1 2018-05-18 Martin Liska <mliska@suse.cz>
3 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
4 * tree-pass.h (make_pass_lower_switch_O0): New function.
5 * tree-switch-conversion.c (node_has_low_bound): Remove.
6 (node_has_high_bound): Likewise.
7 (node_is_bounded): Likewise.
8 (class pass_lower_switch): Make it a template type and create
10 (pass_lower_switch::execute): Add template argument.
11 (make_pass_lower_switch): New function.
12 (make_pass_lower_switch_O0): New function.
13 (do_jump_if_equal): Remove.
14 (emit_case_nodes): Simplify to just handle all 3 cases and leave
15 all the hard work to tree optimization passes.
17 2018-05-18 Martin Liska <mliska@suse.cz>
19 * dbgcnt.c (limit_low): Renamed from limit.
20 (limit_high): New variable.
21 (dbg_cnt_is_enabled): Check for upper limit.
22 (dbg_cnt): Adjust dumping.
23 (dbg_cnt_set_limit_by_index): Add new argument for high
25 (dbg_cnt_set_limit_by_name): Likewise.
26 (dbg_cnt_process_single_pair): Parse new format.
27 (dbg_cnt_process_opt): Use strtok.
28 (dbg_cnt_list_all_counters): Remove 'value' and add
30 * doc/invoke.texi: Document changes.
32 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
34 * doc/sourcebuild.texi (scalar_all_fma): Document.
35 * tree.def (FMA_EXPR): Delete.
36 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
37 * internal-fn.c (ternary_direct): New macro.
38 (expand_ternary_optab_fn): Likewise.
39 (direct_ternary_optab_supported_p): Likewise.
40 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
41 * builtins.c (fold_builtin_fma): Delete.
42 (fold_builtin_3): Don't call it.
43 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
44 * expr.c (expand_expr_real_2): Likewise.
45 * fold-const.c (operand_equal_p): Likewise.
46 (fold_ternary_loc): Likewise.
47 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
48 * gimple.c (DEFTREECODE): Likewise.
49 * gimplify.c (gimplify_expr): Likewise.
50 * optabs-tree.c (optab_for_tree_code): Likewise.
51 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
52 * tree-eh.c (operation_could_trap_p): Likewise.
53 (stmt_could_throw_1_p): Likewise.
54 * tree-inline.c (estimate_operator_cost): Likewise.
55 * tree-pretty-print.c (dump_generic_node): Likewise.
56 (op_code_prio): Likewise.
57 * tree-ssa-loop-im.c (stmt_cost): Likewise.
58 * tree-ssa-operands.c (get_expr_operands): Likewise.
59 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
60 * fold-const-call.h (fold_fma): Delete.
61 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
62 CFN_FNMA and CFN_FNMS.
64 * genmatch.c (combined_fn): New enum.
65 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
66 (commutative_op): New function.
67 (commutate): Use it. Handle more than 2 operands.
68 (dt_operand::gen_gimple_expr): Use commutative_op.
69 (parser::parse_expr): Allow :c to be used with non-binary
70 operators if the commutative operand is known.
71 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
72 CFN_FMS, CFN_FNMA and CFN_FNMS.
73 (backprop::process_assign_use): Remove FMA_EXPR handling.
74 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
75 (gen_hsa_fma): New function.
76 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
77 IFN_FNMA and IFN_FNMS.
78 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
79 * gimple-fold.h (follow_all_ssa_edges): Declare.
80 * gimple-fold.c (follow_all_ssa_edges): New function.
81 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
82 gimple_build interface and use follow_all_ssa_edges to fold the result.
83 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
84 instead of checking for optabs directly.
85 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
86 rather than FMA_EXPRs.
87 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
88 call to IFN_FMA instead of an FMA_EXPR.
90 2018-05-17 Jim Wilson <jimw@sifive.com>
92 * expr.c (do_tablejump): When converting index to Pmode, if we have a
93 sign extended promoted subreg, and the range does not have the sign bit
94 set, then do a sign extend.
96 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
97 test, check for sign extended subreg and/or constant operands, and
98 do a sign extend in that case.
100 2018-05-17 Steve Ellcey <sellcey@cavium.com>
102 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
103 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
105 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
106 Change logics_shift_reg to logics_shift_imm.
107 (thunderx2t99_fp_loadpair_basic): Delete.
108 (thunderx2t99_fp_storepair_basic): Delete.
109 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
110 (thunderx2t99_asimd_polynomial): Delete.
111 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
112 and neon_fp_mul_d_scalar_q.
113 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
114 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
115 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
116 (thunderx2t99_asimd_lut): Add missing tbl types.
117 (thunderx2t99_asimd_ext): Delete.
118 (thunderx2t99_asimd_load1_1_mult): Delete.
119 (thunderx2t99_asimd_load1_2_mult): Delete.
120 (thunderx2t99_asimd_load1_ldp): New.
121 (thunderx2t99_asimd_load1): New.
122 (thunderx2t99_asimd_load2): Add missing *load2* types.
123 (thunderx2t99_asimd_load3): New.
124 (thunderx2t99_asimd_load4): New.
125 (thunderx2t99_asimd_store1_1_mult): Delete.
126 (thunderx2t99_asimd_store1_2_mult): Delete.
127 (thunderx2t99_asimd_store2_mult): Delete.
128 (thunderx2t99_asimd_store2_onelane): Delete.
129 (thunderx2t99_asimd_store_stp): New.
130 (thunderx2t99_asimd_store1): New.
131 (thunderx2t99_asimd_store2): New.
132 (thunderx2t99_asimd_store3): New.
133 (thunderx2t99_asimd_store4): New.
135 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
137 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
138 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
140 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
141 Segher Boessenkool <segher@kernel.crashing.org>
144 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
147 2018-05-17 Richard Biener <rguenther@suse.de>
149 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
150 for pruning loop and prune defs feeding only already visited PHIs.
152 2018-05-17 Richard Biener <rguenther@suse.de>
154 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
156 2018-05-17 Bin Cheng <bin.cheng@arm.com>
157 Richard Biener <rguenther@suse.de>
159 PR tree-optimization/85793
160 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
161 for VMAT_ELEMENTWISE.
163 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
165 * internal-fn.h (lookup_internal_fn): Declare
166 * internal-fn.c (lookup_internal_fn): New function.
167 * gimple.c (gimple_build_call_from_tree): Handle calls to
169 * gimple-pretty-print.c (dump_gimple_call): Print "." before
170 internal function names.
171 * tree-pretty-print.c (dump_generic_node): Likewise.
172 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
174 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
176 * gimple-fold.h (gimple_build): Make the function forms take
177 combined_fn rather than built_in_function.
178 (gimple_simplify): Likewise.
179 * gimple-match-head.c (gimple_simplify): Likewise.
180 * gimple-fold.c (gimple_build): Likewise.
181 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
182 rather than gimple_build_call_internal.
183 (get_initial_defs_for_reduction): Likewise.
184 (vect_create_epilog_for_reduction): Likewise.
185 (vectorizable_live_operation): Likewise.
187 2018-05-17 Martin Liska <mliska@suse.cz>
189 * gimple-ssa-sprintf.c (format_directive): Do not use
190 space in between 'G_' and '('.
192 2018-05-17 Jakub Jelinek <jakub@redhat.com>
195 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
196 even if the mask is not all ones.
199 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
201 (ix86_gimple_fold_builtin): Likewise.
204 * config/i386/i386.c: Include tree-vector-builder.h.
205 (ix86_vector_shift_count): New function.
206 (ix86_fold_builtin): Fold shift builtins by scalar count.
207 (ix86_gimple_fold_builtin): Likewise.
209 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
210 _mm512_setzero): New intrinsics.
212 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
213 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
215 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
216 code generation for cases where splatting a value is not useful.
217 * simplify-rtx.c (simplify_ternary_operation): Simplify
218 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
219 mode to a vec_concat.
221 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
223 * config.gcc: Support "goldmont-plus".
224 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
226 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
227 PROCESSOR_GOLDMONT_PLUS.
228 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
229 (processor_target_table): Add "goldmont-plus".
230 (PTA_GOLDMONT_PLUS): Define.
231 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
232 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
233 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
234 (fold_builtin_cpu): Add "goldmont-plus".
235 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
236 (ix86_option_override_internal): Add "goldmont-plus".
237 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
238 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
239 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
240 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
242 2018-05-17 Richard Biener <rguenther@suse.de>
244 PR tree-optimization/85757
245 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
246 remove defs that only feed that PHI from further processing.
248 2018-05-16 Jim Wilson <jimw@sifive.com>
250 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
252 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
254 2018-05-16 Mark Wielaard <mark@klomp.org>
256 * dwarf2out.c (count_index_strings): New function.
257 (output_indirect_strings): Call count_index_strings and generate
258 header for dwarf_version >= 5.
260 2018-05-16 Mark Wielaard <mark@klomp.org>
262 * dwarf2out.c (dwarf_FORM): New function.
263 (set_indirect_string): Use dwarf_FORM.
264 (reset_indirect_string): Likewise.
265 (size_of_die): Likewise.
266 (value_format): Likewise.
267 (output_die): Likewise.
268 (add_skeleton_AT_string): Likewise.
269 (output_macinfo_op): Likewise.
270 (index_string): Likewise.
271 (output_index_string_offset): Likewise.
272 (output_index_string): Likewise.
273 (count_index_strings): Likewise.
275 2018-05-16 Carl Love <cel@us.ibm.com>
277 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
278 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
280 2018-05-16 Martin Jambor <mjambor@suse.cz>
282 * ipa-prop.c (ipa_free_all_edge_args): Remove.
283 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
285 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
287 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
288 (fnma<mode>4): Likewise.
289 (fms<mode>4): Likewise.
290 (fnms<mode>4): Likewise.
291 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
292 (aarch64_fnma<mode>4): Likewise.
293 (aarch64_fms<mode>4): Likewise.
294 (aarch64_fnms<mode>4): Likewise.
295 (aarch64_fnmadd<mode>4): Likewise.
297 2018-05-16 Jason Merrill <jason@redhat.com>
299 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
301 2018-05-16 Richard Biener <rguenther@suse.de>
303 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
304 (dump_stmt_cost): Declare.
305 (add_stmt_cost): Dump cost we add.
306 (add_stmt_costs): New function.
307 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
309 (vect_analyze_stmt): Adjust prototype.
310 (vectorizable_condition): Likewise.
311 (vectorizable_live_operation): Likewise.
312 (vectorizable_reduction): Likewise.
313 (vectorizable_induction): Likewise.
314 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
315 cost vector to pass to vectorizable_ and record afterwards.
316 (vect_model_reduction_cost): Take cost vector argument and adjust.
317 (vect_model_induction_cost): Likewise.
318 (vectorizable_reduction): Likewise.
319 (vectorizable_induction): Likewise.
320 (vectorizable_live_operation): Likewise.
321 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
322 SLP_TREE_NUMBER_OF_VEC_STMTS.
323 (vect_analyze_slp_cost_1): Remove.
324 (vect_analyze_slp_cost): Likewise.
325 (vect_slp_analyze_node_operations): Take visited args and
326 a target cost vector. Avoid processing already visited stmt sets.
327 (vect_slp_analyze_operations): Use a local cost vector to gather
328 costs and register those of non-discarded instances.
329 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
330 (vect_schedule_slp_instance): Remove copying of
331 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
333 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
334 adding cost. Record cost entry location.
335 (vect_prologue_cost_for_slp_op): Function to compute cost of
336 a constant or invariant generated for SLP vect in the prologue,
337 split out from vect_analyze_slp_cost_1.
338 (vect_model_simple_cost): Make static. Adjust for SLP costing.
339 (vect_model_promotion_demotion_cost): Likewise.
340 (vect_model_store_cost): Likewise, make static.
341 (vect_model_load_cost): Likewise.
342 (vectorizable_bswap): Add cost vector arg and adjust.
343 (vectorizable_call): Likewise.
344 (vectorizable_simd_clone_call): Likewise.
345 (vectorizable_conversion): Likewise.
346 (vectorizable_assignment): Likewise.
347 (vectorizable_shift): Likewise.
348 (vectorizable_operation): Likewise.
349 (vectorizable_store): Likewise.
350 (vectorizable_load): Likewise.
351 (vectorizable_condition): Likewise.
352 (vectorizable_comparison): Likewise.
353 (can_vectorize_live_stmts): Likewise.
354 (vect_analyze_stmt): Likewise.
355 (vect_transform_stmt): Adjust calls to vectorizable_*.
356 * tree-vectorizer.c: Include gimple-pretty-print.h.
357 (dump_stmt_cost): New function.
359 2018-05-16 Richard Biener <rguenther@suse.de>
361 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
362 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
363 * tree-ssa-dse.c: Include tree-ssa-loop.h.
364 (check_name): New callback.
365 (dse_classify_store): Track cycles via a visited bitmap of PHI
366 defs and simplify handling of in-loop and across loop dead stores
367 and properly fail for loop-variant refs. Handle byte-tracking with
368 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
371 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
373 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
374 (vect_get_mask_type_for_stmt): Likewise.
375 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
377 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
378 to determine the statement's vector type and the vector type that
379 should be used for calculating nunits. Deal with cases in which
380 the type has to be deferred.
381 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
382 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
383 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
384 (vect_determine_vf_for_stmt): New functions, split out from...
385 (vect_determine_vectorization_factor): ...here.
386 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
387 (vect_get_mask_type_for_stmt): New functions, split out from
388 vect_determine_vectorization_factor.
390 2018-05-16 Richard Biener <rguenther@suse.de>
392 * tree-cfg.c (verify_gimple_assign_ternary): Properly
393 verify the [VEC_]COND_EXPR embedded comparison.
395 2018-05-15 Martin Sebor <msebor@redhat.com>
397 PR tree-optimization/85753
398 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
399 RECORD_TYPE in addition to ARRAY_TYPE.
401 2018-05-15 Martin Sebor <msebor@redhat.com>
404 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
406 2018-05-15 Richard Biener <rguenther@suse.de>
408 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
409 add by_clobber_p one. Change algorithm to collect all defs
410 representing uses we need to walk and try reducing them to
411 a single one before failing.
412 (dse_dom_walker::dse_optimize_stmt): Adjust.
414 2018-05-13 Mark Wielaard <mark@klomp.org>
416 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
417 (size_of_loc_descr): Likewise.
418 (output_loc_operands): Likewise.
419 (output_loc_operands_raw): Likewise.
420 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
421 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
422 (hash_loc_operands): Likewise.
423 (compare_loc_operands): Likewise.
425 2018-05-14 Mark Wielaard <mark@klomp.org>
427 * dwarf2out.c (count_index_addrs): New function.
428 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
430 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
432 PR tree-optimization/83648
433 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
434 return value as malloc candidate.
436 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
439 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
440 as true in call to suggest_attribute.
442 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
444 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
447 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
449 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
451 (aarch64_crypto_aesd_fused): Likewise.
453 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
455 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
456 (movsi_aarch64): Likewise.
457 (load_pairsi): Likewise.
458 (load_pairdi): Likewise.
459 (store_pairsi): Likewise.
460 (store_pairdi): Likewise.
461 (load_pairsf): Likewise.
462 (load_pairdf): Likewise.
463 (store_pairsf): Likewise.
464 (store_pairdf): Likewise.
465 (zero_extend): Likewise.
466 (trunc): Swap alternatives.
467 (fcvt_target): Add '?' to prefer w over r.
469 2018-05-14 Jakub Jelinek <jakub@redhat.com>
472 * config/i386/i386.md: Disallow non-commutative arithmetics in
473 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
474 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
475 in the peephole2 before it.
477 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
479 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
480 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
481 (ix86_handle_option): Handle -mcldemote.
482 * config.gcc: New header.
483 * config/i386/cldemoteintrin.h: New file.
484 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
485 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
487 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
488 OPTION_MASK_ISA_CLDEMOTE.
489 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
490 (ix86_valid_target_attribute_inner_p): Ditto.
491 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
492 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
493 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
494 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
495 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
497 * config/i386/i386.opt: Add -mcldemote.
498 * config/i386/x86intrin.h: New header.
499 * doc/invoke.texi: Add -mcldemote.
501 2018-05-14 Richard Biener <rguenther@suse.de>
503 * doc/match-and-simplify.texi: Adjust :s documentation.
505 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
507 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
508 intended memcpy size.
509 (REORDER_45): Likewise.
