10930 2018-01-10 Michael Collison <michael.collison@arm.com>
10932 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10933 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10934 (AARCH64_FL_FOR_ARCH8_4): New.
10935 (AARCH64_FL_V8_4): New flag.
10936 (doc/invoke.texi): Document new armv8.4-a option.
10938 2018-01-10 Michael Collison <michael.collison@arm.com>
10940 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10941 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10942 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10943 * config/aarch64/aarch64-option-extension.def: Add
10944 AARCH64_OPT_EXTENSION of 'sha2'.
10945 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10946 (crypto): Disable sha2 and aes if crypto disabled.
10947 (crypto): Enable aes and sha2 if enabled.
10948 (simd): Disable sha2 and aes if simd disabled.
10949 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10951 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10952 (TARGET_SHA2): New feature flag for sha2.
10953 (TARGET_AES): New feature flag for aes.
10954 * config/aarch64/aarch64-simd.md:
10955 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10956 conditional on TARGET_AES.
10957 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10958 (aarch64_crypto_sha1hsi): Make pattern conditional
10960 (aarch64_crypto_sha1hv4si): Ditto.
10961 (aarch64_be_crypto_sha1hv4si): Ditto.
10962 (aarch64_crypto_sha1su1v4si): Ditto.
10963 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10964 (aarch64_crypto_sha1su0v4si): Ditto.
10965 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10966 (aarch64_crypto_sha256su0v4si): Ditto.
10967 (aarch64_crypto_sha256su1v4si): Ditto.
10968 (doc/invoke.texi): Document new aes and sha2 options.
10970 2018-01-10 Martin Sebor <msebor@redhat.com>
10972 PR tree-optimization/83781
10973 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10976 2018-01-11 Martin Sebor <msebor@gmail.com>
10977 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10979 PR tree-optimization/83501
10980 PR tree-optimization/81703
10982 * tree-ssa-strlen.c (get_string_cst): Rename...
10983 (get_string_len): ...to this. Handle global constants.
10984 (handle_char_store): Adjust.
10986 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10987 Jim Wilson <jimw@sifive.com>
10989 * config/riscv/riscv-protos.h (riscv_output_return): New.
10990 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10991 (riscv_attribute_table, riscv_output_return),
10992 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10993 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10994 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10996 (riscv_expand_prologue): Add early return for naked function.
10997 (riscv_expand_epilogue): Likewise.
10998 (riscv_function_ok_for_sibcall): Return false for naked function.
10999 (riscv_set_current_function): New.
11000 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11001 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11002 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11003 * doc/extend.texi (RISC-V Function Attributes): New.
11005 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11007 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11008 check for 128-bit long double before checking TCmode.
11009 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11010 128-bit long doubles before checking TFmode or TCmode.
11011 (FLOAT128_IBM_P): Likewise.
11013 2018-01-10 Martin Sebor <msebor@redhat.com>
11015 PR tree-optimization/83671
11016 * builtins.c (c_strlen): Unconditionally return zero for the empty
11018 Use -Warray-bounds for warnings.
11019 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11020 for non-constant array indices with COMPONENT_REF, arrays of
11021 arrays, and pointers to arrays.
11022 (gimple_fold_builtin_strlen): Determine and set length range for
11023 non-constant character arrays.
11025 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11027 PR middle-end/81897
11028 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11031 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11033 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11035 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11038 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11039 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11040 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11041 indexed_or_indirect_operand predicate.
11042 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11043 (*vsx_le_perm_load_v8hi): Likewise.
11044 (*vsx_le_perm_load_v16qi): Likewise.
11045 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11046 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11047 (*vsx_le_perm_store_v8hi): Likewise.
11048 (*vsx_le_perm_store_v16qi): Likewise.
11049 (eight unnamed splitters): Likewise.
11051 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11053 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11054 * config/rs6000/emmintrin.h: Likewise.
11055 * config/rs6000/mmintrin.h: Likewise.
11056 * config/rs6000/xmmintrin.h: Likewise.
11058 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11061 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11063 * tree.c (tree_nop_conversion): Return true for location wrapper
11065 (maybe_wrap_with_location): New function.
11066 (selftest::check_strip_nops): New function.
11067 (selftest::test_location_wrappers): New function.
11068 (selftest::tree_c_tests): Call it.
11069 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11070 (maybe_wrap_with_location): New decl.
11071 (EXPR_LOCATION_WRAPPER_P): New macro.
11072 (location_wrapper_p): New inline function.
11073 (tree_strip_any_location_wrapper): New inline function.
11075 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11078 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11079 stack_realign_offset for the largest alignment of stack slot
11081 (ix86_find_max_used_stack_alignment): New function.
11082 (ix86_finalize_stack_frame_flags): Use it. Set
11083 max_used_stack_alignment if we don't realign stack.
11084 * config/i386/i386.h (machine_function): Add
11085 max_used_stack_alignment.
11087 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11089 * config/arm/arm.opt (-mbranch-cost): New option.
11090 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11093 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11096 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11097 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11099 2018-01-10 Richard Biener <rguenther@suse.de>
11102 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11103 early out so it also covers the case where we have a non-NULL
11106 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11108 PR tree-optimization/83753
11109 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11110 for non-strided grouped accesses if the number of elements is 1.
11112 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11115 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11116 * i386.h (TARGET_USE_GATHER): Define.
11117 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11119 2018-01-10 Martin Liska <mliska@suse.cz>
11122 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11123 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11125 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11126 CLEANUP_NO_PARTITIONING is not set.
11128 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11130 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11131 for vectors, as a partial revert of r254296.
11132 * rtl.h (const_vec_p): Delete.
11133 (const_vec_duplicate_p): Don't test for vector CONSTs.
11134 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11135 * expmed.c (make_tree): Likewise.
11138 * common.md (E, F): Use CONSTANT_P instead of checking for
11140 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11141 checking for CONST_VECTOR.
11143 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11145 PR middle-end/83575
11146 * predict.c (force_edge_cold): Handle in more sane way edges
11147 with no prediction.
11149 2018-01-09 Carl Love <cel@us.ibm.com>
11151 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11153 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11154 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11155 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11156 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11157 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11158 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11159 * config/rs6000/rs6000-protos.h: Add extern defition for
11160 rs6000_generate_float2_double_code.
11161 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11163 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11164 (float2_v2df): Add define_expand.
11166 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11169 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11170 op_mode in the force_to_mode call.
11172 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11174 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11175 instead of checking each element individually.
11176 (aarch64_evpc_uzp): Likewise.
11177 (aarch64_evpc_zip): Likewise.
11178 (aarch64_evpc_ext): Likewise.
11179 (aarch64_evpc_rev): Likewise.
11180 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11181 instead of checking each element individually. Return true without
11183 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11184 whether all selected elements come from the same input, instead of
11185 checking each element individually. Remove calls to gen_rtx_REG,
11186 start_sequence and end_sequence and instead assert that no rtl is
11189 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11191 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11192 order of HIGH and CONST checks.
11194 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11196 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11197 if the destination isn't an SSA_NAME.
11199 2018-01-09 Richard Biener <rguenther@suse.de>
11201 PR tree-optimization/83668
11202 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11204 (canonicalize_loop_form): ... here, renamed from ...
11205 (canonicalize_loop_closed_ssa_form): ... this and amended to
11206 swap successor edges for loop exit blocks to make us use
11207 the RPO order we need for initial schedule generation.
11209 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11211 PR tree-optimization/64811
11212 * match.pd: When optimizing comparisons with Inf, avoid
11213 introducing or losing exceptions from comparisons with NaN.
11215 2018-01-09 Martin Liska <mliska@suse.cz>
11218 * asan.c (shadow_mem_size): Add gcc_assert.
11220 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11222 Don't save registers in main().
11225 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11226 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11227 * config/avr/avr.c (avr_set_current_function): Don't error if
11228 naked, OS_task or OS_main are specified at the same time.
11229 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11231 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11233 * common/config/avr/avr-common.c (avr_option_optimization_table):
11234 Switch on -mmain-is-OS_task for optimizing compilations.
11236 2018-01-09 Richard Biener <rguenther@suse.de>
11238 PR tree-optimization/83572
11239 * graphite.c: Include cfganal.h.
11240 (graphite_transform_loops): Connect infinite loops to exit
11241 and remove fake edges at the end.
11243 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11245 * ipa-inline.c (edge_badness): Revert accidental checkin.
11247 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11250 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11251 symbols; not inline clones.
11253 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11256 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11257 hard registers. Formatting fixes.