511 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
514 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
515 * vec.c (qsort_chk): Use gcc_qsort.
516 * Makefile.in (OBJS-libcommon): Add sort.o.
517 (build/sort.o): New target. Use it...
518 (BUILD_RTL): ... here, and...
519 (build/gencfn-macros): ... here, and...
520 (build/genmatch): ... here.
522 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
523 Chung-Ju Wu <jasonwucj@gmail.com>
525 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
526 * config/nds32/nds32-graywolf.md: New file.
527 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
528 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
530 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
531 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
532 * config/nds32/nds32.md (pipeline_model): Add graywolf.
533 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
534 * config/nds32/pipelines.md: Include n15 settings.
536 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
537 Chung-Ju Wu <jasonwucj@gmail.com>
539 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
540 * config/nds32/nds32-n13.md: New file.
541 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
542 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
544 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
545 * config/nds32/nds32.md (pipeline_model): Add n13.
546 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
547 * config/nds32/pipelines.md: Include n13 settings.
549 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
550 Chung-Ju Wu <jasonwucj@gmail.com>
552 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
553 * config/nds32/nds32-n10.md: New file.
554 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
555 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
557 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
558 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
559 * config/nds32/nds32.md (pipeline_model): Add n10.
560 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
561 * config/nds32/pipelines.md: Include n10 settings.
563 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
564 Kito Cheng <kito.cheng@gmail.com>
565 Chung-Ju Wu <jasonwucj@gmail.com>
567 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
568 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
569 Add enum values for DSP extension instructions.
570 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
572 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
573 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
575 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
576 * config/nds32/nds32-dspext.md: New file for DSP implementation.
577 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
578 * config/nds32/nds32-intrinsic.md: Likewise.
579 * config/nds32/nds32_intrinsic.h: Likewise.
580 * config/nds32/nds32-md-auxiliary.c: Likewise.
581 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
582 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
583 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
584 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
585 * config/nds32/nds32-protos.h: New declarations for DSP extension.
586 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
587 TYPE_DMAC in switch statement.
588 * config/nds32/nds32.c: New checking and implementation for DSP
589 extension instructions.
590 * config/nds32/nds32.h: Likewise.
591 * config/nds32/nds32.md: Likewise.
592 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
593 * config/nds32/predicates.md: Implement new predicates for DSP
596 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
598 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
599 Reformat alternatives and attributes so it is easier to identify
600 which constraints/attributes go with which instruction.
601 (mov<mode>_hardfloat32, FMOVE64): Likewise.
602 (mov<mode>_softfloat32, FMOVE64): Likewise.
603 (mov<mode>_hardfloat64, FMOVE64): Likewise.
604 (mov<mode>_softfloat64, FMOVE64): Likewise.
606 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
608 * doc/extend.texi (PowerPC Built-in Functions): Rename this
610 (Basic PowerPC Built-in Functions): The new name of the
611 subsection previously known as "PowerPC Built-in Functions".
612 (Basic PowerPC Built-in Functions Available on all Configurations):
614 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
615 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
616 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
617 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
619 2018-05-11 Martin Jambor <mjambor@suse.cz>
622 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
625 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
628 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
630 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
632 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
633 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
634 (ix86_handle_option): Handle -mwaitpkg.
635 * config.gcc: New header.
636 * config/i386/cpuid.h (bit_WAITPKG): New bit.
637 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
638 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
640 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
641 OPTION_MASK_ISA_WAITPKG.
642 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
643 (ix86_option_override_internal): Add PTA_WAITPKG.
644 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
645 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
647 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
648 __builtin_ia32_umwait and __builtin_ia32_tpause.
649 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
650 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
651 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
652 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
653 UNSPECV_TPAUSE): New.
654 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
655 * config/i386/i386.opt: Add -mwaitpkg.
656 * config/i386/waitpkgintrin.h: New file.
657 * config/i386/x86intrin.h: New header.
658 * doc/invoke.texi: Add -mwaitpkg.
660 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
663 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
665 (cortex-m0): Use armv6s-m isa.
666 (cortex-m0plus): Likewise.
667 (cortex-m1): Likewise.
668 (cortex-m0.small-multiply): Likewise.
669 (cortex-m0plus.small-multiply): Likewise.
670 (cortex-m1.small-multiply): Likewise.
672 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
673 Jakub Jelinek <jakub@redhat.com>
675 PR tree-optimization/85692
676 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
677 source permute as well.
679 2018-05-11 Martin Liska <mliska@suse.cz>
682 * doc/extend.texi: Document LLVM style format for no_sanitize
685 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
687 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
688 mode_supports_vsx_dform_quad to mode_supports_dq_form.
689 (mode_supports_vsx_dform_quad): Likewise.
690 (mode_supports_vmx_dform): Move these functions to be next to the
691 other mode_supports functions.
692 (mode_supports_dq_form): Likewise.
693 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
694 mode_supports_dq_form.
695 (reg_offset_addressing_ok_p): Likewise.
696 (offsettable_ok_by_alignment): Likewise.
697 (rs6000_legitimate_offset_address_p): Likewise.
698 (legitimate_lo_sum_address_p): Likewise.
699 (rs6000_legitimize_address): Likewise.
700 (rs6000_legitimize_reload_address): Likewise.
701 (rs6000_secondary_reload_inner): Likewise.
702 (rs6000_preferred_reload_class): Likewise.
703 (rs6000_output_move_128bit): Likewise.
705 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
707 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
708 Generate SImode target register for null target.
709 <case IX86_BUILTIN_XGETBV>: Ditto.
710 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
711 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
713 2018-05-10 Carl Love <cel@us.ibm.com>
715 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
716 dcbtt and dcbtstt if operands[2] is 0.
718 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
721 * config/i386/sse.md (usadv64qi): New expander.
723 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
725 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
726 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
727 -maltivec=be support.
728 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
729 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
730 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
731 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
732 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
733 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
734 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
735 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
736 altivec_vsumsws): Adjust.
737 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
738 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
739 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
741 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
742 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
743 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
744 (altivec_lve<VI_char>x): Delete expand.
745 (*altivec_lve<VI_char>x_internal): Rename to...
746 (altivec_lve<VI_char>x): ... this.
747 (altivec_lvxl_<mode>): Delete expand.
748 (*altivec_lvxl_<mode>_internal): Rename to ...
749 (altivec_lvxl_<mode>): ... this.
750 (altivec_stvxl_<mode>): Delete expand.
751 (*altivec_stvxl_<mode>_internal): Rename to ...
752 (altivec_stvxl_<mode>): ... this.
753 (altivec_stve<VI_char>x): Delete expand.
754 (*altivec_stve<VI_char>x_internal): Rename to ...
755 (altivec_stve<VI_char>x): ... this.
756 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
757 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
758 reduc_plus_scal_<mode>): Adjust.
759 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
761 (rs6000_cpu_cpp_builtins): Adjust.
762 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
763 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
764 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
765 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
766 -maltivec=be support.
767 (rs6000_split_vec_extract_var): Adjust.
768 (rs6000_split_v4si_init): Adjust.
769 (swap_selector_for_mode): Delete.
770 (altivec_expand_lvx_be, altivec_expand_stvx_be,
771 altivec_expand_stvex_be): Delete.
772 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
773 -maltivec=be support.
774 (rs6000_gimple_fold_builtin): Ditto.
775 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
777 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
778 (TARGET_DIRECT_MOVE_64BIT): Adjust.
779 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
780 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
781 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
782 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
783 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
784 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
785 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
786 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
787 anonymous split): Adjust.
788 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
789 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
791 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
793 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
794 when --with-gxx-include-dir is also specified.
795 * configure: Regenerate.
797 2018-05-09 Jim Wilson <jimw@sifive.com>
800 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
801 * config/riscv/t-withmultilib: New.
802 * config/riscv/withmultilib.h: New.
803 * doc/install.texi: Document RISC-V --with-multilib-list support.
805 2018-05-09 Richard Biener <rguenther@suse.de>
807 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
809 (vect_bb_vectorization_profitable_p): Adjust. Compute
810 actual scalar cost using the cost vector and the add_stmt_cost
813 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
815 PR rtl-optimization/85645
816 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
817 in the REG_CFA_REGISTER note for LR, don't leave it empty.
819 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
821 PR rtl-optimization/85645
822 * shrink-wrap.c (spread_components): Return a boolean saying if
823 anything was changed.
824 (try_shrink_wrapping_separate): Iterate spread_components until
825 nothing changes anymore.
827 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
829 PR rtl-optimization/85645
830 * regrename.c (build_def_use): Also kill the chains that include the
831 destination of a REG_CFA_REGISTER note.
833 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
835 PR rtl-optimization/85645
836 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
837 insn that has a REG_CFA_REGISTER note.
839 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
841 * cfgexpand.c (expand_clobber): New function.
842 (expand_gimple_stmt_1): Use it.
843 * tree-vect-stmts.c (vect_clobber_variable): New function,
845 (vectorizable_simd_clone_call): ...here.
846 (vectorizable_store): Emit a clobber either side of an
847 IFN_STORE_LANES sequence.
848 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
850 2018-05-09 Tom de Vries <tom@codesourcery.com>
853 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
854 (define_insn "trap_if_false"): Add exit after trap.
856 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
858 PR rtl-optimization/85638
859 * bb-reorder.c: Include common/common-target.h.
860 (create_forwarder_block): New function extracted from...
861 (fix_up_crossing_landing_pad): ...here. Rename into...
862 (dw2_fix_up_crossing_landing_pad): ...this.
863 (sjlj_fix_up_crossing_landing_pad): New function.
864 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
865 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
866 from both partitions and exit the loop after one iteration.
868 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
871 * doc/extend.texi (PowerPC Built-in Functions): Rename this
873 (Basic PowerPC Built-in Functions): The new name of the
874 subsection previously known as "PowerPC Built-in Functions".
875 (Basic PowerPC Built-in Functions Available on all Configurations):
877 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
879 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
881 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
883 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
886 2018-05-08 Jim Wilson <jimw@sifive.com>
888 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
889 (LD_EMUL_SUFFIX): New.
892 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
894 * doc/extend.texi (PowerPC Built-in Functions): Rename this
896 (Basic PowerPC Built-in Functions): The new name of the
897 subsection previously known as "PowerPC Built-in Functions".
898 (Basic PowerPC Built-in Functions Available on all Configurations):
900 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
902 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
904 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
906 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
909 2018-05-08 Jakub Jelinek <jakub@redhat.com>
912 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
913 after cmpelim optimization.
915 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
917 * config.gcc: Support "goldmont".
918 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
919 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
921 * config/i386/i386.c (m_GOLDMONT): Define.
922 (processor_target_table): Add "goldmont".
923 (PTA_GOLDMONT): Define.
924 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
925 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
926 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
927 (fold_builtin_cpu): Add "goldmont".
928 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
929 (ix86_option_override_internal): Add "goldmont".
930 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
931 (processor_type): Add PROCESSOR_GOLDMONT.
932 * config/i386/i386.md: Add CPU "glm".
933 * config/i386/glm.md: New file.
934 * config/i386/x86-tune.def: Add m_GOLDMONT.
935 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
937 2018-05-08 Jakub Jelinek <jakub@redhat.com>
940 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
942 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
943 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
944 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
947 * config/i386/i386.c (ix86_fold_builtin): Handle
948 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
951 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
952 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
954 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
957 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
958 (check_arch): Likewise.
959 (check_fpu): Return the result rather than printing it.
960 (end arch): Fix operator precedence.
962 (END): Print the result from check_fpu.
964 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
965 Alan Hayward <alan.hayward@arm.com>
966 David Sherwood <david.sherwood@arm.com>
968 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
969 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
970 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
971 (*fcmuo<mode>_and): New patterns.
973 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
975 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
976 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
977 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
978 (cmp_op, sve_imm_con): New code attributes.
979 (SVE_COND_INT_CMP, imm_con): Delete.
980 (cmp_op): Remove above unspecs from int attribute.
981 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
983 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
984 comparison-specific unspecs.
985 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
986 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
987 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
988 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
989 (*vec_fcm<cmp_op><mode>): Rename to...
990 (*fcm<cmp_op><mode>): ...this and adjust likewise.
991 (*vec_fcmuo<mode>): Rename to...
992 (*fcmuo<mode>): ...this and adjust likewise.
993 (*pred_fcm<cmp_op><mode>): New pattern.
994 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
995 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
997 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
999 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1000 (aarch64_emit_sve_predicated_cond): New function.
1001 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1002 (aarch64_emit_unspec_cond_or): Replace with...
1003 (aarch64_emit_sve_or_conds): ...this new function. Use
1004 aarch64_emit_sve_ptrue_op for the individual comparisons and
1005 aarch64_emit_binop to OR them together.
1006 (aarch64_emit_inverted_unspec_cond): Replace with...
1007 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1008 aarch64_emit_sve_ptrue_op for the comparison and
1009 aarch64_emit_unop to invert the result.
1010 (aarch64_expand_sve_vec_cmp_float): Update after the above
1011 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1013 2018-05-07 Nathan Sidwell <nathan@acm.org>
1015 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1016 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1017 (Backwards Compatibility): Likewise.
1019 2018-05-07 Luis Machado <luis.machado@linaro.org>
1023 2018-05-07 Luis Machado <luis.machado@linaro.org>
1025 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1026 <prefetch_dynamic_strides>: New const bool field.
1027 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1028 prefetch_dynamic_strides.
1029 (exynosm1_prefetch_tune): Likewise.
1030 (thunderxt88_prefetch_tune): Likewise.
1031 (thunderx_prefetch_tune): Likewise.
1032 (thunderx2t99_prefetch_tune): Likewise.
1033 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1034 (aarch64_override_options_internal): Update to set
1035 PARAM_PREFETCH_DYNAMIC_STRIDES.
1036 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1037 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1038 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1039 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1040 prefetch-dynamic-strides setting.
1042 2018-05-07 Luis Machado <luis.machado@linaro.org>
1044 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1045 <minimum_stride>: New const int field.
1046 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1047 minimum_stride field.
1048 (exynosm1_prefetch_tune): Likewise.
1049 (thunderxt88_prefetch_tune): Likewise.
1050 (thunderx_prefetch_tune): Likewise.
1051 (thunderx2t99_prefetch_tune): Likewise.
1052 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1053 (aarch64_override_options_internal): Update to set
1054 PARAM_PREFETCH_MINIMUM_STRIDE.
1055 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1056 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1057 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1058 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1059 stride is constant and is below the minimum stride threshold.
1061 2018-05-07 Luis Machado <luis.machado@linaro.org>
1063 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
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 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1107 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1108 the type is addressable. Don't force op into register if it has
1111 2018-05-05 Roland McGrath <mcgrathr@google.com>
1114 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1115 any section for which we don't know a specific type it should have,
1116 regardless of name. Previously this was done only for the exact
1117 names ".init_array", ".fini_array", and ".preinit_array".
1118 (default_elf_asm_named_section): Add comment about
1119 relationship with default_section_type_flags and SECTION_NOTYPE.
1120 (get_section): Don't consider it a type conflict if one side has
1121 SECTION_NOTYPE and the other doesn't, as long as neither has the
1122 SECTION_BSS et al used in the default_section_type_flags logic.
1124 2018-05-05 Tom de Vries <tom@codesourcery.com>
1127 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1128 (workaround_barsyncs): New function.
1129 (nvptx_reorg): Use workaround_barsyncs.
1130 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1131 (define_expand "nvptx_membar_cta"): New define_expand.
1132 (define_insn "*nvptx_membar_cta"): New insn.
1134 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1136 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1137 To improve optimization opportunities.
1138 * builtin-types.def: The new needed builtin types for the above.
1140 2018-05-04 Richard Biener <rguenther@suse.de>
1142 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1143 * gimple-ssa-store-merging.c
1144 (imm_store_chain_info::output_merged_store): Remove redundant create,
1145 release split_store vector contents on failure.
1146 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1147 scalar stmt vector on cache hit.
1149 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1151 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1153 * config.gcc (powerpc-xilinx-eabi*): Remove.
1154 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1156 (fusion_addis_mem_combo_load): Ditto.
1157 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1159 (rs6000_cpu_cpp_builtins): Ditto.
1160 * config/rs6000/rs6000-linux.c
1161 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1162 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1163 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1165 (rs6000_setup_reg_addr_masks): Ditto.
1166 (rs6000_init_hard_regno_mode_ok): Ditto.
1167 (rs6000_option_override_internal): Ditto.
1168 (legitimate_lo_sum_address_p): Ditto.
1169 (rs6000_legitimize_address): Ditto.
1170 (rs6000_legitimize_reload_address): Ditto.
1171 (rs6000_legitimate_address_p): Ditto.
1172 (abi_v4_pass_in_fpr): Ditto.
1173 (setup_incoming_varargs): Ditto.
1174 (rs6000_gimplify_va_arg): Ditto.
1175 (rs6000_split_multireg_move): Ditto.
1176 (rs6000_savres_strategy): Ditto.
1177 (rs6000_emit_prologue_components): Ditto.
1178 (rs6000_emit_epilogue_components): Ditto.
1179 (rs6000_emit_prologue): Ditto.
1180 (rs6000_emit_epilogue): Ditto.
1181 (rs6000_elf_file_end): Ditto.
1182 (rs6000_function_value): Ditto.
1183 (rs6000_libcall_value): Ditto.
1184 * config/rs6000/rs6000.h: Ditto.
1185 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1186 (TARGET_MINMAX): ... this. New.
1187 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1188 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1189 (*movsi_internal1_single): Delete.
1190 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1191 mfpu=, mxilinx-fpu): Delete.
1192 * config/rs6000/singlefp.h: Delete.
1193 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1194 * config/rs6000/t-rs6000: Ditto.
1195 * config/rs6000/t-xilinx: Delete.
1196 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1197 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1198 (VStype_simple): Delete.
1199 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1200 * config/rs6000/xfpu.h: Delete.
1201 * config/rs6000/xfpu.md: Delete.
1202 * config/rs6000/xilinx.h: Delete.
1203 * config/rs6000/xilinx.opt: Delete.
1204 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1205 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1207 2018-05-04 Tom de Vries <tom@codesourcery.com>
1210 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1213 2018-05-04 Richard Biener <rguenther@suse.de>
1216 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1217 (expand_complex_div_wide): Likewise.
1218 (expand_complex_operations_1): Likewise.
1219 (expand_complex_libcall): Preserve EH info of the original stmt.
1220 (tree_lower_complex): Handle removed blocks.
1221 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1222 on complex multiplication and division libcall builtins.
1224 2018-05-04 Richard Biener <rguenther@suse.de>
1227 * fold-const.c (negate_expr_p): Restrict negation of operand
1228 zero of a division to when we know that can happen without
1230 (fold_negate_expr_1): Likewise.
1232 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1235 * real.h (real_nextafter): Declare.
1236 * real.c (real_nextafter): New function.
1237 * fold-const-call.c (fold_const_nextafter): New function.
1238 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1239 CASE_CFN_NEXTTOWARD.
1240 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1241 even when arg1_mode is different from arg0_mode.
1243 2018-05-03 Nathan Sidwell <nathan@acm.org>
1245 * doc/extend.texi (Deprecated Features): Remove
1247 (Backwards Compatibility): Likewise.
1248 * doc/invoke.texi (C++ Language Options): Likewise.
1249 (C++ Dialect Options): Likewise.
1251 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1254 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1255 _mm512_mask_mullox_epi64): New intrinsics.
1257 2018-05-03 Tom de Vries <tom@codesourcery.com>
1260 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1261 dump files): Add offload-tree.
1263 2018-05-03 Richard Biener <rguenther@suse.de>
1265 PR tree-optimization/85615
1266 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1267 to loops not nested in BBs loop father to avoid creating multi-entry
1270 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1272 PR tree-optimization/70291
1273 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1274 arguments. Change return type to tree. Emit libcall as a new
1275 statement rather than replacing existing one when inplace_p is true.
1276 (expand_complex_multiplication_components): New function.
1277 (expand_complex_multiplication): Expand floating-point complex
1278 multiplication using the above.
1279 (expand_complex_division): Rename inner_type parameter to type.
1280 Update expand_complex_libcall call-site.
1281 (expand_complex_operations_1): Update expand_complex_multiplication
1282 and expand_complex_division call-sites.
1284 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1287 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1288 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1289 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1290 the highest significant bit of the shift count mask is clear. In
1291 check whether and[sq]i3 is needed verify that all significant bits
1292 of the shift count other than the highest are set.
1294 2018-05-02 Tom de Vries <tom@codesourcery.com>
1297 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1298 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1299 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1300 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1301 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1302 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1303 __builtin_goacc_parlevel_size.
1305 2018-05-02 Richard Biener <rguenther@suse.de>
1307 PR tree-optimization/85597
1308 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1309 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1312 2018-05-02 Tom de Vries <tom@codesourcery.com>
1315 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1316 dump files): Add ltrans-tree.
1318 2018-05-02 Tom de Vries <tom@codesourcery.com>
1321 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1322 dump files): Add wpa-ipa.
1324 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1326 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1327 powerpc*-*-linux*paired* target.
1328 * config/rs6000/750cl.h: Delete.
1329 * config/rs6000/paired.h: Delete.
1330 * config/rs6000/paired.md: Delete.
1331 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1333 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1334 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1335 comment. Remove paired float support.
1336 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1337 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1339 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1340 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1342 * config/rs6000/rs6000.c: Remove paired float support.
1343 (paired_expand_vector_init, paired_expand_vector_move,
1344 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1345 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1346 paired_expand_builtin, paired_expand_predicate_builtin,
1347 paired_init_builtins): Delete.
1348 * config/rs6000/rs6000.h: Remove paired float support.
1349 * config/rs6000/rs6000.md: Remove paired float support.
1350 (move_from_CR_ov_bit): Delete.
1351 * config/rs6000/rs6000.opt (mpaired): Delete.
1352 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1353 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1355 2018-05-02 Richard Biener <rguenther@suse.de>
1358 * gimplify.c (gimplify_save_expr): When in SSA form allow
1359 SAVE_EXPRs to compute to SSA vars.
1361 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1364 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1365 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1366 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1367 clobber operands[2], instead use a new pseudo. Formatting fixes.
1369 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1371 PR tree-optimization/85586
1372 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1373 exit early for statements in the same group if the accesses are
1376 2018-05-02 Tom de Vries <tom@codesourcery.com>
1379 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1382 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1384 PR tree-optimization/85143
1385 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1387 2018-05-01 Tom de Vries <tom@codesourcery.com>
1390 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1391 not found" error message.
1393 2018-05-01 Tom de Vries <tom@codesourcery.com>
1396 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1397 * vec.c (test_ordered_remove_if): New function.
1398 (vec_c_tests): Call test_ordered_remove_if.
1399 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1400 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1401 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1402 VEC_ORDERED_REMOVE_IF.
1404 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1406 PR tree-optimization/82665
1407 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1408 pointer subtraction where arguments come from a memchr call.
1410 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1412 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1413 --push-state --as-needed and --pop-state instead of --as-needed and
1414 --no-as-needed if ld supports it.
1415 * configure: Regenerated.
1418 * doc/install.texi2html: Replace _002d with - and _002a with * in
1419 generated html files using sed.
1421 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1424 * gcc-rich-location.c (blank_line_before_p): New function.
1425 (use_new_line): New function.
1426 (gcc_rich_location::add_fixit_insert_formatted): New function.
1427 * gcc-rich-location.h
1428 (gcc_rich_location::add_fixit_insert_formatted): New function.
1430 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1432 * selftest.c (assert_streq): Rename "expected" and "actual" to
1433 "val1" and "val2". Extend NULL-handling to cover both inputs
1434 symmetrically, while still requiring both to be non-NULL for a pass.
1435 * selftest.h (assert_streq): Rename "expected" and "actual" to
1437 (ASSERT_EQ): Likewise.
1438 (ASSERT_EQ_AT): Likewise.
1439 (ASSERT_KNOWN_EQ): Likewise.
1440 (ASSERT_KNOWN_EQ_AT): Likewise.
1441 (ASSERT_NE): Likewise.
1442 (ASSERT_MAYBE_NE): Likewise.
1443 (ASSERT_MAYBE_NE_AT): Likewise.
1444 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1445 the assertion to pass.
1446 (ASSERT_STREQ_AT): Likewise.
1448 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1450 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1451 interaction with -pie.
1453 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1455 * selftest.h: Fix alphabetization of per-source-file selftest
1458 2018-04-30 Jason Merrill <jason@redhat.com>
1460 PR c++/61982 - dead stores to destroyed objects.
1461 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1464 2018-04-30 Jason Merrill <jason@redhat.com>
1466 * tree.c (build_clobber): New.
1467 * tree.h: Declare it.
1468 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1470 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1472 * diagnostic-show-locus.c (layout::layout): Update for
1473 location_get_source_line returning a char_span.
1474 (struct char_span): Move to input.h.
1475 (struct correction): Update for fields in char_span becoming
1477 (struct source_line): Update for location_get_source_line
1478 returning a char_span.
1479 (layout::print_line): Likewise.
1480 * edit-context.c (edited_file::print_content): Likewise.
1481 (edited_file::print_diff_hunk): Likewise.
1482 (edited_file::print_run_of_changed_lines): Likewise.
1483 (edited_file::get_num_lines): Likewise.
1484 (edited_line::edited_line): Likewise.
1485 * final.c (asm_show_source): Likewise.
1486 * input.c (location_get_source_line): Convert return type
1487 from const char * to char_span, losing the final "line_len"
1489 (dump_location_info): Update for the above.
1490 (get_substring_ranges_for_loc): Likewise. Use a char_span
1491 when handling the literal within the line.
1492 (test_reading_source_line): Update for location_get_source_line
1493 returning a char_span.
1494 * input.h (class char_span): Move here from
1495 diagnostic-show-locus.c, converting from a struct to a class.
1496 Make data members private.
1497 (char_span::operator bool): New.
1498 (char_span::length): New.
1499 (char_span::get_buffer): New.
1500 (char_span::operator[]): New.
1501 (char_span::subspan): Make const.
1502 (char_span::xstrdup): New.
1503 (location_get_source_line): Convert return type from const char *
1504 to char_span, losing the final "line_size" param.
1506 2018-04-30 Jan Hubicka <jh@suse.cz>
1508 * lto-wrapper.c (ltrans_priorities): New static var.
1509 (cmp_priority): New.
1510 (run_gcc): Read priorities and if doing parallel build order
1511 the Makefile by them.
1513 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1515 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1517 2018-04-30 Richard Biener <rguenther@suse.de>
1519 * tree-cfg.c (verify_address): Remove base argument, add
1520 flag whether to check TREE_ADDRESSABLE and do that.
1521 (verify_expr): Remove.
1522 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1523 (verify_gimple_assign_single): Likewise.
1524 (verify_gimple_switch): Likewise.
1525 (verify_expr_location_1): Dereference tp once. Add (disabled)
1526 piece from verify_expr.
1527 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1529 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1531 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1533 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1535 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1536 (small_data_pattern): Likewise.
1537 (arc_rewrite_small_data): Likewise.
1538 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1539 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1540 (get_symbol_alignment): New function.
1541 (legitimate_small_data_address_p): Likewise.
1542 (legitimate_scaled_address): Update, call
1543 legitimate_small_data_address_p.
1544 (output_sdata): New static variable.
1545 (arc_print_operand): Update how we handle small data operands.
1546 (arc_print_operand_address): Likewise.
1547 (arc_legitimate_address_p): Update, use
1548 legitimate_small_data_address_p.
1549 (arc_rewrite_small_data_p): Remove.
1550 (arc_rewrite_small_data_1): Likewise.
1551 (arc_rewrite_small_data): Likewise.
1552 (small_data_pattern): Likewise.
1553 (compact_sda_memory_operand): Update to use
1554 legitimate_small_data_address_p and get_symbol_alignment.
1555 (prepare_move_operands): Don't rewite sdata pattern.
1556 (prepare_extend_operands): Remove.
1557 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1559 (zero_extendqisi2): Likewise.
1560 (zero_extendhisi2): Likewise.
1561 (extendqihi2): Likewise.
1562 (extendqisi2): Likewise.
1563 (extendhisi2): Likewise.
1567 * config/arc/constraints.md (Usd): Change it to memory constraint.
1569 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1571 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1572 as source of std instructions.
1573 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1574 allow 6-bit constants as source for store instructions.
1575 (movdi_insn): Update instruction pattern to allow 6-bit constants
1576 as source for store instructions.
1578 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1580 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1582 2018-04-30 Nathan Sidwell <nathan@acm.org>
1583 Sandra Loosemore <sandra@codesourcery.com>
1585 * dumpfile.c (dump_open): Allow '-' for stdout.
1586 * doc/invoke.texi (Developer Options): Document dump filename
1587 determination early. Document stdin/stdout selection.
1589 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1591 Microblaze Target: PIC data text relative
1593 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1594 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1596 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1597 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1598 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1599 New addressing mode for data-text relative position indepenedent code.
1600 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1601 'ADDRESS_SYMBOLIC_TXT_REL'.
1602 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1603 (microblaze_legitimate_pic_operand): Exclude function calls from
1604 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1605 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1607 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1608 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1609 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1610 for 'address + offset'.
1611 (microblaze_expand_prologue): Add new function prologue call for
1613 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1614 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1615 table in case of TARGET_PIC_DATA_TEXT_REL.
1616 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1617 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1618 Add new macros 'UNSPEC_TEXT',
1619 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1620 + exclude function calls from 'UNSPEC_PLT' in case of data text
1622 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1623 new target hook for generating address diff vector tables in case of
1625 * doc/tm.texi : Regenerate.
1626 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1627 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1628 of addr diff vector generation.
1629 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1630 target hook definition.
1631 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1632 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1633 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1634 option for data text relative.
1636 2018-04-30 Richard Biener <rguenther@suse.de>
1638 * tree-chrec.h (evolution_function_is_constant_p): Remove
1640 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1642 2018-04-30 Richard Biener <rguenther@suse.de>
1645 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1647 2018-04-30 Richard Biener <rguenther@suse.de>
1649 PR tree-optimization/28364
1650 PR tree-optimization/85275
1651 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1652 copying first exit test.
1654 2018-04-28 Mark Wielaard <mark@klomp.org>
1656 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1658 (dwarf_AT): Handle DW_AT_addr_base.
1659 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1661 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1664 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1665 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1666 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1667 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1669 2018-04-28 Richard Biener <rguenther@suse.de>
1671 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1672 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1673 to reflect use. Only add interesting stmts.
1675 2018-04-27 Martin Jambor <mjambor@suse.cz>
1678 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1679 the jump function allows for passing through aggregate values.
1681 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1683 * input.h (in_system_header_at): Convert from macro to inline
1685 (from_macro_expansion_at): Likewise.
1686 (from_macro_definition_at): Likewise.
1688 2018-04-27 Jeff Law <law@redhat.com>
1690 * config.gcc: Mark tile* targets as deprecated/obsolete.
1692 2018-04-27 Richard Biener <rguenther@suse.de>
1694 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1697 2018-04-27 Richard Biener <rguenther@suse.de>
1699 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1701 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1703 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1704 with Yd constraint. Set "preferred_for_speed" attribute from
1705 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1707 (*movdi_internal): Ditto.
1708 (movti_interunit splitters): Remove
1709 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1710 (movdi_interunit splitters): Ditto.
1711 * config/i386/constraints.md (Ye): Remove.
1712 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1714 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1717 * config/aarch64/constraints.md (Usg): Limit to 31.
1720 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1722 PR tree-optimization/85529
1723 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1724 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1725 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1726 zero extension or masking of the MSB bit.
1727 (optimize_range_tests): Add FIRST_BB argument, pass it through
1728 to optimize_range_tests_var_bound.
1729 (maybe_optimize_range_tests, reassociate_bb): Adjust
1730 optimize_range_tests callers.
1732 2018-04-26 Richard Biener <rguenther@suse.de>
1733 Jakub Jelinek <jakub@redhat.com>
1735 * cgraph.h (symbol_table): Just declare debug method here.
1736 * symtab.c (symbol_table::debug): Define.
1738 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1740 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1742 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1744 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1745 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1746 (*movdi_internal): Substitute Yi and Yj constraint with x
1747 and Ym and Yn constraint with y constraint. Update "isa"
1748 attribute and set "preferred_for_speed" attribute from
1749 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1750 (*movsi_internal): Ditto.
1751 (*movdf_internal): Ditto.
1752 (*movsf_internal): Ditto.
1753 (*zero_extendsidi2): Ditto.
1754 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1755 (sse2_loadld): Ditto.
1756 (*vec_extract<ssevecmodelower>_0): Ditto.