11259 PR preprocessor/83722
11260 * gcc.c (try_generate_repro): Pass
11261 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11262 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11265 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11266 Kito Cheng <kito.cheng@gmail.com>
11268 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11269 (riscv_leaf_function_p): Delete.
11270 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11272 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11274 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11276 (do_ifelse): New function.
11277 (do_isel): New function.
11278 (do_sub3): New function.
11279 (do_add3): New function.
11280 (do_load_mask_compare): New function.
11281 (do_overlap_load_compare): New function.
11282 (expand_compare_loop): New function.
11283 (expand_block_compare): Call expand_compare_loop() when appropriate.
11284 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11285 option description.
11286 (-mblock-compare-inline-loop-limit): New option.
11288 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11291 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11292 Reverse order of second and third operands in first alternative.
11293 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11294 of first and second elements in UNSPEC_VPERMR vector.
11295 (altivec_expand_vec_perm_le): Likewise.
11297 2018-01-08 Jeff Law <law@redhat.com>
11299 PR rtl-optimizatin/81308
11300 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11301 (process_switch): If group_case_labels makes a change, then set
11303 (pass_convert_switch::execute): If a switch is converted, then
11304 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11306 PR rtl-optimization/81308
11307 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11310 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11312 PR target/83663 - Revert r255946
11313 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11314 generation for cases where splatting a value is not useful.
11315 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11316 across a vec_duplicate and a paradoxical subreg forming a vector
11317 mode to a vec_concat.
11319 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11321 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11322 -march=armv8.3-a variants.
11323 * config/arm/t-multilib: Likewise.
11324 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11326 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11328 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11330 (cceq_ior_compare_complement): Give it a name so I can use it, and
11331 change boolean_or_operator predicate to boolean_operator so it can
11332 be used to generate a crand.
11333 (eqne): New code iterator.
11334 (bd/bd_neg): New code_attrs.
11335 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11336 a single define_insn.
11337 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11338 decrement (bdnzt/bdnzf/bdzt/bdzf).
11339 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11340 with the new names of the branch decrement patterns, and added the
11341 names of the branch decrement conditional patterns.
11343 2018-01-08 Richard Biener <rguenther@suse.de>
11345 PR tree-optimization/83563
11346 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11349 2018-01-08 Richard Biener <rguenther@suse.de>
11351 PR middle-end/83713
11352 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11354 2018-01-08 Richard Biener <rguenther@suse.de>
11356 PR tree-optimization/83685
11357 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11358 references to abnormals.
11360 2018-01-08 Richard Biener <rguenther@suse.de>
11363 * dwarf2out.c (output_indirect_strings): Handle empty
11364 skeleton_debug_str_hash.
11365 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11367 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11369 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11370 (emit_store_direct): Likewise.
11371 (arc_trampoline_adjust_address): Likewise.
11372 (arc_asm_trampoline_template): New function.
11373 (arc_initialize_trampoline): Use asm_trampoline_template.
11374 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11375 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11376 * config/arc/arc.md (flush_icache): Delete pattern.
11378 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11380 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11381 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11384 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11387 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11388 by not USED_FOR_TARGET.
11389 (make_pass_resolve_sw_modes): Likewise.
11391 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11393 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11396 2018-01-08 Richard Biener <rguenther@suse.de>
11398 PR middle-end/83580
11399 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11401 2018-01-08 Richard Biener <rguenther@suse.de>
11403 PR middle-end/83517
11404 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11406 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11408 PR middle-end/81897
11409 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11410 basic blocks with a small number of successors.
11411 (convert_control_dep_chain_into_preds): Improve handling of
11413 (dump_predicates): Split apart into...
11414 (dump_pred_chain): ...here...
11415 (dump_pred_info): ...and here.
11416 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11417 (can_chain_union_be_invalidated_p): Improve check for invalidation
11419 (uninit_uses_cannot_happen): Avoid unnecessary if
11420 convert_control_dep_chain_into_preds yielded nothing.
11422 2018-01-06 Martin Sebor <msebor@redhat.com>
11424 PR tree-optimization/83640
11425 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11426 subtracting negative offset from size.
11427 (builtin_access::overlap): Adjust offset bounds of the access to fall
11428 within the size of the object if possible.
11430 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11432 PR rtl-optimization/83699
11433 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11434 extract_bit_field_as_subreg to cases in which the extracted
11435 value is also a vector.
11437 * lra-constraints.c (process_alt_operands): Test for the equivalence
11438 substitutions when detecting a possible reload cycle.
11440 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11443 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11444 by default if flag_selective_schedling{,2}. Formatting fixes.
11446 PR rtl-optimization/83682
11447 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11448 if it has non-VECTOR_MODE element mode.
11449 (vec_duplicate_p): Likewise.
11451 PR middle-end/83694
11452 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11453 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11455 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11458 * config/i386/i386-builtin.def
11459 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11460 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11461 Require also OPTION_MASK_ISA_AVX512F in addition to
11462 OPTION_MASK_ISA_GFNI.
11463 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11464 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11465 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11466 to OPTION_MASK_ISA_GFNI.
11467 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11468 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11469 OPTION_MASK_ISA_AVX512BW.
11470 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11471 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11472 addition to OPTION_MASK_ISA_GFNI.
11473 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11474 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11475 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11476 to OPTION_MASK_ISA_GFNI.
11477 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11478 a requirement for all ISAs rather than any of them with a few
11480 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11482 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11483 bitmasks to be enabled with 3 exceptions, instead of requiring any
11484 enabled ISA with lots of exceptions.
11485 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11486 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11487 Change avx512bw in isa attribute to avx512f.
11488 * config/i386/sgxintrin.h: Add license boilerplate.
11489 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11490 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11491 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11492 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11494 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11495 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11496 temporarily sse2 rather than sse if not enabled already.
11499 * config/i386/sse.md (VI248_VLBW): Rename to ...
11500 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11501 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11502 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11503 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11504 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11505 mode iterator instead of VI248_VLBW.
11507 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11509 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11510 (record_modified): Skip clobbers; add debug output.
11511 (param_change_prob): Use sreal frequencies.
11513 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11515 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11516 punt for user-aligned variables.
11518 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11520 * tree-chrec.c (chrec_contains_symbols): Return true for
11523 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11526 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11527 of (x|y) == x for BICS pattern.
11529 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11531 PR tree-optimization/83605
11532 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11533 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11536 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11538 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11539 * config/epiphany/rtems.h: New file.
11541 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11542 Uros Bizjak <ubizjak@gmail.com>
11545 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11546 QIreg_operand instead of register_operand predicate.
11547 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11548 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11549 comments instead of -fmitigate[-_]rop.
11551 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11554 * cgraphunit.c (symbol_table::compile): Switch to text_section
11555 before calling assembly_start debug hook.
11556 * run-rtl-passes.c (run_rtl_passes): Likewise.
11559 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11561 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11562 range_int_cst_p rather than !symbolic_range_p before calling
11563 extract_range_from_multiplicative_op_1.
11565 2018-01-04 Jeff Law <law@redhat.com>
11567 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11568 redundant test in assertion.
11570 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11572 * doc/rtl.texi: Document machine_mode wrapper classes.
11574 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11576 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11577 using tree_to_uhwi.
11579 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11581 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11582 the VEC_PERM_EXPR fold to fail.
11584 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11587 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11588 to switched_sections.
11590 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11593 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11594 test for d.testing.
11596 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11599 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11600 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11602 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11605 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11606 is BLKmode and bitpos not zero or mode change is needed.
11608 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11611 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11614 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11617 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11618 instead of MULT rtx. Update all corresponding splitters.
11619 (*saddl_se): Ditto.
11620 (*ssub<modesuffix>): Ditto.
11621 (*ssubl_se): Ditto.
11622 (*cmp_sadd_di): Update split patterns.
11623 (*cmp_sadd_si): Ditto.
11624 (*cmp_sadd_sidi): Ditto.
11625 (*cmp_ssub_di): Ditto.
11626 (*cmp_ssub_si): Ditto.
11627 (*cmp_ssub_sidi): Ditto.
11628 * config/alpha/predicates.md (const23_operand): New predicate.
11629 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11630 Look for ASHIFT, not MULT inner operand.
11631 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11633 2018-01-04 Martin Liska <mliska@suse.cz>
11635 PR gcov-profile/83669
11636 * gcov.c (output_intermediate_file): Add version to intermediate
11638 * doc/gcov.texi: Document new field 'version' in intermediate
11639 file format. Fix location of '-k' option of gcov command.
11641 2018-01-04 Martin Liska <mliska@suse.cz>
11644 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11646 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11648 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11650 2018-01-03 Martin Sebor <msebor@redhat.com>
11652 PR tree-optimization/83655
11653 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11654 checking calls with invalid arguments.
11656 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11658 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11659 (vectorizable_mask_load_store): Delete.
11660 (vectorizable_call): Return false for masked loads and stores.
11661 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11662 instead of gimple_assign_rhs1.
11663 (vectorizable_load): Handle IFN_MASK_LOAD.
11664 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11668 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11670 (vectorizable_mask_load_store): ...here.
11671 (vectorizable_load): ...and here.
11673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11675 * tree-vect-stmts.c (vect_build_all_ones_mask)
11676 (vect_build_zero_merge_argument): New functions, split out from...
11677 (vectorizable_load): ...here.
11679 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11681 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11683 (vectorizable_mask_load_store): ...here.
11684 (vectorizable_store): ...and here.
11686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11688 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11690 (vectorizable_mask_load_store): ...here.
11692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11694 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11695 (vect_model_store_cost): Take a vec_load_store_type instead of a
11697 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11698 (vect_model_store_cost): Take a vec_load_store_type instead of a
11700 (vectorizable_mask_load_store): Update accordingly.
11701 (vectorizable_store): Likewise.
11702 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11706 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11707 IFN_MASK_LOAD calls here rather than...
11708 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11711 Alan Hayward <alan.hayward@arm.com>
11712 David Sherwood <david.sherwood@arm.com>
11714 * expmed.c (extract_bit_field_1): For vector extracts,
11715 fall back to extract_bit_field_as_subreg if vec_extract
11718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11719 Alan Hayward <alan.hayward@arm.com>
11720 David Sherwood <david.sherwood@arm.com>
11722 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11723 they are variable or constant sized.
11724 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11725 slots for constant-sized data.
11727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11728 Alan Hayward <alan.hayward@arm.com>
11729 David Sherwood <david.sherwood@arm.com>
11731 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11732 handling COND_EXPRs with boolean comparisons, try to find a better
11733 basis for the mask type than the boolean itself.
11735 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11737 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11738 is calculated and how it can be overridden.
11739 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11740 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11742 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11746 Alan Hayward <alan.hayward@arm.com>
11747 David Sherwood <david.sherwood@arm.com>
11749 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11750 Remove the mode argument.
11751 (aarch64_simd_valid_immediate): Remove the mode and inverse
11753 * config/aarch64/iterators.md (bitsize): New iterator.
11754 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11755 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11756 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11757 aarch64_simd_valid_immediate.
11758 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11759 (aarch64_reg_or_bic_imm): Likewise.
11760 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11761 with an insn_type enum and msl with a modifier_type enum.
11762 Replace element_width with a scalar_mode. Change the shift
11763 to unsigned int. Add constructors for scalar_float_mode and
11764 scalar_int_mode elements.
11765 (aarch64_vect_float_const_representable_p): Delete.
11766 (aarch64_can_const_movi_rtx_p)
11767 (aarch64_simd_scalar_immediate_valid_for_move)
11768 (aarch64_simd_make_constant): Update call to
11769 aarch64_simd_valid_immediate.
11770 (aarch64_advsimd_valid_immediate_hs): New function.
11771 (aarch64_advsimd_valid_immediate): Likewise.
11772 (aarch64_simd_valid_immediate): Remove mode and inverse
11773 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11774 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11775 and aarch64_float_const_representable_p on the result.
11776 (aarch64_output_simd_mov_immediate): Remove mode argument.
11777 Update call to aarch64_simd_valid_immediate and use of
11778 simd_immediate_info.
11779 (aarch64_output_scalar_simd_mov_immediate): Update call
11782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11783 Alan Hayward <alan.hayward@arm.com>
11784 David Sherwood <david.sherwood@arm.com>
11786 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11787 (mode_nunits): Likewise CONST_MODE_NUNITS.
11788 * machmode.def (ADJUST_NUNITS): Document.
11789 * genmodes.c (mode_data::need_nunits_adj): New field.
11790 (blank_mode): Update accordingly.
11791 (adj_nunits): New variable.
11792 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11794 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11795 listed in adj_nunits.
11796 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11797 listed in adj_nunits. Don't emit case statements for such modes.
11798 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11799 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11800 nothing if adj_nunits is nonnull.
11801 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11802 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11803 (emit_mode_fbit): Update use of print_maybe_const_decl.
11804 (emit_move_size): Likewise. Treat the array as non-const
11806 (emit_mode_adjustments): Handle adj_nunits.
11808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11810 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11811 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11812 (VECTOR_MODES): Use it.
11813 (make_vector_modes): Take the prefix as an argument.
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 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11820 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11821 for MODE_VECTOR_BOOL.
11822 * machmode.def (VECTOR_BOOL_MODE): Document.
11823 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11824 (make_vector_bool_mode): New function.
11825 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11827 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11828 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11830 * stor-layout.c (int_mode_for_mode): Likewise.
11831 * tree.c (build_vector_type_for_mode): Likewise.
11832 * varasm.c (output_constant_pool_2): Likewise.
11833 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11834 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11835 for MODE_VECTOR_BOOL.
11836 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11837 of mode class checks.
11838 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11839 instead of a list of mode class checks.
11840 (expand_vector_scalar_condition): Likewise.
11841 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11843 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11844 Alan Hayward <alan.hayward@arm.com>
11845 David Sherwood <david.sherwood@arm.com>
11847 * machmode.h (mode_size): Change from unsigned short to
11849 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11850 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11851 or if measurement_type is not polynomial.
11852 (fixed_size_mode::includes_p): Check for constant-sized modes.
11853 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11854 return a poly_uint16 rather than an unsigned short.
11855 (emit_mode_size): Change the type of mode_size from unsigned short
11856 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11857 (emit_mode_adjustments): Cope with polynomial vector sizes.
11858 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11860 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11862 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11863 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11864 * caller-save.c (setup_save_areas): Likewise.
11865 (replace_reg_with_saved_mem): Likewise.
11866 * calls.c (emit_library_call_value_1): Likewise.
11867 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11868 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11869 (gen_lowpart_for_combine): Likewise.
11870 * convert.c (convert_to_integer_1): Likewise.
11871 * cse.c (equiv_constant, cse_insn): Likewise.
11872 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11873 (cselib_subst_to_values): Likewise.
11874 * dce.c (word_dce_process_block): Likewise.
11875 * df-problems.c (df_word_lr_mark_ref): Likewise.
11876 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11877 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11878 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11879 (rtl_for_decl_location): Likewise.
11880 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11881 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11882 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11883 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11884 (expand_expr_real_1): Likewise.
11885 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11886 (pad_below): Likewise.
11887 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11888 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11889 * ira.c (get_subreg_tracking_sizes): Likewise.
11890 * ira-build.c (ira_create_allocno_objects): Likewise.
11891 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11892 (ira_sort_regnos_for_alter_reg): Likewise.
11893 * ira-costs.c (record_operand_costs): Likewise.
11894 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11895 (resolve_simple_move): Likewise.
11896 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11897 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11898 (lra_constraints): Likewise.
11899 (CONST_POOL_OK_P): Reject variable-sized modes.
11900 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11901 (add_pseudo_to_slot, lra_spill): Likewise.
11902 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11903 * optabs-query.c (get_best_extraction_insn): Likewise.
11904 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11905 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11906 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11907 * recog.c (offsettable_address_addr_space_p): Likewise.
11908 * regcprop.c (maybe_mode_change): Likewise.
11909 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11910 * regrename.c (build_def_use): Likewise.
11911 * regstat.c (dump_reg_info): Likewise.
11912 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11913 (find_reloads, find_reloads_subreg_address): Likewise.
11914 * reload1.c (eliminate_regs_1): Likewise.
11915 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11916 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11917 (simplify_binary_operation_1, simplify_subreg): Likewise.
11918 * targhooks.c (default_function_arg_padding): Likewise.
11919 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11920 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11921 (verify_gimple_assign_ternary): Likewise.
11922 * tree-inline.c (estimate_move_cost): Likewise.
11923 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11924 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11925 (get_address_cost_ainc): Likewise.
11926 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11927 (vect_supportable_dr_alignment): Likewise.
11928 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11929 (vectorizable_reduction): Likewise.
11930 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11931 (vectorizable_operation, vectorizable_load): Likewise.
11932 * tree.c (build_same_sized_truth_vector_type): Likewise.