1757 (*vec_extractv4si_0_zext_sse4): Ditto.
1758 (vec_concatv2di): Ditto.
1759 (*vec_dup<mode>): Ditto.
1760 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1761 * config/i386/constraints.md (Yi): Remove.
1766 2018-04-26 Nathan Sidwell <nathan@acm.org>
1768 * dumpfile.c (dump_open): New.
1769 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1770 (dump_finish): Detect stdio/stderr by value not name.
1772 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1774 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1776 2018-04-26 Tom de Vries <tom@codesourcery.com>
1779 * config/nvptx/nvptx.c (verify_neutering_jumps)
1780 (verify_neutering_labels): New function
1781 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1783 2018-04-26 Tom de Vries <tom@codesourcery.com>
1786 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1787 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1790 2018-04-26 Richard Biener <rguenther@suse.de>
1791 Tom de Vries <tom@codesourcery.com>
1794 * lto-streamer-out.c (output_function): Fixup loops if required to match
1795 discovery done in the reader.
1797 2018-04-26 Richard Biener <rguenther@suse.de>
1799 PR tree-optimization/85116
1800 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1801 have a loop exit from the single latch predecessor. Remove
1802 case of header with just condition.
1803 (ch_base::copy_headers): Exclude infinite loops from any
1805 (pass_ch::execute): Record exits.
1807 2018-04-26 Richard Biener <rguenther@suse.de>
1809 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1810 prologue cost vector and pass it to vect_get_load_cost.
1811 (vect_get_peeling_costs_all_drs): Likewise.
1812 (vect_peeling_hash_get_lowest_cost): Likewise.
1813 (vect_enhance_data_refs_alignment): Likewise.
1815 2018-04-26 Richard Biener <rguenther@suse.de>
1818 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1819 checking of integer<->pointer conversions.
1820 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1821 sign-/zero-extending pointer types.
1822 (expand_omp_for_static_chunk): Likewise.
1824 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1825 Jean Lee <xiaoyur347@gmail.com>
1827 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1828 (TARGET_ASAN_SHADOW_OFFSET): Define.
1829 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1830 true for -fsanitize=address.
1832 2018-04-25 Mark Wielaard <mark@klomp.org>
1834 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1837 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1839 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1840 than "alu", remove explicit "memory" and "imm_disp" attributes.
1841 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1844 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1845 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1846 gen_lowpart_no_emit.
1848 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1851 * config/i386/i386.c (ix86_expand_builtin): Change memory
1852 operand to XI, extend p0 to Pmode.
1853 * config/i386/i386.md: Change unspec volatile and operand
1854 1 mode to XI, change operand 0 mode to P.
1856 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1858 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1859 GET_MODE_MASK before any checking.
1860 (nds32_can_use_bset_p): Likewise.
1861 (nds32_can_use_btgl_p): Likewise.
1863 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1865 * config/nds32/nds32-doubleword.md: New define_split pattern for
1866 illegal register number.
1868 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1870 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1872 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1874 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1876 2018-04-25 Richard Biener <rguenther@suse.de>
1878 * lto-streamer.h (LTO_major_version): Bump to 8.
1880 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1882 * BASE-VER: Set to 9.0.0.
1884 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1886 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1887 in __abskf2 and __powikf2.
1889 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1892 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1893 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1894 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1895 Use the above on operand 2. Reindent.
1896 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1898 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1901 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1903 * config/i386/i386.opt (mcet): Removed.
1904 * doc/install.texi: Remove -mcet documentation.
1905 * doc/invoke.texi: Likewise.
1907 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1910 * doc/install.texi: Remove -mcet from bootstrap-cet.
1912 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1915 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1916 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1920 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1921 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1922 containing a CONST_VECTOR.
1924 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1926 * doc/install.texi: Update newlib dependency for nvptx.
1928 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1931 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1932 instead of INTVAL when shifting x left.
1934 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1936 PR tree-optimization/85478
1937 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1938 vect_grouped_store_supported for single element vectors.
1940 2018-04-24 Richard Biener <rguenther@suse.de>
1943 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1944 load cost increase to the case of non-constant step.
1946 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1949 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1950 destination if any_malformed_asm.
1952 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1955 * expr.c (store_field): In the bitfield case, if the value comes from
1956 a function call and is returned in registers by means of a PARALLEL,
1957 do not change the mode of the temporary unless BLKmode and VOIDmode.
1959 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1961 PR rtl-optimization/85423
1962 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1963 dependencies to debug insns when the previous insn is non-debug.
1965 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1967 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1968 enums into a single definition.
1969 (fls): Fix predicates and printing.
1972 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1974 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1975 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1976 and short u6 immediate.
1977 (check_if_valid_sleep_operand): Remove.
1978 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1980 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1982 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1983 flag_always_save_lp condition.
1984 * config/nds32/nds32.opt (malways-save-lp): New option.
1986 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1988 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1989 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1990 * config/nds32/nds32.h
1991 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1992 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1994 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1996 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1997 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1999 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2000 Chung-Ju Wu <jasonwucj@gmail.com>
2002 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2004 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2005 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2007 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2009 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2011 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2013 * config/nds32/nds32-protos.h (nds32_data_alignment,
2014 nds32_local_alignment): Declare.
2015 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2016 nds32_local_alignment): New functions.
2017 (TARGET_CONSTANT_ALIGNMENT): Define.
2018 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2020 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2022 * config/nds32/nds32.c
2023 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2024 (TARGET_MODES_TIEABLE_P): Likewise.
2026 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2028 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2031 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2032 Chung-Ju Wu <jasonwucj@gmail.com>
2034 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2035 for unaligned access.
2036 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2038 * config/nds32/nds32-intrinsic.md: Likewise.
2039 * config/nds32/nds32_intrinsic.h: Likewise.
2040 * config/nds32/nds32.h (nds32_builtins): Likewise.
2041 * config/nds32/nds32.opt (munaligned-access): New option.
2042 * config/nds32/nds32.c (nds32_asm_file_start): Display
2043 flag_unaligned_access status.
2045 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2047 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2048 -mno-relax is present.
2049 * config/riscv/linux.h (LINK_SPEC): Ditto.
2051 2018-04-20 Martin Sebor <msebor@redhat.com>
2054 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2056 (gimple_fold_builtin_stxcpy_chk): Same.
2057 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2059 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2062 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2063 __powikf2 when long double is IEEE 128-bit.
2065 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2067 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2068 step to make sure stack always aligned.
2070 2018-04-20 Carl Love <cel@us.ibm.com>
2073 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2074 size check for arg0.
2076 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2077 Tom de Vries <tom@codesourcery.com>
2080 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2081 Emit insns for calls too.
2082 (nvptx_find_par): Always look for worker-level predecessor insn.
2083 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2085 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2086 (nvptx_process_pars): Propagate frames for calls.
2088 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2091 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2093 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2094 (ix86_handle_option): Don't handle OPT_mibt.
2095 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2097 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2099 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2100 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2101 (ix86_target_macros): Define __CET__ with flag_cf_protection
2102 for -fcf-protection.
2103 * config/i386/i386.c (isa2_opts): Remove -mibt.
2104 * config/i386/i386.h (TARGET_IBT): Removed.
2105 (TARGET_IBT_P): Likewise.
2106 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2107 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2108 * config/i386/i386.opt (mcet): Update help message.
2111 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2112 -mcet as an alias for -mshstk.
2114 2018-04-20 Richard Biener <rguenther@suse.de>
2117 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2118 complexity by forcing a single use of the multiply operand.
2120 2018-04-20 Martin Jambor <mjambor@suse.cz>
2123 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2124 recursion dependency to only apply to non-clones.
2126 2018-04-20 Martin Jambor <mjambor@suse.cz>
2129 * ipa-cp.c (create_specialized_node): Check that clones of
2130 self-recursive edges exist during IPA-CP.
2132 2018-04-19 Toon Moene <toon@moene.org>
2134 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2137 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2139 PR tree-optimization/85467
2140 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2141 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2142 VECTOR_CST element to type.
2144 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2147 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2148 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2149 (builtin_longjmp): Likewise.
2150 (save_stack_nonlocal): New pattern.
2151 (restore_stack_nonlocal): Likewise.
2153 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2156 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2157 Replace ASM_OUTPUT_LABEL with fprintf.
2159 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2162 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2163 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2164 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2165 define __IBT__ and __SHSTK__ for -fcf-protection.
2166 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2168 (ix86_trampoline_init): Likewise.
2169 (x86_output_mi_thunk): Likewise.
2170 (ix86_notrack_prefixed_insn_p): Likewise.
2171 (ix86_option_override_internal): Don't disallow -fcf-protection.
2172 * config/i386/i386.md (rdssp<mode>): Also enable for
2174 (incssp<mode>): Likewise.
2175 (nop_endbr): Likewise.
2176 * config/i386/i386.opt (mcet): Change help message to built-in
2180 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2181 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2182 enable CET built-in functions.
2184 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2186 * common/config/i386/i386-common.c
2187 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2188 OPTION_MASK_ISA_MOVDIRI_UNSET,
2189 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2190 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2191 * config.gcc (movdirintrin.h): New header.
2192 * config/i386/cpuid.h (bit_MOVDIRI,
2193 bit_MOVDIR64B): New bits.
2194 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2196 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2197 (VOID, PVOID, PCVOID)): New function types.
2198 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2199 __builtin_ia32_directstoreu_u64,
2200 __builtin_ia32_movdir64b): New builtins.
2201 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2202 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2204 (ix86_valid_target_attribute_inner_p): Ditto.
2205 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2206 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2207 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2208 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2209 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2210 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2211 (movdiri<mode>, movdir64b_<mode>): New.
2212 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2213 * config/i386/immintrin.h: Include movdirintrin.h.
2214 * config/i386/movdirintrin.h: New file.
2215 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2217 2018-04-19 Richard Biener <rguenther@suse.de>
2220 * cfg.c (clear_bb_flags): When loop state says we have
2221 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2223 2018-04-19 Richard Biener <rguenther@suse.de>
2225 PR tree-optimization/84737
2226 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2227 copying restrict info.
2228 (vect_setup_realignment): Use it.
2229 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2230 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2231 the first DR to all generated stores.
2232 (vectorizable_load): Likewise for loads.
2234 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2236 PR tree-optimization/85446
2237 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2238 the integral and pointer types to have the same precision.
2240 * doc/install.texi: Document --disable-cet being the default and
2243 2018-04-18 Martin Liska <mliska@suse.cz>
2245 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2248 2018-04-18 Martin Liska <mliska@suse.cz>
2251 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2254 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2255 arguments if they are comparable.
2257 2018-04-18 Martin Liska <mliska@suse.cz>
2260 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2263 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2266 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2269 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2270 ENDBR after calling __morestack.
2272 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2275 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2276 by using gcc_base_ver to generate a gcc_driver_version, and use
2277 it when generating GCC_DRIVER_NAME.
2278 * configure: Regenerate.
2280 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2283 * config.gcc: Obsolete powerpc*-*-*spe*.
2285 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2288 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2289 (stabstr_D): Change type of unum from unsigned int to
2290 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2293 2018-04-17 Jim Wilson <jimw@sifive.com>
2296 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2297 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2298 Set arg_pointer_offset after using pretend_args_size.
2300 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2302 PR rtl-optimization/85431
2303 * dse.c (record_store): Ignore zero width stores.
2306 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2307 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2308 __builtin_stack_restore rather than after it.
2309 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2310 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2311 argument instead of virtual_dynamic_stack_rtx.
2313 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2315 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2317 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2318 Add note to error message to explain internal mapping of overloaded
2319 built-in function name to non-overloaded built-in function name.
2320 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2323 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2326 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2327 where the inputs overlap with the output.
2329 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2331 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2332 (=v, v) alternative and explicit "memory" attribute.
2333 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2334 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2336 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2337 "sselog1" type instead of "sselog".
2338 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2339 "sselog". Remove explicit "memory" attribute.
2340 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2341 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2343 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2344 "sselog1" type instead of "sselog". Remove explicit "memory"
2346 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2347 use "sselog1" type instead of "sselog". Remove explicit "memory"
2349 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2350 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2352 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2353 "sselog1" type instead of "sselog". Remove explicit "memory"
2355 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2356 use "sselog1" type instead of "sselog". Remove explicit "memory"
2360 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2363 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2366 2018-04-17 Martin Jambor <mjambor@suse.cz>
2369 * ipa-cp.c (create_specialized_node): Call
2370 expand_all_artificial_thunks if necessary.
2372 2018-04-17 Martin Liska <mliska@suse.cz>
2375 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2376 in message, remote space in between '_G' and '('.
2378 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2381 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2382 avx512f_vmcmp<mode>3<round_saeonly_name>,
2383 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2384 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2385 avx512f_rndscale<mode><round_saeonly_name>,
2386 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2387 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2388 Use %<iptr>2 instead of %2 for -masm=intel.
2389 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2390 avx512f_vcvttss2usi<round_saeonly_name>,
2391 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2393 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2394 avx512f_vcvttsd2usi<round_saeonly_name>,
2395 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2396 Use %q1 instead of %1 for -masm=intel.
2397 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2398 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2399 of %3 for -masm=intel.
2400 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2402 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2404 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2406 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2407 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2409 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2411 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2412 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2413 %g1 and one with %0 and %1.
2414 (avx512er_vmrcp28<mode><round_saeonly_name>,
2415 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2417 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2418 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2419 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2420 of %0 and %{%4%} for -masm=intel.
2421 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2422 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2423 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2424 order of %0 and %{%5%}%{z%} for -masm=intel.
2426 2018-04-17 Jan Hubicka <jh@suse.cz>
2429 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2431 2018-04-17 Martin Liska <mliska@suse.cz>
2434 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2435 for target_clone error message. Make default implementation
2436 clone to be a local declaration.
2437 (separate_attrs): Add new argument and check for an empty
2439 (expand_target_clones): Handle it.
2440 (ipa_target_clone): Make redirection just for target_clones
2443 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2444 Tom de Vries <tom@codesourcery.com>
2447 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2448 tiled basic blocks without omp continue statements.
2450 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2453 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2454 vec_extract expression as having side effects to make sure it gets
2457 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2460 * config/i386/i386.c (get_builtin_code_for_version): Check
2463 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2466 * gcc/config.gcc: Support "skylake".
2467 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2469 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2470 (processor_target_table): Add "skylake".
2471 (ix86_option_override_internal): Add "skylake".
2472 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2473 PROCESSOR_CANNONLAKE.
2474 (get_builtin_code_for_version): Fix priority for
2475 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2476 PROCESSOR_SKYLAKE-AVX512.
2477 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2478 (processor_type): Add PROCESSOR_SKYLAKE.
2480 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2481 Jason Merrill <jason@redhat.com>
2484 * convert.c (convert_to_integer_1): Use direct recursion for
2485 enumeral types and types with a precision less than the number
2486 of bits in their mode.
2488 2018-04-16 Julia Koval <julia.koval@intel.com>
2491 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2492 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2494 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2497 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2498 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2499 and -mno-direct-move.
2501 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2504 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2505 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2506 Ensure negative shifts result in {0}.
2508 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2510 PR rtl-optimization/79916
2511 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2512 regs (if any) to define how to gnerate SD moves when LRA is in
2515 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2517 PR rtl-optimization/85393
2518 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2519 * except.c (expand_dw2_landing_pad_for_region): Make static.
2520 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2521 a label and unconditional jump to old_bb, rather than
2522 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2525 PR rtl-optimization/85376
2526 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2527 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2528 instead of a specific value.
2530 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2531 Bin Cheng <bin.cheng@arm.com>
2533 PR tree-optimization/82965
2534 PR tree-optimization/83991
2535 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2536 by_profile_only parameter.
2537 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2538 information if the loop was predicted to iterate too many times.
2539 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2541 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2544 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2547 2018-04-13 Martin Liska <mliska@suse.cz>
2548 Jakub Jelinek <jakub@redhat.com>
2551 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2552 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2553 * builtins.c (expand_builtin_memory_copy_args): Use
2554 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2555 handle dest_addr == pc_rtx.
2557 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2560 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2561 asked to not generate direct moves.
2562 (fix_trunc<mode>si2_stfiwx): Similar.
2563 (fix_trunc<mode>si2_internal): Similar.
2565 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2568 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2569 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2570 lookup if dest in some wider mode is known to be const0_rtx and
2571 if so, record permanent equivalence for it to be ZERO_EXTEND of
2572 the narrower mode destination.