11933 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11934 * var-tracking.c (emit_note_insn_var_location): Likewise.
11935 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11936 (ADDR_VEC_ALIGN): Likewise.
11938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11939 Alan Hayward <alan.hayward@arm.com>
11940 David Sherwood <david.sherwood@arm.com>
11942 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11944 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11945 or if measurement_type is polynomial.
11946 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11947 * combine.c (make_extraction): Likewise.
11948 * dse.c (find_shift_sequence): Likewise.
11949 * dwarf2out.c (mem_loc_descriptor): Likewise.
11950 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11951 (extract_bit_field, extract_low_bits): Likewise.
11952 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11953 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11954 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11955 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11956 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11957 * reload.c (find_reloads): Likewise.
11958 * reload1.c (alter_reg): Likewise.
11959 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11960 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11961 * tree-if-conv.c (predicate_mem_writes): Likewise.
11962 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11963 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11964 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11965 * valtrack.c (dead_debug_insert_temp): Likewise.
11966 * varasm.c (mergeable_constant_section): Likewise.
11967 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11969 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11970 Alan Hayward <alan.hayward@arm.com>
11971 David Sherwood <david.sherwood@arm.com>
11973 * expr.c (expand_assignment): Cope with polynomial mode sizes
11974 when assigning to a CONCAT.
11976 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11977 Alan Hayward <alan.hayward@arm.com>
11978 David Sherwood <david.sherwood@arm.com>
11980 * machmode.h (mode_precision): Change from unsigned short to
11982 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11984 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11985 or if measurement_type is not polynomial.
11986 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11987 in which the mode is already known to be a scalar_int_mode.
11988 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11989 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11991 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11992 for GET_MODE_PRECISION.
11993 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11994 for GET_MODE_PRECISION.
11995 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11997 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11998 (expand_field_assignment, make_extraction): Likewise.
11999 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12000 (get_last_value): Likewise.
12001 * convert.c (convert_to_integer_1): Likewise.
12002 * cse.c (cse_insn): Likewise.
12003 * expr.c (expand_expr_real_1): Likewise.
12004 * lra-constraints.c (simplify_operand_subreg): Likewise.
12005 * optabs-query.c (can_atomic_load_p): Likewise.
12006 * optabs.c (expand_atomic_load): Likewise.
12007 (expand_atomic_store): Likewise.
12008 * ree.c (combine_reaching_defs): Likewise.
12009 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12010 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12011 * tree.h (type_has_mode_precision_p): Likewise.
12012 * ubsan.c (instrument_si_overflow): Likewise.
12014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12015 Alan Hayward <alan.hayward@arm.com>
12016 David Sherwood <david.sherwood@arm.com>
12018 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12019 polynomial numbers of units.
12020 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12021 (valid_vector_subparts_p): New function.
12022 (build_vector_type): Remove temporary shim and take the number
12023 of units as a poly_uint64 rather than an int.
12024 (build_opaque_vector_type): Take the number of units as a
12025 poly_uint64 rather than an int.
12026 * tree.c (build_vector_from_ctor): Handle polynomial
12027 TYPE_VECTOR_SUBPARTS.
12028 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12029 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12030 (build_vector_from_val): If the number of units is variable,
12031 use build_vec_duplicate_cst for constant operands and
12032 VEC_DUPLICATE_EXPR otherwise.
12033 (make_vector_type): Remove temporary is_constant ().
12034 (build_vector_type, build_opaque_vector_type): Take the number of
12035 units as a poly_uint64 rather than an int.
12036 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12038 * cfgexpand.c (expand_debug_expr): Likewise.
12039 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12040 (store_constructor, expand_expr_real_1): Likewise.
12041 (const_scalar_mask_from_tree): Likewise.
12042 * fold-const-call.c (fold_const_reduction): Likewise.
12043 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12044 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12045 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12046 (fold_relational_const): Likewise.
12047 (native_interpret_vector): Likewise. Change the size from an
12048 int to an unsigned int.
12049 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12050 TYPE_VECTOR_SUBPARTS.
12051 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12052 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12053 duplicating a non-constant operand into a variable-length vector.
12054 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12055 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12056 * ipa-icf.c (sem_variable::equals): Likewise.
12057 * match.pd: Likewise.
12058 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12059 * print-tree.c (print_node): Likewise.
12060 * stor-layout.c (layout_type): Likewise.
12061 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12062 * tree-cfg.c (verify_gimple_comparison): Likewise.
12063 (verify_gimple_assign_binary): Likewise.
12064 (verify_gimple_assign_ternary): Likewise.
12065 (verify_gimple_assign_single): Likewise.
12066 * tree-pretty-print.c (dump_generic_node): Likewise.
12067 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12068 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12069 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12070 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12071 (vect_shift_permute_load_chain): Likewise.
12072 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12073 (expand_vector_condition, optimize_vector_constructor): Likewise.
12074 (lower_vec_perm, get_compute_type): Likewise.
12075 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12076 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12077 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12078 (vect_recog_mask_conversion_pattern): Likewise.
12079 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12080 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12081 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12082 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12083 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12084 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12085 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12086 (supportable_widening_operation): Likewise.
12087 (supportable_narrowing_operation): Likewise.
12088 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12090 * varasm.c (output_constant): Likewise.
12092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12093 Alan Hayward <alan.hayward@arm.com>
12094 David Sherwood <david.sherwood@arm.com>
12096 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12097 so that both the length == 3 and length != 3 cases set up their
12098 own permute vectors. Add comments explaining why we know the
12099 number of elements is constant.
12100 (vect_permute_load_chain): Likewise.
12102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12103 Alan Hayward <alan.hayward@arm.com>
12104 David Sherwood <david.sherwood@arm.com>
12106 * machmode.h (mode_nunits): Change from unsigned char to
12108 (ONLY_FIXED_SIZE_MODES): New macro.
12109 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12110 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12111 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12113 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12114 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12115 or if measurement_type is not polynomial.
12116 * genmodes.c (ZERO_COEFFS): New macro.
12117 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12119 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12120 Use ZERO_COEFFS when emitting initializers.
12121 * data-streamer.h (bp_pack_poly_value): New function.
12122 (bp_unpack_poly_value): Likewise.
12123 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12124 for GET_MODE_NUNITS.
12125 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12126 for GET_MODE_NUNITS.
12127 * tree.c (make_vector_type): Remove temporary shim and make
12128 the real function take the number of units as a poly_uint64
12129 rather than an int.
12130 (build_vector_type_for_mode): Handle polynomial nunits.
12131 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12132 * emit-rtl.c (const_vec_series_p_1): Likewise.
12133 (gen_rtx_CONST_VECTOR): Likewise.
12134 * fold-const.c (test_vec_duplicate_folding): Likewise.
12135 * genrecog.c (validate_pattern): Likewise.
12136 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12137 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12138 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12139 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12140 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12141 * rtlanal.c (subreg_get_info): Likewise.
12142 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12143 (vect_grouped_load_supported): Likewise.
12144 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12145 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12146 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12147 (simplify_const_unary_operation, simplify_binary_operation_1)
12148 (simplify_const_binary_operation, simplify_ternary_operation)
12149 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12150 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12151 instead of CONST_VECTOR_NUNITS.
12152 * varasm.c (output_constant_pool_2): Likewise.
12153 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12154 explicit-encoded elements in the XVEC for variable-length vectors.
12156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12158 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12160 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12161 Alan Hayward <alan.hayward@arm.com>
12162 David Sherwood <david.sherwood@arm.com>
12164 * coretypes.h (fixed_size_mode): Declare.
12165 (fixed_size_mode_pod): New typedef.
12166 * builtins.h (target_builtins::x_apply_args_mode)
12167 (target_builtins::x_apply_result_mode): Change type to
12168 fixed_size_mode_pod.
12169 * builtins.c (apply_args_size, apply_result_size, result_vector)
12170 (expand_builtin_apply_args_1, expand_builtin_apply)
12171 (expand_builtin_return): Update accordingly.
12173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12175 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12176 * cselib.c (cselib_hash_rtx): Likewise.
12177 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12178 CONST_VECTOR encoding.
12180 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12181 Jeff Law <law@redhat.com>
12184 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12185 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12186 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12187 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12190 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12191 explicitly probe *sp in a noreturn function if there were any callee
12192 register saves or frame pointer is needed.
12194 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12197 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12198 BLKmode for ternary, binary or unary expressions.
12201 * var-tracking.c (delete_vta_debug_insn): New inline function.
12202 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12203 insns from get_insns () to NULL instead of each bb separately.
12204 Use delete_vta_debug_insn. No longer static.
12205 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12206 delete_vta_debug_insns callers.
12207 * rtl.h (delete_vta_debug_insns): Declare.
12208 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12209 instead of variable_tracking_main.
12211 2018-01-03 Martin Sebor <msebor@redhat.com>
12213 PR tree-optimization/83603
12214 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12215 arguments past the endof the argument list in functions declared
12216 without a prototype.
12217 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12218 Avoid checking when arguments are null.
12220 2018-01-03 Martin Sebor <msebor@redhat.com>
12223 * doc/extend.texi (attribute const): Fix a typo.
12224 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12225 issuing -Wsuggest-attribute for void functions.
12227 2018-01-03 Martin Sebor <msebor@redhat.com>
12229 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12230 offset_int::from instead of wide_int::to_shwi.
12231 (maybe_diag_overlap): Remove assertion.
12232 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12233 * gimple-ssa-sprintf.c (format_directive): Same.
12234 (parse_directive): Same.
12235 (sprintf_dom_walker::compute_format_length): Same.
12236 (try_substitute_return_value): Same.
12238 2018-01-03 Jeff Law <law@redhat.com>
12240 PR middle-end/83654
12241 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12242 non-constant residual for zero at runtime and avoid probing in
12243 that case. Reorganize code for trailing problem to mirror handling
12246 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12248 PR tree-optimization/83501
12249 * tree-ssa-strlen.c (get_string_cst): New.
12250 (handle_char_store): Call get_string_cst.
12252 2018-01-03 Martin Liska <mliska@suse.cz>
12254 PR tree-optimization/83593
12255 * tree-ssa-strlen.c: Include tree-cfg.h.
12256 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12257 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12258 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12260 (strlen_dom_walker::before_dom_children): Call
12261 gimple_purge_dead_eh_edges. Dump tranformation with details
12263 (strlen_dom_walker::before_dom_children): Update call by adding
12264 new argument cleanup_eh.
12265 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12267 2018-01-03 Martin Liska <mliska@suse.cz>
12270 * cif-code.def (VARIADIC_THUNK): New enum value.
12271 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12274 2018-01-03 Jan Beulich <jbeulich@suse.com>
12276 * sse.md (mov<mode>_internal): Tighten condition for when to use
12277 vmovdqu<ssescalarsize> for TI and OI modes.
12279 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12281 Update copyright years.
12283 2018-01-03 Martin Liska <mliska@suse.cz>
12286 * ipa-visibility.c (function_and_variable_visibility): Skip
12287 functions with noipa attribure.
12289 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12291 * gcc.c (process_command): Update copyright notice dates.
12292 * gcov-dump.c (print_version): Ditto.
12293 * gcov.c (print_version): Ditto.
12294 * gcov-tool.c (print_version): Ditto.
12295 * gengtype.c (create_file): Ditto.
12296 * doc/cpp.texi: Bump @copying's copyright year.
12297 * doc/cppinternals.texi: Ditto.
12298 * doc/gcc.texi: Ditto.
12299 * doc/gccint.texi: Ditto.
12300 * doc/gcov.texi: Ditto.
12301 * doc/install.texi: Ditto.
12302 * doc/invoke.texi: Ditto.
12304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12306 * vector-builder.h (vector_builder::m_full_nelts): Change from
12307 unsigned int to poly_uint64.
12308 (vector_builder::full_nelts): Update prototype accordingly.
12309 (vector_builder::new_vector): Likewise.
12310 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12311 (vector_builder::operator ==): Likewise.
12312 (vector_builder::finalize): Likewise.
12313 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12314 Take the number of elements as a poly_uint64 rather than an
12316 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12317 from unsigned int to poly_uint64.
12318 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12319 (vec_perm_indices::new_vector): Likewise.
12320 (vec_perm_indices::length): Likewise.
12321 (vec_perm_indices::nelts_per_input): Likewise.
12322 (vec_perm_indices::input_nelts): Likewise.
12323 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12324 number of elements per input as a poly_uint64 rather than an
12325 unsigned int. Use the original encoding for variable-length
12326 vectors, rather than clamping each individual element.
12327 For the second and subsequent elements in each pattern,
12328 clamp the step and base before clamping their sum.
12329 (vec_perm_indices::series_p): Handle polynomial element counts.
12330 (vec_perm_indices::all_in_range_p): Likewise.
12331 (vec_perm_indices_to_tree): Likewise.
12332 (vec_perm_indices_to_rtx): Likewise.
12333 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12334 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12335 (tree_vector_builder::new_binary_operation): Handle polynomial
12336 element counts. Return false if we need to know the number
12337 of elements at compile time.
12338 * fold-const.c (fold_vec_perm): Punt if the number of elements
12339 isn't known at compile time.
12341 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12343 * vec-perm-indices.h (vec_perm_builder): Change element type
12344 from HOST_WIDE_INT to poly_int64.
12345 (vec_perm_indices::element_type): Update accordingly.
12346 (vec_perm_indices::clamp): Handle polynomial element_types.
12347 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12348 (vec_perm_indices::all_in_range_p): Likewise.
12349 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12351 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12352 polynomial vec_perm_indices element types.
12353 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12354 * fold-const.c (fold_vec_perm): Likewise.
12355 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12356 * tree-vect-generic.c (lower_vec_perm): Likewise.
12357 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12358 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12359 element type to HOST_WIDE_INT.
12361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12362 Alan Hayward <alan.hayward@arm.com>
12363 David Sherwood <david.sherwood@arm.com>
12365 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12366 rather than an int. Use plus_constant.
12367 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12368 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12371 Alan Hayward <alan.hayward@arm.com>
12372 David Sherwood <david.sherwood@arm.com>
12374 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12375 a HOST_WIDE_INT to a poly_int64.
12377 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12378 Alan Hayward <alan.hayward@arm.com>
12379 David Sherwood <david.sherwood@arm.com>
12381 * calls.c (load_register_parameters): Cope with polynomial
12382 mode sizes. Require a constant size for BLKmode parameters
12383 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12384 forces a parameter to be padded at the lsb end in order to
12385 fill a complete number of words, require the parameter size
12386 to be ordered wrt UNITS_PER_WORD.
12388 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12389 Alan Hayward <alan.hayward@arm.com>
12390 David Sherwood <david.sherwood@arm.com>
12392 * reload1.c (spill_stack_slot_width): Change element type
12393 from unsigned int to poly_uint64_pod.
12394 (alter_reg): Treat mode sizes as polynomial.
12396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12397 Alan Hayward <alan.hayward@arm.com>
12398 David Sherwood <david.sherwood@arm.com>
12400 * reload.c (complex_word_subreg_p): New function.
12401 (reload_inner_reg_of_subreg, push_reload): Use it.
12403 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12404 Alan Hayward <alan.hayward@arm.com>
12405 David Sherwood <david.sherwood@arm.com>
12407 * lra-constraints.c (process_alt_operands): Reject matched
12408 operands whose sizes aren't ordered.
12409 (match_reload): Refer to this check here.
12411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12412 Alan Hayward <alan.hayward@arm.com>
12413 David Sherwood <david.sherwood@arm.com>
12415 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12416 that the mode size is in the set {1, 2, 4, 8, 16}.
12418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12419 Alan Hayward <alan.hayward@arm.com>
12420 David Sherwood <david.sherwood@arm.com>
12422 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12423 Use plus_constant instead of gen_rtx_PLUS.
12425 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12426 Alan Hayward <alan.hayward@arm.com>
12427 David Sherwood <david.sherwood@arm.com>
12429 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12430 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12431 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12432 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12433 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12434 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12435 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12436 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12437 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12438 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12440 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12441 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12442 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12443 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12444 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12445 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12446 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12447 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12448 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12449 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12451 * expr.c (emit_move_resolve_push): Treat the input and result
12452 of PUSH_ROUNDING as a poly_int64.
12453 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12454 (emit_push_insn): Likewise.
12455 * lra-eliminations.c (mark_not_eliminable): Likewise.
12456 * recog.c (push_operand): Likewise.
12457 * reload1.c (elimination_effects): Likewise.
12458 * rtlanal.c (nonzero_bits1): Likewise.
12459 * calls.c (store_one_arg): Likewise. Require the padding to be
12460 known at compile time.
12462 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12463 Alan Hayward <alan.hayward@arm.com>
12464 David Sherwood <david.sherwood@arm.com>
12466 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12467 Use plus_constant instead of gen_rtx_PLUS.