2574 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2576 * lto-streamer-out.c (output_function): Revert 259346.
2577 * omp-expand.c (expand_oacc_for): Likewise.
2579 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2581 PR rtl-optimization/85354
2582 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2583 * sel-sched.c (sel_global_init): ... here.
2585 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2588 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2589 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2590 mode for PE-COFF targets.
2591 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2592 (i386_pe_asm_lto_end): Likewise.
2593 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2594 (TARGET_ASM_LTO_END): Likewise.
2595 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2596 (i386_pe_asm_lto_start): New function.
2597 (i386_pe_asm_lto_end): Likewise.
2599 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2600 Richard Biener <rguenther@suse.de>
2603 * lto-streamer-out.c (output_function): Fix CFG loop state before
2605 * omp-expand.c (expand_oacc_for): Handle calls to internal
2606 functions like regular functions.
2608 2018-04-12 Richard Biener <rguenther@suse.de>
2611 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2612 for the early LTO debug to properly generate references to it
2613 during DIE emission. Do not re-use that for the skeleton for
2615 (dwarf2out_early_finish): Likewise.
2617 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2620 * config/i386/sse.md
2621 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2622 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2623 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2624 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2625 and output is a reg, avoid creating invalid lowpart subreg, but
2626 instead split into a 512-bit move. Don't split if not AVX512VL,
2627 input is xmm16+ reg and output is a mem.
2628 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2629 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2630 xmm16+ reg and output is a mem.
2632 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2634 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2635 also for flag_dwarf2_cfi_asm.
2637 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2639 PR rtl-optimization/85342
2640 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2641 a bool scalar var inside of the loop instead. Don't try to update
2642 recog_data.operand after failed apply_change_group.
2644 2018-04-12 Tom de Vries <tom@codesourcery.com>
2647 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2648 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2649 array with flexible array member as array without given dimension.
2650 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2651 argument for undefined param to true.
2653 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2656 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2657 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2658 from PowerPC section.
2659 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2660 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2661 help text that is too long.
2662 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2663 help text that is too long.
2664 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2665 help text that is too long.
2667 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2669 * config/alpha/alpha.md (stack_probe_internal): Rename
2670 from "probe_stack". Update all callers.
2672 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2674 PR rtl-optimization/84566
2675 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2676 sched_macro_fuse_insns.
2678 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2681 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2682 (compute_block_dependences): ... from here.
2684 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2686 PR tree-optimization/85331
2687 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2688 from int to HOST_WIDE_INT.
2690 2018-04-11 Martin Jambor <mjambor@suse.cz>
2693 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2694 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2695 not the same as the source val.
2696 (cgraph_edge_brings_value_p): New parameter.
2697 (gather_edges_for_value): Pass destination value to
2698 cgraph_edge_brings_value_p.
2699 (perhaps_add_new_callers): Likewise.
2700 (get_info_about_necessary_edges): Likewise and exclude values brought
2701 only by self-recursive edges.
2702 (create_specialized_node): Redirect only clones of self-calling edges.
2703 (+self_recursive_pass_through_p): New function.
2704 (find_more_scalar_values_for_callers_subset): Use it.
2705 (find_aggregate_values_for_callers_subset): Likewise.
2706 (known_aggs_to_agg_replacement_list): Removed.
2707 (decide_whether_version_node): Re-calculate known constants for all
2708 remaining context clones.
2710 2018-04-11 Richard Biener <rguenther@suse.de>
2713 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2714 from early DWARF output.
2715 (dwarf2out_early_finish): Output line info unconditionally into
2716 early DWARF and add reference to it.
2718 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2721 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2722 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2723 other than V2DFmode using iptr mode attribute.
2724 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2726 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2728 PR rtl-optimization/84659
2729 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2731 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2734 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2736 (output_loc_list): Pass address of a dummy size variable even in the
2737 locview handling loop.
2738 (index_location_lists): Add comment on why skip_loc_list_entry can't
2741 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2744 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2747 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2750 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2751 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2752 and -mstring-compare-inline-limit.
2754 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2757 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2758 for stack clash protection in a register whenever we need it to be in
2761 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2763 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2764 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2766 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2769 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2771 (mlong-double-): Ditto.
2772 * config/rs6000/sysv4.opt (msdata=): Ditto.
2773 (mtls-size=): Ditto.
2775 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2777 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2778 erroneous entries for
2779 "vector int vec_ldl (int, long int *)", and
2780 "vector unsigned int vec_ldl (int, unsigned long int *)".
2781 Add comments and entries for
2782 "vector bool char vec_ldl (int, bool char *)",
2783 "vector bool short vec_ldl (int, bool short *)",
2784 "vector bool int vec_ldl (int, bool int *)",
2785 "vector bool long long vec_ldl (int, bool long long *)",
2786 "vector pixel vec_ldl (int, pixel *)",
2787 "vector long long vec_ldl (int, long long *)",
2788 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2789 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2790 type tree bool_long_long_type_node and correct definition of
2791 bool_V2DI_type_node to make reference to this new type tree.
2792 (rs6000_mangle_type): Replace erroneous reference to
2793 bool_long_type_node with bool_long_long_type_node.
2794 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2795 comments to emphasize sign distinctions for char and int types and
2796 replace RS6000_BTI_bool_long constant with
2797 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2798 use of RS6000_BTI_pixel.
2799 (bool_long_type_node): Remove this macro definition.
2800 (bool_long_long_type_node): New macro definition
2802 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2804 PR rtl-optimization/85300
2805 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2806 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2807 simplify_unary_operation fails.
2809 2018-04-10 Martin Liska <mliska@suse.cz>
2811 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2812 cgraph_edge and ipa_ref.
2814 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2818 * config/i386/sse.md
2819 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2820 computation of the VEC_MERGE selector from mask.
2821 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2822 Fix decoding of the VEC_MERGE selector into mask.
2824 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2826 PR tree-optimization/85286
2827 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2829 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2831 * final.c (final_1): Set insn_last_address as well as
2832 insn_current_address.
2834 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2837 * explow.c (emit_stack_probe): Call validize_mem on memory location
2838 before passing it to gen_probe_stack. Create address operand and
2839 legitimize it for the probe_stack_address case.
2841 2018-04-09 Jan Hubicka <jh@suse.cz>
2844 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2845 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2846 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2848 (free_lang_data): Rebuild type inheritance graph.
2850 2018-04-09 Martin Sebor <msebor@redhat.com>
2852 * invoke.texi (-finline-small-functions): Mention other optimization
2854 (-findirect-inlining, -fpartial-inlining): Same.
2855 (-finline-functions-called-once): Same.
2856 (-freorder-blocks-and-partition): Same.
2858 2018-04-09 Jan Hubicka <jh@suse.cz>
2861 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2862 jumps; choose last target that matches the criteria (i.e.
2863 no partition changes for non-crossing jumps).
2864 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2865 support for redirecting crossing jumps to non-crossing.
2867 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2869 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2870 also for naked functions.
2872 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2874 * config/arc/arc.md (add_shift): New pattern.
2875 (add_shift2): Likewise.
2876 (sub_shift): Likewise.
2877 (sub_shift_cmp0_noout): Likewise.
2878 (compare_si_ashiftsi): Likewise.
2879 (xbfu_cmp0_noout): New combine pattern.
2880 (xbfu_cmp0"): Likewise.
2881 (movsi_set_cc_insn): Place the predicable variant first.
2882 (commutative_binary_cmp0_noout): Remove clobber.
2883 (commutative_binary_cmp0): New pattern.
2884 (noncommutative_binary_cmp0): Likewise.
2885 (noncommutative_binary_cmp0_noout): Likewise.
2886 (noncommutative_binary_comparison_result_used): Removed.
2887 (rsub_cmp0): New pattern.
2888 (rsub_cmp0_noout): Likewise.
2889 (extzvsi): Changed, keep only meaningful variants.
2890 (SQH, SEZ): New iterators.
2891 (SQH_postfix): New mode attribute.
2892 (SEZ_prefix): New code attribute.
2893 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2894 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2895 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2897 (noncommutative_operator): Check the availability of barrel
2900 2018-04-09 Richard Biener <rguenther@suse.de>
2902 PR tree-optimization/85284
2903 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2904 Only use the niter constraining form of simple_iv when the exit
2907 2018-04-09 Tom de Vries <tom@codesourcery.com>
2910 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2911 (define_expand "*memory_barrier"): New define_expand.
2912 (define_insn "memory_barrier"): New insn.
2914 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2916 PR rtl-optimization/80463
2917 PR rtl-optimization/83972
2918 PR rtl-optimization/83480
2920 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2921 correct producer for the insn.
2922 (tidy_control_flow): Fixup seqnos in case of debug insns.
2924 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2926 PR rtl-optimization/83913
2928 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2929 different sched-times when merging exprs.
2931 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2933 PR rtl-optimization/83962
2935 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2936 tidy_fallthru_edge and tidy_control_flow.
2938 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2940 PR rtl-optimization/83530
2942 * sel-sched.c (force_next_insn): New global variable.
2943 (remove_insn_for_debug): When force_next_insn is true, also leave only
2944 next insn in the ready list.
2945 (sel_sched_region): When the region wasn't scheduled, make another pass
2946 over it with force_next_insn set to 1.
2948 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2950 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2952 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2953 for interrupt control.
2954 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2955 functions for interrupt control.
2956 * config/nds32/nds32-intrinsic.md: Likewise.
2957 * config/nds32/nds32_intrinsic.h: Likewise.
2958 * config/nds32/nds32.h (nds32_builtins): Likewise.
2960 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2962 * config/nds32/nds32.c (nds32_init_machine_status,
2963 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2964 strict_aligned_p field.
2965 (nds32_expand_to_rtl_hook): New function.
2966 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2967 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2969 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2970 Chung-Ju Wu <jasonwucj@gmail.com>
2972 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2973 * config/nds32/nds32-n7.md: New file.
2974 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2975 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2977 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2978 * config/nds32/nds32.md (pipeline_model): Add n7.
2979 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2980 * config/nds32/pipelines.md: Include n7 settings.
2982 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2983 Chung-Ju Wu <jasonwucj@gmail.com>
2985 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2986 * config/nds32/nds32-e8.md: New file.
2987 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2988 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2990 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2991 * config/nds32/nds32.md (pipeline_model): Add e8.
2992 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2993 * config/nds32/pipelines.md: Include e8 settings.
2995 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2996 Chung-Ju Wu <jasonwucj@gmail.com>
2998 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2999 * config/nds32/nds32-n8.md: New file.
3000 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3001 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3003 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3004 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3005 * config/nds32/nds32.md (pipeline_model): Add n8.
3006 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3007 * config/nds32/pipelines.md: Include n8 settings.
3009 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3010 Chung-Ju Wu <jasonwucj@gmail.com>
3012 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3013 * config/nds32/nds32-n9-2r1w.md: New file.
3014 * config/nds32/nds32-n9-3r2w.md: New file.
3015 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3016 nds32_register_ports): New or modify for cpu n9.
3017 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3019 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3020 * config/nds32/nds32-utils.c: New file.
3021 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3022 TARGET_MUL_SLOW): Define.
3023 * config/nds32/nds32.md (pipeline_model): New attribute.
3024 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3025 New options that support cpu n9.
3026 * config/nds32/pipelines.md: Include n9 settings.
3027 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3029 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3031 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3032 information if necessary.
3033 (output_cond_branch_compare_zero): Likewise.
3034 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3035 (nds32_target_alignment): Refine for alignment.
3036 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3037 (FUNCTION_BOUNDARY): Modify.
3038 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3040 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3042 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3044 * config/nds32/constants.md (unspec_volatile_element): Add values for
3045 TLB operation and data prefetch.
3046 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3047 functions for TLB operation and data prefetch.
3048 * config/nds32/nds32-intrinsic.md: Likewise.
3049 * config/nds32/nds32_intrinsic.h: Likewise.
3050 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3051 (nds32_print_operand): Likewise.
3052 * config/nds32/nds32.h (nds32_builtins): Likewise.
3054 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3055 Andrew Pinski <pinsika@gcc.gnu.org>
3058 * match.pd: Use constant_boolean_node of correct type instead of
3059 boolean_true_node or boolean_false_node for simplifying
3060 pointer comparisons to zero.
3062 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3064 PR tree-optimization/80021
3065 * tree.c (verify_type_variant): Make error call in verify_variant_match
3066 translatable and remove final full stop.
3068 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3070 * config/nds32/constants.md (unspec_volatile_element): Add
3071 UNSPEC_VOLATILE_EH_RETURN.
3072 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3073 nds32_output_stack_pop): Support dwarf exception handling process.
3074 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3075 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3076 exception handling process.
3077 (nds32_compute_stack_frame): Likewise.
3078 (nds32_return_addr_rtx): Likewise.
3079 (nds32_initial_elimination_offset): Likewise.
3080 (nds32_expand_prologue): Likewise.
3081 (nds32_expand_epilogue): Likewise.
3082 (nds32_dynamic_chain_address): New function.
3083 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3085 (DYNAMIC_CHAIN_ADDRESS): Define.
3086 (EH_RETURN_DATA_REGNO): Define.
3087 (EH_RETURN_STACKADJ_RTX): Define.
3088 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3089 patterns for dwarf exception handling.
3091 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3093 * config/nds32/nds32.h: Clean up obsolete macros.
3095 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3097 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3098 Add enum values for particular instructions.
3099 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3100 particular intrinsic functions.
3101 * config/nds32/nds32-intrinsic.md: Likewise.
3102 * config/nds32/nds32_intrinsic.h: Likewise.
3103 * config/nds32/nds32.h (nds32_builtins): Likewise.
3104 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3105 (btst, ave): New patterns for particular instructions.
3107 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3109 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3110 Add enum values for atomic load/store and memory sync.
3111 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3113 * config/nds32/nds32-intrinsic.md: Likewise.
3114 * config/nds32/nds32_intrinsic.h: Likewise.
3115 * config/nds32/nds32.h (nds32_builtins): Likewise.
3117 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3119 PR tree-optimization/85257
3120 * fold-const.c (native_encode_vector): If not all elts could fit
3121 and off is -1, return 0 rather than offset.
3122 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3123 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3124 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3125 adjust buffer in native_interpret_expr call.
3127 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3129 * config/nds32/constants.md (unspec_volatile_element): Add cache
3130 control enum values.
3131 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3132 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3133 * config/nds32/nds32.c (nds32_cctl_names): New.
3134 (nds32_print_operand): Handle cache control register names.
3135 * config/nds32/nds32.h (nds32_builtins): New enum values.
3136 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3138 * config/nds32/nds32.md (type): Add mmu.
3139 * config/nds32/pipelines.md (simple_insn): Add mmu.
3141 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3143 * config/nds32/nds32.md (type): Remove call.
3144 * config/nds32/pipelines.md (simple_insn): Likewise.
3146 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3148 * config/nds32/constants.md (unspec_volatile_element): Add
3149 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3150 UNSPEC_VOLATILE_FMFCFG.
3151 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3152 description for fmfcfg and fmfcsr.
3153 (bdesc_1arg): Add fmtcsr.
3154 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3155 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3156 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3157 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3158 unspec_fmfcfg): New patterns.
3159 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3160 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3161 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3162 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3163 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3164 __nds32__fmfcfg): Define.
3166 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3168 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3169 intrinsic register names.
3170 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3171 intrinsic register enum values and macros.
3173 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3175 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3176 for load/store addressing form.
3177 (nds32_print_operand_address): Likewise.
3179 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3182 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3183 based on LABEL_REF. Remove useless assertion.
3184 (pic_address_needs_scratch): Fix formatting.
3185 (sparc_legitimize_pic_address): Minor tweaks.
3186 (sparc_delegitimize_address): Adjust assertion accordingly.
3187 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3188 into symbolic_operand.
3189 (movsi_high_pic_label_ref): Likewise.
3190 (movsi_lo_sum_pic_label_ref): Likewise.
3191 (movdi_pic_label_ref): Likewise.
3192 (movdi_high_pic_label_ref): Likewise.
3193 (movdi_lo_sum_pic_label_ref): Likewise.
3195 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3197 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3198 custom LIB_SPEC setup.
3200 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3201 Kito Cheng <kito.cheng@gmail.com>
3203 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3204 * config/riscv/freebsd.h: New.