12469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12470 Alan Hayward <alan.hayward@arm.com>
12471 David Sherwood <david.sherwood@arm.com>
12473 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12474 rather than an int.
12476 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12477 Alan Hayward <alan.hayward@arm.com>
12478 David Sherwood <david.sherwood@arm.com>
12480 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12481 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12482 via stack temporaries. Treat the mode size as polynomial too.
12484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12485 Alan Hayward <alan.hayward@arm.com>
12486 David Sherwood <david.sherwood@arm.com>
12488 * expr.c (expand_expr_real_2): When handling conversions involving
12489 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12490 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12491 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12499 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12500 Alan Hayward <alan.hayward@arm.com>
12501 David Sherwood <david.sherwood@arm.com>
12503 * combine.c (can_change_dest_mode): Handle polynomial
12504 REGMODE_NATURAL_SIZE.
12505 * expmed.c (store_bit_field_1): Likewise.
12506 * expr.c (store_constructor): Likewise.
12507 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12508 and polynomial REGMODE_NATURAL_SIZE.
12509 (gen_lowpart_common): Likewise.
12510 * reginfo.c (record_subregs_of_mode): Likewise.
12511 * rtlanal.c (read_modify_subreg_p): Likewise.
12513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12514 Alan Hayward <alan.hayward@arm.com>
12515 David Sherwood <david.sherwood@arm.com>
12517 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12518 numbers of elements.
12520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12521 Alan Hayward <alan.hayward@arm.com>
12522 David Sherwood <david.sherwood@arm.com>
12524 * match.pd: Cope with polynomial numbers of vector elements.
12526 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12527 Alan Hayward <alan.hayward@arm.com>
12528 David Sherwood <david.sherwood@arm.com>
12530 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12531 in a POINTER_PLUS_EXPR.
12533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12534 Alan Hayward <alan.hayward@arm.com>
12535 David Sherwood <david.sherwood@arm.com>
12537 * omp-simd-clone.c (simd_clone_subparts): New function.
12538 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12539 (ipa_simd_modify_function_body): Likewise.
12541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12542 Alan Hayward <alan.hayward@arm.com>
12543 David Sherwood <david.sherwood@arm.com>
12545 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12546 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12547 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12548 (expand_vector_condition, vector_element): Likewise.
12549 (subparts_gt): New function.
12550 (get_compute_type): Use subparts_gt.
12551 (count_type_subparts): Delete.
12552 (expand_vector_operations_1): Use subparts_gt instead of
12553 count_type_subparts.
12555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12556 Alan Hayward <alan.hayward@arm.com>
12557 David Sherwood <david.sherwood@arm.com>
12559 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12560 (vect_compile_time_alias): ...this new function. Do the calculation
12561 on poly_ints rather than trees.
12562 (vect_prune_runtime_alias_test_list): Update call accordingly.
12564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12565 Alan Hayward <alan.hayward@arm.com>
12566 David Sherwood <david.sherwood@arm.com>
12568 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12570 (vect_schedule_slp_instance): Likewise.
12572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12573 Alan Hayward <alan.hayward@arm.com>
12574 David Sherwood <david.sherwood@arm.com>
12576 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12577 constant and extern definitions for variable-length vectors.
12578 (vect_get_constant_vectors): Note that the number of units
12579 is known to be constant.
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 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12586 of units as polynomial. Choose between WIDE and NARROW based
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 * tree-vect-stmts.c (simd_clone_subparts): New function.
12594 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12597 Alan Hayward <alan.hayward@arm.com>
12598 David Sherwood <david.sherwood@arm.com>
12600 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12601 vectors as polynomial. Use build_index_vector for
12602 IFN_GOMP_SIMD_LANE.
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 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12609 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12610 for variable-length vectors.
12611 (vectorizable_mask_load_store): Treat the number of units as
12612 polynomial, asserting that it is constant if the condition has
12613 already been enforced.
12614 (vectorizable_store, vectorizable_load): Likewise.
12616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12617 Alan Hayward <alan.hayward@arm.com>
12618 David Sherwood <david.sherwood@arm.com>
12620 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12621 of units as polynomial. Punt if we can't tell at compile time
12622 which vector contains the final result.
12624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12625 Alan Hayward <alan.hayward@arm.com>
12626 David Sherwood <david.sherwood@arm.com>
12628 * tree-vect-loop.c (vectorizable_induction): Treat the number
12629 of units as polynomial. Punt on SLP inductions. Use an integer
12630 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12631 cast of such a series for variable-length floating-point
12634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12635 Alan Hayward <alan.hayward@arm.com>
12636 David Sherwood <david.sherwood@arm.com>
12638 * tree.h (build_index_vector): Declare.
12639 * tree.c (build_index_vector): New function.
12640 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12641 of units as polynomial, forcibly converting it to a constant if
12642 vectorizable_reduction has already enforced the condition.
12643 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12644 to create a {1,2,3,...} vector.
12645 (vectorizable_reduction): Treat the number of units as polynomial.
12646 Choose vectype_in based on the largest scalar element size rather
12647 than the smallest number of units. Enforce the restrictions
12650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12651 Alan Hayward <alan.hayward@arm.com>
12652 David Sherwood <david.sherwood@arm.com>
12654 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12655 number of units as polynomial.
12657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12658 Alan Hayward <alan.hayward@arm.com>
12659 David Sherwood <david.sherwood@arm.com>
12661 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12662 * target.def (autovectorize_vector_sizes): Return the vector sizes
12663 by pointer, using vector_sizes rather than a bitmask.
12664 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12665 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12666 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12668 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12669 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12670 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12671 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12672 * omp-general.c (omp_max_vf): Likewise.
12673 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12674 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12675 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12676 * tree-vect-slp.c (vect_slp_bb): Likewise.
12677 * doc/tm.texi: Regenerate.
12678 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12680 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12681 the vector size as a poly_uint64 rather than an unsigned int.
12682 (current_vector_size): Change from an unsigned int to a poly_uint64.
12683 (get_vectype_for_scalar_type): Update accordingly.
12684 * tree.h (build_truth_vector_type): Take the size and number of
12685 units as a poly_uint64 rather than an unsigned int.
12686 (build_vector_type): Add a temporary overload that takes
12687 the number of units as a poly_uint64 rather than an unsigned int.
12688 * tree.c (make_vector_type): Likewise.
12689 (build_truth_vector_type): Take the number of units as a poly_uint64
12690 rather than an unsigned int.
12692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12693 Alan Hayward <alan.hayward@arm.com>
12694 David Sherwood <david.sherwood@arm.com>
12696 * target.def (get_mask_mode): Take the number of units and length
12697 as poly_uint64s rather than unsigned ints.
12698 * targhooks.h (default_get_mask_mode): Update accordingly.
12699 * targhooks.c (default_get_mask_mode): Likewise.
12700 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12701 * doc/tm.texi: Regenerate.
12703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12704 Alan Hayward <alan.hayward@arm.com>
12705 David Sherwood <david.sherwood@arm.com>
12707 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12708 * omp-general.c (omp_max_vf): Likewise.
12709 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12710 (expand_omp_simd): Handle polynomial safelen.
12711 * omp-low.c (omplow_simd_context): Add a default constructor.
12712 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12713 (lower_rec_simd_input_clauses): Update accordingly.
12714 (lower_rec_input_clauses): Likewise.
12716 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12717 Alan Hayward <alan.hayward@arm.com>
12718 David Sherwood <david.sherwood@arm.com>
12720 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12721 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12722 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12723 (vect_analyze_slp_cost): Likewise.
12724 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12725 (vect_model_load_cost): Likewise.
12727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12728 Alan Hayward <alan.hayward@arm.com>
12729 David Sherwood <david.sherwood@arm.com>
12731 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12732 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12733 from an unsigned int * to a poly_uint64_pod *.
12734 (calculate_unrolling_factor): New function.
12735 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12738 Alan Hayward <alan.hayward@arm.com>
12739 David Sherwood <david.sherwood@arm.com>
12741 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12742 from an unsigned int to a poly_uint64.
12743 (_loop_vec_info::slp_unrolling_factor): Likewise.
12744 (_loop_vec_info::vectorization_factor): Change from an int
12746 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12747 (vect_get_num_vectors): New function.
12748 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12749 (vect_get_num_copies): Use vect_get_num_vectors.
12750 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12751 to an unsigned int *.
12752 (vect_analyze_data_refs): Change min_vf from an int * to a
12754 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12755 than an unsigned HOST_WIDE_INT.
12756 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12757 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12758 to an unsigned int *.
12759 (vect_analyze_data_ref_dependences): Likewise.
12760 (vect_compute_data_ref_alignment): Handle polynomial vf.
12761 (vect_enhance_data_refs_alignment): Likewise.
12762 (vect_prune_runtime_alias_test_list): Likewise.
12763 (vect_shift_permute_load_chain): Likewise.
12764 (vect_supportable_dr_alignment): Likewise.
12765 (dependence_distance_ge_vf): Take the vectorization factor as a
12766 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12767 (vect_analyze_data_refs): Change min_vf from an int * to a
12769 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12770 vfm1 as a poly_uint64 rather than an int. Make the same change
12771 for the returned bound_scalar.
12772 (vect_gen_vector_loop_niters): Handle polynomial vf.
12773 (vect_do_peeling): Likewise. Update call to
12774 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12775 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12777 * tree-vect-loop.c (vect_determine_vectorization_factor)
12778 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12779 (vect_get_known_peeling_cost): Likewise.
12780 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12781 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12782 (vect_transform_loop): Likewise. Use the lowest possible VF when
12783 updating the upper bounds of the loop.
12784 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12785 rather than an int.
12786 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12787 polynomial unroll factors.
12788 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12789 (vect_make_slp_decision): Likewise.
12790 (vect_supported_load_permutation_p): Likewise, and polynomial
12792 (vect_analyze_slp_cost): Handle polynomial vf.
12793 (vect_slp_analyze_node_operations): Likewise.
12794 (vect_slp_analyze_bb_1): Likewise.
12795 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12796 than an unsigned HOST_WIDE_INT.
12797 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12798 (vectorizable_load): Handle polynomial vf.
12799 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12801 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12804 Alan Hayward <alan.hayward@arm.com>
12805 David Sherwood <david.sherwood@arm.com>
12807 * match.pd: Handle bit operations involving three constants
12808 and try to fold one pair.
12810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12812 * tree-vect-loop-manip.c: Include gimple-fold.h.
12813 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12814 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12815 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12816 Add a path that uses a step of VF instead of 1, but disable it
12818 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12819 and niters_no_overflow parameters. Update calls to
12820 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12821 Create a new SSA name if the latter choses to use a ste other
12822 than zero, and return it via niters_vector_mult_vf_var.
12823 * tree-vect-loop.c (vect_transform_loop): Update calls to
12824 vect_do_peeling, vect_gen_vector_loop_niters and
12825 slpeel_make_loop_iterate_ntimes.
12826 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12827 (vect_gen_vector_loop_niters): Update declarations after above changes.
12829 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12831 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12832 128-bit round to integer instructions.
12833 (ceil<mode>2): Likewise.
12834 (btrunc<mode>2): Likewise.
12835 (round<mode>2): Likewise.
12837 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12839 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12840 unaligned VSX load/store on P8/P9.
12841 (expand_block_clear): Allow the use of unaligned VSX
12842 load/store on P8/P9.
12844 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12846 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12848 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12849 swap associated with both a load and a store.
12851 2018-01-02 Andrew Waterman <andrew@sifive.com>
12853 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12854 * config/riscv/riscv.md (clear_cache): Use it.
12856 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12858 * web.c: Remove out-of-date comment.
12860 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12862 * expr.c (fixup_args_size_notes): Check that any existing
12863 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12864 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12865 (emit_single_push_insn): ...here.
12867 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12869 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12870 (const_vector_encoded_nelts): New function.
12871 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12872 (const_vector_int_elt, const_vector_elt): Declare.
12873 * emit-rtl.c (const_vector_int_elt_1): New function.
12874 (const_vector_elt): Likewise.
12875 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12876 of CONST_VECTOR_ELT.
12878 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12880 * expr.c: Include rtx-vector-builder.h.
12881 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12882 directly on the tree encoding.
12883 (const_vector_from_tree): Likewise.
12884 * optabs.c: Include rtx-vector-builder.h.
12885 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12886 sequence of "u" values.
12887 * vec-perm-indices.c: Include rtx-vector-builder.h.
12888 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12889 directly on the vec_perm_indices encoding.
12891 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12893 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12894 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12895 * rtx-vector-builder.h: New file.
12896 * rtx-vector-builder.c: Likewise.
12897 * rtl.h (rtx_def::u2): Add a const_vector field.
12898 (CONST_VECTOR_NPATTERNS): New macro.
12899 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12900 (CONST_VECTOR_DUPLICATE_P): Likewise.
12901 (CONST_VECTOR_STEPPED_P): Likewise.
12902 (CONST_VECTOR_ENCODED_ELT): Likewise.
12903 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12904 (unwrap_const_vec_duplicate): Likewise.
12905 (const_vec_series_p): Check for a non-duplicated vector encoding.
12906 Say that the function only returns true for integer vectors.
12907 * emit-rtl.c: Include rtx-vector-builder.h.
12908 (gen_const_vec_duplicate_1): Delete.
12909 (gen_const_vector): Call gen_const_vec_duplicate instead of
12910 gen_const_vec_duplicate_1.
12911 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12912 (gen_const_vec_duplicate): Use rtx_vector_builder.
12913 (gen_const_vec_series): Likewise.
12914 (gen_rtx_CONST_VECTOR): Likewise.
12915 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12916 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12917 Build a new vector rather than modifying a CONST_VECTOR in-place.
12918 (handle_special_swappables): Update call accordingly.
12919 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12920 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12921 Build a new vector rather than modifying a CONST_VECTOR in-place.
12922 (handle_special_swappables): Update call accordingly.
12924 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12926 * simplify-rtx.c (simplify_const_binary_operation): Use
12927 CONST_VECTOR_ELT instead of XVECEXP.
12929 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12931 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12932 the selector elements to be different from the data elements
12933 if the selector is a VECTOR_CST.
12934 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12935 ssizetype for the selector.
12937 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12939 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12940 before testing each element individually.
12941 * tree-vect-generic.c (lower_vec_perm): Likewise.
12943 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12945 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12946 * selftest-run-tests.c (selftest::run_tests): Call it.
12947 * vector-builder.h (vector_builder::operator ==): New function.
12948 (vector_builder::operator !=): Likewise.
12949 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12950 (vec_perm_indices::all_from_input_p): New function.
12951 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12952 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12953 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12954 instead of reading the VECTOR_CST directly. Detect whether both
12955 vector inputs are the same before constructing the vec_perm_indices,
12956 and update the number of inputs argument accordingly. Use the
12957 utility functions added above. Only construct sel2 if we need to.
12959 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12961 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12962 the broadcast of the low byte.
12963 (expand_mult_highpart): Use an explicit encoding for the permutes.
12964 * optabs-query.c (can_mult_highpart_p): Likewise.
12965 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12966 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12967 (vectorizable_bswap): Likewise.
12968 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12969 explicit encoding for the power-of-2 permutes.
12970 (vect_permute_store_chain): Likewise.
12971 (vect_grouped_load_supported): Likewise.
12972 (vect_permute_load_chain): Likewise.
12974 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12976 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12977 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12978 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12979 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12980 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12981 (vect_gen_perm_mask_any): Likewise.
12983 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12985 * int-vector-builder.h: New file.
12986 * vec-perm-indices.h: Include int-vector-builder.h.
12987 (vec_perm_indices): Redefine as an int_vector_builder.
12988 (auto_vec_perm_indices): Delete.
12989 (vec_perm_builder): Redefine as a stand-alone class.
12990 (vec_perm_indices::vec_perm_indices): New function.
12991 (vec_perm_indices::clamp): Likewise.
12992 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12993 (vec_perm_indices::new_vector): New function.
12994 (vec_perm_indices::new_expanded_vector): Update for new
12995 vec_perm_indices class.
12996 (vec_perm_indices::rotate_inputs): New function.
12997 (vec_perm_indices::all_in_range_p): Operate directly on the
12998 encoded form, without computing elided elements.
12999 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13000 encoding. Update for new vec_perm_indices class.
13001 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13002 the given vec_perm_builder.
13003 (expand_vec_perm_var): Update vec_perm_builder constructor.
13004 (expand_mult_highpart): Use vec_perm_builder instead of
13005 auto_vec_perm_indices.
13006 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13007 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13008 or double series encoding as appropriate.
13009 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13010 vec_perm_indices instead of auto_vec_perm_indices.
13011 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13012 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13013 (vect_permute_store_chain): Likewise.
13014 (vect_grouped_load_supported): Likewise.
13015 (vect_permute_load_chain): Likewise.