3206 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3208 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3209 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3212 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3213 Kito Cheng <kito.cheng@gmail.com>
3215 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3216 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3217 * config/nds32/nds32-protos.h (nds32_output_call,
3218 nds32_output_return): Declare.
3219 * config/nds32/nds32.md: Refine all the call and return patterns.
3221 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3224 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3225 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3227 PR rtl-optimization/84872
3228 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3229 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3232 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3234 * expr.c (copy_blkmode_to_reg): Revert 254862.
3235 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3237 2018-04-06 Richard Biener <rguenther@suse.de>
3240 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3241 after seeing a component reference with an adjacent field. Treat
3242 refs to arrays at struct end of external decls similar to
3243 refs to unconstrained commons.
3245 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3248 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3249 look through SAVE_EXPRs with non-side-effects argument. Adjust
3251 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3254 2018-04-06 Richard Biener <rguenther@suse.de>
3257 * alias.c (find_base_term): New wrapper around find_base_term
3258 unwinding CSELIB_VAL_PTR changes.
3259 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3262 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3264 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3266 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3267 constant definitions.
3268 ("nop"): lr 0,0 -> nopr r0
3269 ("nop_lr0", "nop_lr1"): New insn definitions.
3271 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3273 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3274 NDS32_V3PUSH_AVAILABLE_P macro.
3276 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3277 Chung-Ju Wu <jasonwucj@gmail.com>
3279 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3280 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3281 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3282 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3283 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3284 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3285 * config/nds32/constraints.md: New constraints and checking for hard
3286 float configuration.
3287 * config/nds32/iterators.md: New mode iterator and attribute for hard
3288 float configuration.
3289 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3291 * config/nds32/nds32-fpu.md: New file.
3292 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3293 deal with hard float code generation.
3294 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3296 (abi_type, float_reg_number): New enum type.
3297 * config/nds32/nds32-predicates.c: New predicates for hard float.
3298 * config/nds32/nds32-protos.h: Declare functions for hard float.
3299 * config/nds32/nds32.c: Implementation for hard float configuration.
3300 * config/nds32/nds32.h: Definitions for hard float configuration.
3301 * config/nds32/nds32.md: Include hard float machine description and
3302 modify patterns for hard float configuration.
3303 * config/nds32/nds32.opt: New options for hard float configuration.
3304 * config/nds32/predicates.md: New predicates for hard float
3307 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3309 * common/config/nds32/nds32-common.c
3310 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3312 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3315 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3316 CONSTRUCTOR_ELT (ctor, ...)->value.
3318 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3321 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3323 2018-04-05 Tom de Vries <tom@codesourcery.com>
3326 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3329 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3330 Kito Cheng <kito.cheng@gmail.com>
3332 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3333 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3334 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3335 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3337 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3338 Kito Cheng <kito.cheng@gmail.com>
3340 * config/nds32/constraints.md (Ufe): New memory constraint.
3341 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3342 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3343 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3345 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3346 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3348 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3350 * config/nds32/nds32.md: Use optimize_size in the condition for
3351 alu-shift instructions.
3353 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3355 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3357 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3359 * config/nds32/nds32.md (negsi2): Refine pattern.
3361 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3362 Chung-Ju Wu <jasonwucj@gmail.com>
3364 * config/nds32/iterators.md (shift_rotate): New code iterator.
3365 (shift): New code attribute.
3366 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3367 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3368 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3369 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3370 bit-wise operations.
3371 (andsi3, *andsi3): Ditto.
3372 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3373 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3374 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3375 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3376 nds32_ior_operand, nds32_xor_operand): New predicates.
3378 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3380 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3381 (addsi3, subsi3): ... this.
3383 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3385 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3387 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3389 * config/nds32/nds32.md: Adjust indention.
3391 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3393 * config/nds32/nds32.md (feature): New attribute.
3395 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3397 * config/nds32/nds32.md (subtype): New attribute.
3399 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3402 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3403 expansion to perform a bitwise AND of the argument followed by a
3404 boolean negation of the result.
3406 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3408 PR rtl-optimization/84878
3409 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3410 the basic block. Assert the use reference is not artificial and that
3411 it has an associated insn.
3413 2018-04-04 Michael Matz <matz@suse.de>
3415 * builtins.c (compute_objsize): Pass correct operand
3416 to array_at_struct_end_p.
3418 2018-04-04 Richard Biener <rguenther@suse.de>
3421 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3422 from contexts for DINFO_LEVEL_TERSE and below.
3424 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3426 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3427 resiter_operand condition.
3428 * config/nds32/nds32.md (*move<mode>): Ditto.
3430 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3431 Monk Chiang <sh.chiang04@gmail.com>
3433 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3435 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3437 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3439 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3440 Kito Cheng <kito.cheng@gmail.com>
3442 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3443 nds32_cond_code_str, output_cond_branch,
3444 output_cond_branch_compare_zero, nds32_expand_cbranch,
3445 nds32_expand_cstore, nds32_expand_movcc,
3446 nds32_output_cbranchsi4_equality_zero,
3447 nds32_output_cbranchsi4_equality_reg,
3448 nds32_output_cbranchsi4_equality_reg_or_const_int,
3449 nds32_output_cbranchsi4_greater_less_zero: New functions.
3450 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3451 nds32_expand_cstore, nds32_expand_movcc,
3452 nds32_output_cbranchsi4_equality_zero,
3453 nds32_output_cbranchsi4_equality_reg,
3454 nds32_output_cbranchsi4_equality_reg_or_const_int,
3455 nds32_output_cbranchsi4_greater_less_zero): Declare.
3456 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3457 nds32_rimm11s_operand): New predicates.
3458 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3459 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3462 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3464 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3465 * config/nds32/nds32.md: Ditto.
3466 * config/nds32/pipelines.md: Ditto.
3468 2018-04-04 Richard Biener <rguenther@suse.de>
3470 PR tree-optimization/85168
3471 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3472 propagating abnormals.
3474 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3476 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3478 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3479 Kito Cheng <kito.cheng@gmail.com>
3481 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3482 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3483 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3484 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3485 * config/nds32/nds32.md (sibcall_internal): New.
3486 (sibcall_register): Remove.
3487 (sibcall_immediate): Remove.
3488 (sibcall_value_internal): New.
3489 (sibcall_value_register): Remove.
3490 (sibcall_value_immediate): Remove.
3491 * config/nds32/predicates.md (nds32_general_register_operand): New.
3492 (nds32_call_address_operand): New.
3494 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3496 PR rtl-optimization/85167
3497 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3498 bb_defs if *split_p, instead preinitialize it to NULL.
3500 PR tree-optimization/85156
3501 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3502 evaluating the argument multiple times.
3504 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3506 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3508 (_mm_cvtpd_ps): Likewise.
3509 (_mm_cvttpd_epi32): Likewise.
3510 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3511 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3512 vector, pixel, and bool following altivec.h include.
3514 2018-04-03 Martin Sebor <msebor@redhat.com>
3516 * doc/extend.texi (Common Function Attributes): Clarify.
3517 (const attribute): Likewise.
3518 (pure attribute): Likewise.
3520 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3523 * config/i386/i386.c (ix86_expand_vector_set): Use
3524 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3526 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3528 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3529 instructions when changing rounding bits to preserve precision bits
3530 in the x87 control word.
3532 2018-04-03 Martin Liska <mliska@suse.cz>
3534 PR tree-optimization/82491
3535 * rtl.h (strip_offset_and_add): Replace += suboffset with
3536 poly_uint64 () + suboffset.
3538 2018-03-29 Martin Liska <mliska@suse.cz>
3539 Martin Jambor <mjambor@suse.cz>
3542 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3543 param_type is not an integral or pointer type.
3545 2018-04-03 Richard Biener <rguenther@suse.de>
3547 * sese.h (recompute_all_dominators): Remove.
3549 2018-04-02 Martin Sebor <msebor@redhat.com>
3551 * doc/invoke.texi (-Wrestrict): Fix typos.
3553 2018-04-02 Jim Wilson <jimw@sifive.com>
3555 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3556 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3557 (<optab>di3, <optab>si3_extend): Likewise.
3558 (<optab>si3_mask, <optab>si3_mask_1): New.
3559 (<optab>di3_mask, <optab>di3_mask_1): New.
3560 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3561 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3562 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3564 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3566 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3569 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3571 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3572 (nds32_canonicalize_comparison): New function.
3574 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3575 Kito Cheng <kito.cheng@gmail.com>
3576 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3578 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3579 * config/nds32/constants.md (unspec_volatile_element): Add
3580 UNSPEC_VOLATILE_RELAX_GROUP.
3581 * config/nds32/nds32-relax-opt.c: New file.
3582 * config/nds32/nds32-predicates.c
3583 (nds32_symbol_load_store_p): New function.
3584 * config/nds32/nds32-protos.h
3585 (nds32_symbol_load_store_p): Declare function.
3586 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3587 * config/nds32/nds32.c
3588 (nds32_register_pass): New function to register pass.
3589 (nds32_register_passes): New function to register passes.
3590 * config/nds32/nds32.md (relax_group): New pattern.
3591 * config/nds32/nds32.opt (mrelax-hint): New option.
3592 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3594 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3596 * config/nds32/t-nds32: Modify files dependency.
3598 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3600 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3601 (PROFILE_HOOK): Define its implementation.
3603 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3605 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3606 type and 32-bit size.
3608 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3611 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3612 (V_128_256): New mode iterator.
3613 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3614 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3615 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3617 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3618 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3620 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3623 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3624 NaN inputs correctly.
3626 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3629 * config/rs6000/vsx.md (??r): New mode attribute.
3630 (*vsx_mov<mode>_64bit): Use it.
3631 (*vsx_mov<mode>_32bit): Likewise.
3633 2018-03-30 Martin Sebor <msebor@redhat.com>
3635 PR tree-optimization/84818
3636 * builtins.c (check_access): Use warning_n.
3638 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3641 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3643 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3646 2018-03-30 Julia Koval <julia.koval@intel.com>
3649 * x86-tune.def (movx, partial_reg_dependency): Enable for
3652 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3655 * lra-constraints.c (process_alt_operands): Move setting
3656 this_alternative_matches below.
3658 2018-03-29 Martin Liska <mliska@suse.cz>
3661 * doc/invoke.texi: Document how LTO works with debug info.
3662 Describe auto-load support of binutils. Mention 'x86-64'
3663 as valid option value of -march option.
3665 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3667 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3670 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3671 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3672 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3675 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3678 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3679 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3680 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3681 for RS6000_BTM_POWERPC64.
3682 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3683 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3684 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3689 2018-03-28 Carl Love <cel@us.ibm.com>
3692 2017-09-27 Carl Love <cel@us.ibm.com>
3694 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3695 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3696 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3699 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3701 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3702 instead of __vector bool.
3703 (_mm_max_pu8): Likewise.
3704 (_mm_min_pi16): Likewise.
3706 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3709 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3710 (DIVWEUO): Likewise.
3712 (DIVDEUO): Likewise.
3713 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3714 DIVWEUO and DIVDEUO.
3715 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3716 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3717 (div_extend): Likewise.
3718 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3720 (__builtin_divweuo): Likewise.
3721 (__builtin_divdeo): Likewise.
3722 (__builtin_divdeuo): Likewise.
3724 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3727 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3728 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3730 PR tree-optimization/82004
3731 * gimple-match-head.c (optimize_pow_to_exp): New function.
3732 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3733 Don't fold to exp if optimize_pow_to_exp is false.
3735 2018-03-28 Martin Liska <mliska@suse.cz>
3738 * calls.c (initialize_argument_information): Fix trailing space.
3739 * common.opt: Fix typo and provide better explanation for
3740 -fsanitize-coverage option.
3741 * config/i386/i386.opt: Fix typo.
3743 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3744 Martin Liska <mliska@suse.cz>
3747 * gimplify.c (asan_poison_variable): Don't do the check for
3748 gimplify_omp_ctxp here.
3749 (gimplify_decl_expr): Do it here.
3750 (gimplify_target_expr): Likewise.
3752 2018-03-28 Martin Liska <mliska@suse.cz>
3755 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3756 chkp_type_bounds_count if MPX is not enabled.
3758 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3760 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3762 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3765 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3766 function to create the function decl for complex long double
3767 multiply and divide for -mabi=ieeelongdouble.
3768 (init_float128_ieee): Call it.
3770 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3773 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3774 -fcf-protection=branch -mibt.
3775 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3777 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3780 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3782 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3785 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3786 extern array declarations.
3788 2018-03-27 Richard Biener <rguenther@suse.de>
3791 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3792 explicit single_use checks.
3794 2018-03-27 Richard Biener <rguenther@suse.de>
3796 PR tree-optimization/85082
3797 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3800 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3802 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3803 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3804 Turn on fasynchronous-unwind-tables and funwind-tables.
3806 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3809 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3810 (*bmi_blsr_<mode>_ccz): Ditto.
3812 2018-03-26 Tom de Vries <tom@codesourcery.com>
3814 PR tree-optimization/85063
3815 * omp-general.c (offloading_function_p): New function. Factor out
3817 * omp-offload.c (pass_omp_target_link::gate): ... here.
3818 * omp-general.h (offloading_function_p): Declare.
3819 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3820 with attribute omp declare target for offloading functions.
3822 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3824 PR tree-optimization/84005
3825 * tree-data-ref.h (get_base_for_alignment): Declare.
3826 * tree-data-ref.c (get_base_for_alignment_1): New function.
3827 (get_base_for_alignment): Likewise.
3828 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3829 get_base_for_alignment to find a suitable base object, instead
3830 of always using drb->base_address.
3832 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3835 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3836 known size by default.
3838 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3841 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3842 and non BLKmode operands.
3844 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3847 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3848 Clean up attributes.
3850 2018-03-23 Richard Biener <rguenther@suse.de>
3853 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3854 we are going to emit early debug for LTO.
3856 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3859 * function.c (match_asm_constraints_1): Don't optimize if input
3860 doesn't satisfy general_operand predicate for output's mode.
3863 * alias.c (write_dependence_p): Don't require for x_canonicalized
3864 non-VOIDmode if x has VOIDmode.
3867 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3868 just don't try to optimize it rather than assert it never happens.
3870 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3872 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3873 macro expansions for definition of ST_INTERNAL_<mode> and
3874 LD_INTERNAL_<mode> builtins.
3875 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3877 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3879 (altivec_expand_st_builtin): Likewise.
3880 (altivec_expand_builtin): Remove calls to deleted functions.
3881 (rs6000_address_for_altivec): Delete this function.
3882 * config/rs6000/vector.md: Remove expands for
3883 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3885 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3888 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3889 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3890 re-computing once computed.
3891 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3892 (arm_init_machine_status): Initialize
3893 machine->static_chain_stack_bytes.
3895 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3898 * doc/extend.texi: Add four new prototypes for vec_ld.
3899 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3900 definitions for more logical presentation.
3901 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3902 entries for V1TI variants of __builtin_altivec_ld builtin.
3903 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3904 handling of V1TI variant of LVX icode pattern.
3905 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3906 (rs6000_gimple_fold_builtin): Likewise.
3907 (altivec_init_builtins): Add code to define
3908 __builtin_altivec_lvx_v1ti function.
3910 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3913 * function.c (match_asm_constraints_1): Don't do the optimization
3914 if input isn't a REG, SUBREG, MEM or constant.
3916 2018-03-22 Tom de Vries <tom@codesourcery.com>
3918 PR tree-optimization/84956
3919 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3920 bb_has_abnormal_pred.
3922 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3925 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3926 DECL_INITIAL (decl) to decl at the end.
3927 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3930 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3932 * doc/extend.texi (__builtin_tgmath): Document when complex
3933 integer types are treated as _Complex _Float64.
3935 2018-03-21 Tom de Vries <tom@codesourcery.com>
3937 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3939 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3941 PR tree-optimization/84960
3942 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3943 if it is ENTRY block, move them into single succ of ENTRY in that case.
3945 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3947 PR tree-optimization/84811
3948 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3949 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3950 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3951 (known_subrange_p): Update accordingly. Cast each value involved
3952 in the size comparison, rather than casting the result of the
3955 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3957 PR tree-optimization/84982
3958 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3959 by flipping the least significant bit rather than all bits from
3960 bitpos to bitpos + bitsize - 1.
3962 2018-03-21 Nathan Sidwell <nathan@acm.org>
3964 * doc/extend.texi (Deprecated Features): Remove mention of
3965 long-deleted deprecations.
3967 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3970 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3971 * configure: Regenerate.
3973 2018-03-21 Tom de Vries <tom@codesourcery.com>
3975 PR tree-optimization/83126
3976 * tree-parloops.c (num_phis): New function.
3977 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3979 2018-03-21 Nathan Sidwell <nathan@acm.org>
3981 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3982 mention anon-struct/union members and trailing attributes.
3984 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3986 PR tree-optimization/84969
3987 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3988 builtin memset partitions if they set different rhs values.
3990 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3992 PR rtl-optimization/84989
3993 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3994 VEC_DUPLICATE with scalar result mode.
3996 2018-03-21 Martin Liska <mliska@suse.cz>
3999 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4000 not intended return statement.
4002 2018-03-21 Martin Liska <mliska@suse.cz>
4005 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4006 (chkp_find_bound_slots_1): Limit number of iterations.
4008 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4011 * Minor grammar fixes for x86 options.
4013 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4016 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4017 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4019 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4022 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4023 (altivec_lvx_<mode>_1op): Likewise.
4024 (altivec_stvx_<mode>_2op): Likewise.
4025 (altivec_stvx_<mode>_1op): Likewise.
4026 (altivec_lvx_<VM2:mode>): New define_expand.
4027 (altivec_stvx_<VM2:mode>): Likewise.
4028 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4029 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4030 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4031 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4032 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4033 (rs6000_gen_lvx): Likewise.
4034 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4035 (altivec_expand_stv_builtin): Likewise.
4036 (altivec_expand_builtin): Likewise.
4037 * config/rs6000/vector.md: Likewise.
4039 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4042 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4045 2018-03-20 Richard Biener <rguenther@suse.de>
4048 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4049 sign-conversions as zero, fall back to standard scalar_stmt
4052 2018-03-20 Martin Liska <mliska@suse.cz>
4055 * predict.c (rebuild_frequencies): Handle case when we have
4056 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4058 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4061 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4062 flag_section_anchors.
4063 * varasm.c (use_blocks_for_decl_p): Remove hack for
4064 dw2_force_const_mem.
4067 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4069 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4070 be created, use lowpart_subreg of operands[0] rather than operands[0]
4072 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4073 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4074 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4075 and n constraint instead of aarch64_shift_imm_di and Usd.
4076 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4077 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4079 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4082 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4083 to favor GPR over NEON registers.
4084 (<shift>di3_neon): Likewise.
4086 2018-03-20 Tom de Vries <tom@codesourcery.com>
4089 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4090 (nvptx_process_pars): Emit bar.sync asap and alap.
4092 2018-03-20 Tom de Vries <tom@codesourcery.com>
4095 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4096 seen_label if seen_label is already set.
4098 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4101 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4102 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4103 Use 1U instead of 1. Formatting fixes.
4106 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4107 instead of TREE_TYPE (s1) for the return value.
4109 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4111 PR tree-optimization/84946
4112 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4113 bitsize + bitsize in poly_uint64 rather than poly_int64.
4116 * dwarf2asm.c: Include fold-const.c.
4117 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4118 of decl rather than decl itself.
4120 PR rtl-optimization/84643
4121 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4123 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4126 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4127 calling assemble_variable.
4129 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4132 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4133 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4135 2018-03-19 Jim Wilson <jimw@sifive.com>
4138 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4139 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4140 (riscv_first_stack_step): Likewise.
4141 (riscv_option_override): Use STACK_BOUNDARY instead of
4143 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4145 (BIGGEST_ALIGNMENT): Set to 128.
4146 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4147 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4150 2018-03-19 Richard Biener <rguenther@suse.de>
4152 PR tree-optimization/84933
4153 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4154 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4156 2018-03-19 Richard Biener <rguenther@suse.de>
4158 PR tree-optimization/84859
4159 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4160 (cond_if_else_store_replacement): Perform sinking operation on
4161 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4162 Generalize what a BB with a single eligible store is.
4164 2018-03-19 Richard Biener <rguenther@suse.de>
4166 PR tree-optimization/84929
4167 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4168 chrec_is_positive against non-chrec arg.
4170 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4173 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4175 2018-03-18 Martin Liska <mliska@suse.cz>
4177 PR rtl-optimization/84635
4178 * regrename.c (build_def_use): Use matches_mode only when
4181 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4183 PR tree-optimization/84913
4184 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4185 vectorize chains of COND_EXPRs.
4187 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4189 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4191 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4193 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4195 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4197 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4199 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4200 Kito Cheng <kito.cheng@gmail.com>
4202 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4203 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4204 (nds32_adjust_reg_alloc_order): New function.
4205 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4207 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4209 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4210 nds32_print_operand, nds32_print_operand_address): Use
4211 HOST_WIDE_INT_PRINT_DEC instead.
4213 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4215 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4217 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4220 * config/i386/i386.c (initial_ix86_tune_features,
4221 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4223 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4224 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4225 rather than 1u << ix86_tune. Formatting fix.
4226 (ix86_option_override_internal): Change ix86_arch_mask from
4227 unsigned int to unsigned HOST_WIDE_INT, initialize to
4228 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4229 (ix86_function_specific_restore): Likewise.
4231 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4234 * postreload.c (reload_combine_recognize_pattern): Perform
4235 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4236 truncate_int_for_mode the result for the destination's mode.
4239 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4240 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4242 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4245 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4248 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4249 regno_allocno_class_array and sorted_pseudos.
4250 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4251 insns where regno is used.
4253 2018-03-16 Martin Liska <mliska@suse.cz>
4256 * multiple_target.c (create_dispatcher_calls): Redirect
4257 reference in the symbol table.
4259 2018-03-16 Martin Liska <mliska@suse.cz>
4262 * multiple_target.c (create_dispatcher_calls): Redirect also
4265 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4269 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4270 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4271 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4273 2018-03-16 Julia Koval <julia.koval@intel.com>
4275 * doc/invoke.texi (Skylake Server): Add CLWB.
4276 Cannonlake): Remove CLWB.
4278 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4280 PR tree-optimization/84841
4281 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4283 (FLOAT_ONE_CONST_TYPE): Define.
4284 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4285 (sort_by_operand_rank): Put entries with higher constant_type last
4286 rather than first to match comments.
4288 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4290 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4293 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4296 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4297 check and use address of target always.
4299 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4302 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4303 (indirect_thunk_bnd_needed): Likewise.
4304 (indirect_thunks_used): Likewise.
4305 (indirect_thunks_bnd_used): Likewise.
4306 (indirect_return_needed): New.
4307 (indirect_return_bnd_needed): Likewise.
4308 (output_indirect_thunk_function): Add a bool argument for
4310 (output_indirect_thunk_function): Don't generate alias for
4311 function return thunk.
4312 (ix86_code_end): Call output_indirect_thunk_function to generate
4313 function return thunks.
4314 (ix86_output_function_return): Set indirect_return_bnd_needed
4315 and indirect_return_needed instead of indirect_thunk_bnd_needed
4316 and indirect_thunk_needed.
4318 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4320 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4321 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4322 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4324 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4325 Paul Hua <paul.hua.gm@gmail.com>
4328 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4330 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4332 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4333 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4336 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4339 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4340 instead of GET_MODE_SIZE when comparing Units.
4342 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4345 * varasm.c (hash_section): Return an unchangeble hash value
4346 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4347 Return !aarch64_can_use_per_function_literal_pools_p ().
4349 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4352 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4353 rather than address of cmode as last argument to prepare_cmp_insn.
4355 2018-03-15 Julia Koval <julia.koval@intel.com>
4357 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4358 F_AVX512VNNI, F_AVX512BITALG): New.
4360 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4363 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4364 insn for floating-point loads and stores.
4366 2018-03-14 Carl Love <cel@us.ibm.com>
4368 * config/rs6000/rs6000-c.c: Add macro definitions for
4369 ALTIVEC_BUILTIN_VEC_PERMXOR.
4370 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4371 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4372 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4373 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4375 * config/doc/extend.texi: Add prototypes for vec_permxor.
4377 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4380 * diagnostic-show-locus.c (class layout_point): Convert m_line
4381 from int to linenum_type.
4382 (line_span::comparator): Use linenum "compare" function when
4383 comparing line numbers.
4384 (test_line_span): New function.
4385 (layout_range::contains_point): Convert param "row" from int to
4387 (layout_range::intersects_line_p): Likewise.
4388 (layout::will_show_line_p): Likewise.
4389 (layout::print_source_line): Likewise.
4390 (layout::should_print_annotation_line_p): Likewise.
4391 (layout::print_annotation_line): Likewise.
4392 (layout::print_leading_fixits): Likewise.
4393 (layout::annotation_line_showed_range_p): Likewise.
4394 (struct line_corrections): Likewise for field m_row.
4395 (line_corrections::line_corrections): Likewise for param "row".
4396 (layout::print_trailing_fixits): Likewise.
4397 (layout::get_state_at_point): Likewise.
4398 (layout::get_x_bound_for_row): Likewise.
4399 (layout::print_line): Likewise.
4400 (diagnostic_show_locus): Likewise for locals "last_line" and
4402 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4403 * input.c (selftest::test_linenum_comparisons): New function.
4404 (selftest::input_c_tests): Call it.
4405 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4406 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4407 * selftest.h (ASSERT_GT): New macro.
4408 (ASSERT_GT_AT): New macro.
4409 (ASSERT_LT): New macro.
4410 (ASSERT_LT_AT): New macro.
4412 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4414 PR rtl-optimization/84780
4415 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4417 2018-03-14 Martin Liska <mliska@suse.cz>
4419 * tree.c (record_node_allocation_statistics): Use
4420 get_stats_node_kind.
4421 (get_stats_node_kind): New function extracted from
4422 record_node_allocation_statistics.
4423 (free_node): Use get_stats_node_kind.
4425 2018-03-14 Richard Biener <rguenther@suse.de>
4427 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4428 that the value-set of ANTIC_IN doesn't grow.
4431 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4433 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4434 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4436 2018-03-14 Julia Koval <julia.koval@intel.com>
4438 * config.gcc (icelake-client, icelake-server): New.
4440 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4441 (initial_ix86_arch_features): Ditto.
4442 (PTA_SKYLAKE): Add SGX.
4443 (PTA_ICELAKE): Remove.
4444 (PTA_ICELAKE_CLIENT): New.
4445 (PTA_ICELAKE_SERVER): New.
4446 (ix86_option_override_internal): Split up icelake on icelake client and
4448 (get_builtin_code_for_version): Ditto.
4449 (fold_builtin_cpu): Ditto.
4450 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4451 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4452 * config/i386/i386.h (processor_type): Ditto.
4453 * doc/invoke.texi: Ditto.
4455 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4458 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4459 INTEGER_CST offset, add it together with bitpos / 8 and
4460 sign extend based on POINTER_SIZE.
4464 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4467 * config/i386/constraints.md (Yc): New register constraint.
4468 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4469 Use Yc constraint for alternative 2 of operand 0. Remove
4470 preferred_for_speed attribute.
4472 2018-03-14 Richard Biener <rguenther@suse.de>
4474 PR tree-optimization/84830
4475 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4476 with the old one to avoid oscillations.
4478 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4481 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4483 (assign_by_spills): Return a flag of reload assignment failure.
4484 Do not process the reload assignment failures. Do not spill other
4485 reload pseudos if they has the same reg class. Update n if
4487 (lra_assign): Add a return arg. Set up from the result of
4488 assign_by_spills call.
4489 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4490 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4491 usage_insns if it is not NULL.
4492 (spill_hard_reg_in_range): New function.
4493 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4494 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4495 function prototypes.
4496 (lra_assign): Change prototype.
4497 * lra.c (lra): Add code to deal with fails by splitting hard reg
4500 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4502 * config/riscv/riscv.opt (mrelax): New option.
4503 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4504 "norelax" when riscv_mrelax is disabled.
4505 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4507 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4510 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4511 reassociation for int modes.
4513 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4515 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4516 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4518 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4519 * config/aarch64/aarch64-sve.md
4520 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4521 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4522 (*extend<mode><Vwide>2): Rename to...
4523 (aarch64_sve_extend<mode><Vwide>2): ...this.
4524 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4525 renaming the old pattern to...
4526 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4528 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4529 define_expand, renaming the old pattern to...
4530 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4531 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4532 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4533 account when deciding which SVE instruction the optab should use.
4534 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4536 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4538 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4539 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4540 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4541 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4542 (tlsdesc_small_<mode>): Turn a define_expand and use
4543 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4544 (tlsdesc_small_advsimd_<mode>): ...this.
4545 (tlsdesc_small_sve_<mode>): New pattern.
4547 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4549 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4550 (UNSPEC_UMUL_HIGHPART): New constants.
4551 (MUL_HIGHPART): New int iteraor.
4552 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4553 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4555 (*<su>mul<mode>3_highpart): New define_insn.
4557 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4560 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4563 2018-03-13 Martin Liska <mliska@suse.cz>
4566 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4568 (sem_item_optimizer::~sem_item_optimizer): Release it.
4569 (sem_item_optimizer::merge_classes): Register variable aliases.
4570 (sem_item_optimizer::fixup_pt_set): New function.
4571 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4572 * ipa-icf.h: Declare new variables and functions.
4574 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4577 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4578 integer_pow2p@2 and test integer_pow2p in condition.
4579 (A < 0 ? C : 0): Similarly for @1.
4582 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4583 characters starting at p contain '\0' character, don't look beyond
4587 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4588 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4591 * reg-stack.c (change_stack): Change update_end var from int to
4592 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4593 also call set_block_for_insn on the newly added insns and rescan.
4596 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4597 on the last operand.
4600 * tree.c (stabilize_reference_1): Return save_expr (e) for
4601 STATEMENT_LIST even if it doesn't have side-effects.
4603 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4605 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4607 2018-03-12 Renlin Li <renlin.li@arm.com>
4609 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4610 aarch64_output_scalar_simd_mov_immediate.
4612 2018-03-12 Martin Sebor <msebor@redhat.com>
4614 PR tree-optimization/83456
4615 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4616 for perfectly overlapping calls to memcpy.
4617 (gimple_fold_builtin_memory_chk): Same.
4618 (gimple_fold_builtin_strcpy): Handle no-warning.
4619 (gimple_fold_builtin_stxcpy_chk): Same.
4620 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4622 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4624 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4625 parameter. Use it for SFmode.
4626 (rs6000_function_arg_advance_1): Adjust.
4627 (rs6000_function_arg): Adjust.
4628 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4630 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4632 PR rtl-optimization/84169
4633 PR rtl-optimization/84780
4634 * combine.c (can_combine_p): Check for a 2-insn combination whether
4635 the destination register is used between the two insns, too.
4637 2018-03-12 Richard Biener <rguenther@suse.de>
4639 PR tree-optimization/84803
4640 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4641 for refs DR analysis didn't process.
4643 2018-03-12 Richard Biener <rguenther@suse.de>
4645 PR tree-optimization/84777
4646 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4647 force-vectorize loops ignore whether we are optimizing for size.
4649 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4651 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4652 (TARGET_MD_ASM_ADJUST): Define.
4654 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4655 Kito Cheng <kito.cheng@gmail.com>
4656 Chung-Ju Wu <jasonwucj@gmail.com>
4658 * config/nds32/nds32.c (nds32_compute_stack_frame,
4659 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4660 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4661 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4662 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4663 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4664 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4665 * config/nds32/nds32.md (prologue, epilogue): Use macro
4666 NDS32_V3PUSH_AVAILABLE_P to do checking.
4668 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4671 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4672 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4673 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4674 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4675 addition of most attributes on !orig_type_die or the attribute not
4676 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4678 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4679 Chung-Ju Wu <jasonwucj@gmail.com>
4681 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4683 * config/nds32/nds32.opt (mvh): New option.
4685 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4686 Chung-Ju Wu <jasonwucj@gmail.com>
4688 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4690 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4691 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4694 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4695 Chung-Ju Wu <jasonwucj@gmail.com>
4697 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4699 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4700 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4702 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4703 Kito Cheng <kito.cheng@gmail.com>
4704 Chung-Ju Wu <jasonwucj@gmail.com>
4706 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4707 UNSPEC_FFMISM and UNSPEC_FLMISM.
4708 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4709 for ffb, ffmism and flmism.
4710 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4711 (unspec_ffmism): Ditto.
4712 (unspec_flmism): Ditto.
4713 (nds32_expand_builtin_impl): Check if string extension is available.