13016 (vect_shift_permute_load_chain): Likewise.
13017 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13018 (vect_transform_slp_perm_load): Likewise.
13019 (vect_schedule_slp_instance): Likewise.
13020 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13021 (vectorizable_mask_load_store): Likewise.
13022 (vectorizable_bswap): Likewise.
13023 (vectorizable_store): Likewise.
13024 (vectorizable_load): Likewise.
13025 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13026 vec_perm_indices instead of auto_vec_perm_indices. Use
13027 tree_to_vec_perm_builder to read the vector from a tree.
13028 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13029 vec_perm_builder instead of a vec_perm_indices.
13030 (have_whole_vector_shift): Use vec_perm_builder and
13031 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13032 truncation to calc_vec_perm_mask_for_shift.
13033 (vect_create_epilog_for_reduction): Likewise.
13034 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13035 from auto_vec_perm_indices to vec_perm_indices.
13036 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13037 instead of changing individual elements.
13038 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13039 the vector in d.perm.
13040 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13041 from auto_vec_perm_indices to vec_perm_indices.
13042 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13043 instead of changing individual elements.
13044 (arm_vectorize_vec_perm_const): Use new_vector to install
13045 the vector in d.perm.
13046 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13047 Update vec_perm_builder constructor.
13048 (rs6000_expand_interleave): Likewise.
13049 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13050 (rs6000_expand_interleave): Likewise.
13052 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13054 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13055 to qimode could truncate the indices.
13056 * optabs.c (expand_vec_perm_var): Likewise.
13058 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13060 * Makefile.in (OBJS): Add vec-perm-indices.o.
13061 * vec-perm-indices.h: New file.
13062 * vec-perm-indices.c: Likewise.
13063 * target.h (vec_perm_indices): Replace with a forward class
13065 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13066 * optabs.h: Include vec-perm-indices.h.
13067 (expand_vec_perm): Delete.
13068 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13069 (expand_vec_perm_const): Declare.
13070 * target.def (vec_perm_const_ok): Replace with...
13071 (vec_perm_const): ...this new hook.
13072 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13073 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13074 * doc/tm.texi: Regenerate.
13075 * optabs.def (vec_perm_const): Delete.
13076 * doc/md.texi (vec_perm_const): Likewise.
13077 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13078 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13079 expand_vec_perm for constant permutation vectors. Assert that
13080 the mode of variable permutation vectors is the integer equivalent
13081 of the mode that is being permuted.
13082 * optabs-query.h (selector_fits_mode_p): Declare.
13083 * optabs-query.c: Include vec-perm-indices.h.
13084 (selector_fits_mode_p): New function.
13085 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13086 is defined, instead of checking whether the vec_perm_const_optab
13087 exists. Use targetm.vectorize.vec_perm_const instead of
13088 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13089 fit in the vector mode before using a variable permute.
13090 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13091 vec_perm_indices instead of an rtx.
13092 (expand_vec_perm): Replace with...
13093 (expand_vec_perm_const): ...this new function. Take the selector
13094 as a vec_perm_indices rather than an rtx. Also take the mode of
13095 the selector. Update call to shift_amt_for_vec_perm_mask.
13096 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13097 Use vec_perm_indices::new_expanded_vector to expand the original
13098 selector into bytes. Check whether the indices fit in the vector
13099 mode before using a variable permute.
13100 (expand_vec_perm_var): Make global.
13101 (expand_mult_highpart): Use expand_vec_perm_const.
13102 * fold-const.c: Includes vec-perm-indices.h.
13103 * tree-ssa-forwprop.c: Likewise.
13104 * tree-vect-data-refs.c: Likewise.
13105 * tree-vect-generic.c: Likewise.
13106 * tree-vect-loop.c: Likewise.
13107 * tree-vect-slp.c: Likewise.
13108 * tree-vect-stmts.c: Likewise.
13109 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13111 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13112 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13113 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13114 (aarch64_vectorize_vec_perm_const): ...this new function.
13115 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13116 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13117 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13118 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13119 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13120 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13121 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13123 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13124 check for NEON modes.
13125 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13126 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13127 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13128 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13130 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13131 the old VEC_PERM_CONST conditions.
13132 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13133 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13134 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13135 (ia64_vectorize_vec_perm_const_ok): Merge into...
13136 (ia64_vectorize_vec_perm_const): ...this new function.
13137 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13138 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13139 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13140 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13141 * config/mips/mips.c (mips_expand_vec_perm_const)
13142 (mips_vectorize_vec_perm_const_ok): Merge into...
13143 (mips_vectorize_vec_perm_const): ...this new function.
13144 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13145 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13146 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13147 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13148 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13149 (rs6000_expand_vec_perm_const): Delete.
13150 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13152 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13153 (altivec_expand_vec_perm_const_le): Take each operand individually.
13154 Operate on constant selectors rather than rtxes.
13155 (altivec_expand_vec_perm_const): Likewise. Update call to
13156 altivec_expand_vec_perm_const_le.
13157 (rs6000_expand_vec_perm_const): Delete.
13158 (rs6000_vectorize_vec_perm_const_ok): Delete.
13159 (rs6000_vectorize_vec_perm_const): New function.
13160 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13161 an element count and rtx array.
13162 (rs6000_expand_extract_even): Update call accordingly.
13163 (rs6000_expand_interleave): Likewise.
13164 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13165 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13166 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13167 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13168 (rs6000_expand_vec_perm_const): Delete.
13169 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13170 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13171 (altivec_expand_vec_perm_const_le): Take each operand individually.
13172 Operate on constant selectors rather than rtxes.
13173 (altivec_expand_vec_perm_const): Likewise. Update call to
13174 altivec_expand_vec_perm_const_le.
13175 (rs6000_expand_vec_perm_const): Delete.
13176 (rs6000_vectorize_vec_perm_const_ok): Delete.
13177 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13178 reference to the SPE evmerge intructions.
13179 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13180 an element count and rtx array.
13181 (rs6000_expand_extract_even): Update call accordingly.
13182 (rs6000_expand_interleave): Likewise.
13183 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13184 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13186 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13188 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13190 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13191 vector mode and that that mode matches the mode of the data
13193 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13194 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13195 directly using expand_vec_perm_1 when forcing selectors into
13197 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13199 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13201 * optabs-query.h (can_vec_perm_p): Delete.
13202 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13203 * optabs-query.c (can_vec_perm_p): Split into...
13204 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13205 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13206 particular selector is valid.
13207 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13208 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13209 (vect_grouped_load_supported): Likewise.
13210 (vect_shift_permute_load_chain): Likewise.
13211 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13212 (vect_transform_slp_perm_load): Likewise.
13213 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13214 (vectorizable_bswap): Likewise.
13215 (vect_gen_perm_mask_checked): Likewise.
13216 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13217 implementations of variable permutation vectors into account
13218 when deciding which selector to use.
13219 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13220 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13221 with a false third argument.
13222 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13223 to test whether the constant selector is valid and can_vec_perm_var_p
13224 to test whether a variable selector is valid.
13226 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13228 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13229 * optabs-query.c (can_vec_perm_p): Likewise.
13230 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13231 instead of vec_perm_indices.
13232 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13233 (vect_gen_perm_mask_checked): Likewise,
13234 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13235 (vect_gen_perm_mask_checked): Likewise,
13237 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13239 * optabs-query.h (qimode_for_vec_perm): Declare.
13240 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13241 (qimode_for_vec_perm): ...this new function.
13242 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13244 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13246 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13247 does not have a conditional at the top.
13249 2018-01-02 Richard Biener <rguenther@suse.de>
13251 * ipa-inline.c (big_speedup_p): Fix expression.
13253 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13256 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13259 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13263 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13264 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13265 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13266 cond_taken_branch_cost 3->4.
13268 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13270 PR tree-optimization/83581
13271 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13272 TODO_cleanup_cfg if any changes have been made.
13274 PR middle-end/83608
13275 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13276 convert_modes if target mode has the right side, but different mode
13279 PR middle-end/83609
13280 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13281 last argument when extracting from CONCAT. If either from_real or
13282 from_imag is NULL, use expansion through memory. If result is not
13283 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13284 the parts directly to inner mode, if even that fails, use expansion
13287 PR middle-end/83623
13288 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13289 check for bswap in mode rather than HImode and use that in expand_unop
13292 Copyright (C) 2018 Free Software Foundation, Inc.
13294 Copying and distribution of this file, with or without modification,
13295 are permitted in any medium without royalty provided the copyright
13296 notice and this notice are preserved.