4714 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4715 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4717 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4720 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4723 * lra-assigns.c (assign_by_spills): Return a flag of reload
4724 assignment failure. Do not process the reload assignment
4725 failures. Do not spill other reload pseudos if they has the same
4727 (lra_assign): Add a return arg. Set up from the result of
4728 assign_by_spills call.
4729 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4730 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4731 usage_insns if it is not NULL.
4732 (spill_hard_reg_in_range): New function.
4733 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4734 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4735 function prototypes.
4736 (lra_assign): Change prototype.
4737 * lra.c (lra): Add code to deal with fails by splitting hard reg
4740 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4743 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4745 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4748 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4749 (dw_val_node): Add val_symbolic_view.
4750 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4751 (symview_upper_bound): New.
4752 (new_line_info_table): Initialize symviews_since_reset.
4753 (dwarf2out_source_line): Count symviews_since_reset and set
4754 symview_upper_bound.
4755 (dw_val_equal_p): Handle symview.
4756 (add_AT_symview): New.
4757 (print_dw_val): Handle symview.
4758 (attr_checksum, attr_checksum_ordered): Likewise.
4759 (same_dw_val_p, size_of_die): Likewise.
4760 (value_format, output_die): Likewise.
4761 (add_high_low_attributes): Use add_AT_symview for entry_view.
4762 (dwarf2out_finish): Reset symview_upper_bound, clear
4765 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4768 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4769 Add strict argument and use it.
4770 (rs6000_split_multireg_move): Update for new strict argument.
4771 (mem_operand_gpr): Disallow all non-offsettable addresses.
4772 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4774 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4777 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4778 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4779 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4782 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4783 decl, use remap_type if we want to use the type.
4785 2018-03-09 Martin Sebor <msebor@redhat.com>
4787 PR tree-optimization/84526
4788 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4790 (builtin_access::generic_overlap): Be prepared to handle non-array
4793 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4795 PR rtl-optimization/84682
4796 * lra-constraints.c (process_address_1): Check is_address flag
4797 for address constraints.
4798 (process_alt_operands): Likewise.
4799 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4800 preprocess_constraints.
4801 * recog.h (preprocess_constraints): Add oploc parameter.
4803 * recog.c (preprocess_constraints): Test address_operand for
4804 CT_ADDRESS constraints.
4806 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4809 * lra-assigns.c (assign_by_spills): Return a flag of reload
4810 assignment failure. Do not process the reload assignment
4811 failures. Do not spill other reload pseudos if they has the same
4813 (lra_assign): Add a return arg. Set up from the result of
4814 assign_by_spills call.
4815 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4816 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4817 usage_insns if it is not NULL.
4818 (spill_hard_reg_in_range): New function.
4819 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4820 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4821 function prototypes.
4822 (lra_assign): Change prototype.
4823 * lra.c (lra): Add code to deal with fails by splitting hard reg
4826 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4829 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4830 Accept complain bool parameter. Only emit errors if it is true.
4831 (arm_parse_cpu_option_name): Likewise.
4832 (arm_target_thumb_only): Adjust callers of the above.
4833 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4834 prototype to take a default true bool parameter.
4835 (arm_parse_arch_option_name): Likewise.
4837 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4838 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4842 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4843 * configure: Regenerate.
4844 * configure.ac ("linker --version-script option"): New.
4845 ("linker soname option"): New.
4847 2018-03-09 Richard Biener <rguenther@suse.de>
4849 PR tree-optimization/84775
4850 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4851 immediate uses of predicate stmts and mark them modified.
4854 PR tree-optimization/84178
4855 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4857 (version_loop_for_if_conversion): Delay update_ssa call.
4858 (tree_if_conversion): Delay update_ssa until after predicate
4861 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4864 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4865 when the function accesses prior frames.
4867 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4870 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4871 gen_llsym, otherwise call maybe_gen_llsym.
4874 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4875 has ',' character inside of it.
4877 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4880 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4881 as clobbering CC_REGNUM.
4883 2018-03-08 Richard Biener <rguenther@suse.de>
4886 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4887 (follow_copies_to_constant): Do not follow SSA names registered
4890 2018-03-08 Richard Biener <rguenther@suse.de>
4892 PR tree-optimization/84178
4893 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4895 (version_loop_for_if_conversion): Delay update_ssa call.
4896 (tree_if_conversion): Delay update_ssa until after predicate
4899 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4901 PR tree-optimization/84178
4902 * tree-if-conv.c (release_bb_predicate): Remove the
4903 the assertion that the stmts have NULL use_ops.
4904 Discard the statements, asserting that they haven't
4905 yet been added to a BB.
4907 2018-03-08 Richard Biener <rguenther@suse.de>
4909 PR tree-optimization/84746
4910 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4911 (phi_translate): Pass in destination ANTIC_OUT set.
4912 (phi_translate_1): Likewise. For a simplified result lookup
4913 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4914 (phi_translate_set): Adjust.
4915 (do_pre_regular_insertion): Likewise.
4916 (do_pre_partial_partial_insertion): Likewise.
4918 2018-03-08 Martin Liska <mliska@suse.cz>
4920 PR gcov-profile/84735
4921 * doc/gcov.texi: Document usage of profile files.
4922 * gcov-io.h: Document changes in the format.
4924 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4928 * dwarf2out.c (struct dw_line_info_table): Update comments for
4930 (FORCE_RESET_NEXT_VIEW): New.
4931 (FORCE_RESETTING_VIEW_P): New.
4932 (RESETTING_VIEW_P): Check for -1 too.
4933 (ZERO_VIEW_P): Likewise.
4934 (new_line_info_table): Force-reset next view.
4935 (dwarf2out_begin_function): Likewise.
4936 (dwarf2out_source_line): Simplify zero_view_p initialization.
4937 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4938 view directly. Omit view when omitting .loc at line 0.
4940 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4942 PR tree-optimization/84740
4943 * tree-switch-conversion.c (process_switch): Call build_constructors
4944 only if info.phi_count is non-zero.
4946 PR tree-optimization/84739
4947 * tree-tailcall.c (find_tail_calls): Check call arguments against
4948 DECL_ARGUMENTS (current_function_decl) rather than
4949 DECL_ARGUMENTS (func) when checking for tail recursion.
4951 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4953 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4954 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4955 Volker Reichelt's entry and add entries for people that perform
4956 GCC fuzzy testing and report numerous bugs.
4958 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4961 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4962 readonly data in sdata, if that is disabled.
4963 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4964 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4965 -mreadonly-in-sdata option.
4967 2018-03-07 Martin Sebor <msebor@redhat.com>
4969 PR tree-optimization/84468
4970 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4971 basic block when looking for nul assignment.
4973 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4976 * except.h (output_function_exception_table): Adjust prototype.
4977 * except.c (output_function_exception_table): Remove FNNAME parameter
4978 and add SECTION parameter. Ouput one part of the table at a time.
4979 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4980 the first part of the exception table and emit unwind directives.
4981 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4982 (i386_pe_seh_cold_init): Likewise.
4983 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4984 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4985 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4986 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4987 * config/i386/winnt.c: Include except.h.
4988 (struct seh_frame_state): Add reg_offset, after_prologue and
4989 in_cold_section fields.
4990 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4991 (i386_pe_seh_cold_init): New function.
4992 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4993 to seh->in_cold_section.
4994 (seh_emit_push): Record the offset of the push.
4995 (seh_emit_save): Record the offet of the save.
4996 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4997 Test seh->after_prologue to disregard the epilogue.
4998 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4999 (i386_pe_end_cold_function): New function.
5001 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5004 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5005 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5008 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5010 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5011 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5014 * multiple_target.c: Include tree-inline.h and intl.h.
5015 (expand_target_clones): Diagnose and fail if node->definition and
5016 !tree_versionable_function_p (node->decl).
5018 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5020 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5022 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5023 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5024 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5026 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5029 * combine.c (try_combine): Use reg_or_subregno instead of handling
5030 just paradoxical SUBREGs and REGs.
5032 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5034 * config/arc/arc.c (arc_finalize_pic): Remove function.
5035 (arc_must_save_register): We use single base PIC register, remove
5036 checks to save/restore the PIC register.
5037 (arc_expand_prologue): Likewise.
5038 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5040 (arc_verify_short): Likewise.
5041 (arc_attr_type): Likewise.
5042 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5043 (walk_stores): Likewise.
5044 (arc_address_cost): Make it static.
5045 (arc_verify_short): Likewise.
5046 (branch_dest): Likewise.
5047 (arc_attr_type): Likewise.
5048 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5049 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5050 (arc_final_prescan_insn): Remove inserting the nops due to
5051 hardware hazards. It is done in reorg step.
5052 (insn_length_variant_t): Remove.
5053 (insn_length_parameters_t): Likewise.
5054 (arc_insn_length_parameters): Likewise.
5055 (arc_get_insn_variants): Likewise.
5056 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5058 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5061 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5064 PR tree-optimization/84687
5065 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5067 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5069 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5071 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5072 Rename to ppc_speculation_barrier.
5073 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5074 __builtin_ppc_speculation_barrier.
5076 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5079 * combine.c (combine_simplify_rtx): Don't try to simplify if
5080 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5083 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5085 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5086 to 32 bytes when compiling for POWER9.
5088 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5091 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5092 regparm >= 3 with no arg reg available also for calls with
5093 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5096 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5098 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5100 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5103 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5105 2018-03-05 Richard Biener <rguenther@suse.de>
5107 PR tree-optimization/84486
5108 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5109 When inserting a __builtin_assume_aligned call set the LHS
5110 SSA name alignment info accordingly.
5112 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5114 PR tree-optimization/84114
5115 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5116 Avoid reassociation of FLOAT_MODE addition.
5118 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5120 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5121 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5122 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5123 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5124 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5125 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5126 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5128 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5129 __builtin_ia32_wbinvd): New builtins.
5130 (SPECIAL_ARGS2): New.
5131 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5132 (SPECIAL_ARGS2): New.
5133 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5134 (ix86_valid_target_attribute_inner_p): Ditto.
5135 (ix86_init_mmx_sse_builtins): Add special_args2.
5136 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5137 TARGET_WBNOINVD_P): New.
5138 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5139 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5140 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5141 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5142 * config/i386/pconfigintrin.h: New file.
5143 * config/i386/wbnoinvdintrin.h: Ditto.
5144 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5145 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5147 2018-03-05 Richard Biener <rguenther@suse.de>
5149 PR tree-optimization/84670
5150 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5152 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5153 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5154 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5155 doesn't grow if all successors (recursively) were visited at least
5158 2018-03-05 Richard Biener <rguenther@suse.de>
5160 PR tree-optimization/84650
5161 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5162 if executed in the loop pipeline.
5164 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5166 * doc/configfiles.texi (Configuration Files): Move info about
5167 conditionalizing $target-protos.h to...
5168 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5169 differs from $target-protos.h.
5171 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5172 Chung-Ju Wu <jasonwucj@gmail.com>
5174 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5175 * config/nds32/nds32-multiple.md (setmemsi): Define.
5176 * config/nds32/nds32-memory-manipulation.c
5177 (nds32_gen_dup_4_byte_to_word_value): New.
5178 (emit_setmem_word_loop): New.
5179 (emit_setmem_byte_loop): New.
5180 (nds32_expand_setmem_loop): New.
5181 (nds32_expand_setmem_loop_v3m): New.
5182 (nds32_expand_setmem_unroll): New.
5183 (nds32_expand_setmem): New.
5185 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5186 Chung-Ju Wu <jasonwucj@gmail.com>
5188 * config/nds32/nds32-memory-manipulation.c
5189 (nds32_emit_load_store): New.
5190 (nds32_emit_post_inc_load_store): New.
5191 (nds32_emit_mem_move): New.
5192 (nds32_emit_mem_move_block): New.
5193 (nds32_expand_movmemsi_loop_unknown_size): New.
5194 (nds32_expand_movmemsi_loop_known_size): New.
5195 (nds32_expand_movmemsi_loop): New.
5196 (nds32_expand_movmemsi_unroll): New.
5197 (nds32_expand_movmemqi): Rename ...
5198 (nds32_expand_movmemsi): ... to this.
5199 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5200 (movmemsi): ... to this.
5201 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5202 (nds32_expand_movmemsi): ... to this.
5204 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5205 Monk Chiang <sh.chiang04@gmail.com>
5206 Chung-Ju Wu <jasonwucj@gmail.com>
5208 * config/nds32/nds32-protos.h
5209 (nds32_expand_load_multiple): New arguments.
5210 (nds32_expand_store_multiple): Ditto.
5211 (nds32_valid_multiple_load_store): Rename ...
5212 (nds32_valid_multiple_load_store_p): ... to this.
5213 * config/nds32/nds32-memory-manipulation.c
5214 (nds32_expand_load_multiple): Refine implementation.
5215 (nds32_expand_store_multiple): Ditto.
5216 * config/nds32/nds32-multiple.md
5217 (load_multiple): Update nds32_expand_load_multiple interface.
5218 (store_multiple): Update nds32_expand_store_multiple interface.
5219 * config/nds32/nds32-predicates.c
5220 (nds32_valid_multiple_load_store): Rename ...
5221 (nds32_valid_multiple_load_store_p): ... to this and refine
5223 * config/nds32/predicates.md
5224 (nds32_load_multiple_and_update_address_operation): New predicate.
5225 (nds32_store_multiple_and_update_address_operation): New predicate.
5227 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5228 Chung-Ju Wu <jasonwucj@gmail.com>
5230 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5231 (combo): New attribute.
5232 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5234 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5236 * config/nds32/nds32.opt: Change -mcmodel= default value.
5238 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5239 Monk Chiang <sh.chiang04@gmail.com>
5240 Chung-Ju Wu <jasonwucj@gmail.com>
5242 * config/nds32/constants.md (unspec_element): New enum.
5243 * config/nds32/constraints.md (Umw): New constraint.
5244 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5245 * config/nds32/nds32-intrinsic.md: Likewise.
5246 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5247 (nds32_valid_smw_lwm_base_p): New.
5248 (nds32_output_smw_single_word): New.
5249 (nds32_output_lmw_single_word): New.
5250 (nds32_expand_unaligned_load): New.
5251 (nds32_expand_unaligned_store): New.
5252 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5253 (nds32_output_smw_single_word): Declare.
5254 (nds32_output_lmw_single_word): Declare.
5255 (nds32_expand_unaligned_load): Declare.
5256 (nds32_expand_unaligned_store): Declare.
5257 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5258 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5259 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5260 NDS32_BUILTIN_UASTORE_DW.
5261 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5264 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5265 Kito Cheng <kito.cheng@gmail.com>
5266 Chung-Ju Wu <jasonwucj@gmail.com>
5268 * config/nds32/nds32-intrinsic.c
5269 (nds32_expand_builtin_null_ftype_reg): Delete.
5270 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5271 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5272 (nds32_read_argument): New.
5273 (nds32_legitimize_target): Ditto.
5274 (nds32_legitimize_argument): Ditto.
5275 (nds32_check_constant_argument): Ditto.
5276 (nds32_expand_unop_builtin): Ditto.
5277 (nds32_expand_unopimm_builtin): Ditto.
5278 (nds32_expand_binop_builtin): Ditto.
5279 (nds32_builtin_decl_impl): Ditto.
5280 (builtin_description): Ditto.
5281 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5282 (nds32_init_builtins_impl): Ditto.
5283 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5284 (nds32_builtin_decl): New.
5285 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5286 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5288 2018-03-02 Jeff Law <law@redhat.com>
5290 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5291 (redundant_insn, fill_simple_delay_slots): Likewise.
5292 (fill_slots_from_thread): Likewise.
5293 * resource.c (mark_referenced_resources): Likewise.
5294 (mark_set_resources, find_dead_or_set_registers): Likewise.
5296 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5298 * substring-locations.h (format_warning_va): Formatting fix for
5300 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5302 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5303 * substring-locations.c: Include intl.h.
5304 (format_warning_va): Turned into small wrapper around
5305 format_warning_n_va, renamed to ...
5306 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5307 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5309 (format_warning_at_substring_n): New function.
5310 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5311 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5312 format_warning_at_substring with just a shorter name instead of
5313 const function pointer.
5314 (fmtwarn_n): New function.
5315 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5316 appropriate, get rid of all the fmtstr temporaries, move conditionals
5317 with G_() wrapped string literals directly into fmtwarn arguments,
5318 cast dir.len to (int), formatting fixes.
5320 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5322 * doc/invoke.texi: Remove "Cilk Plus" references.
5324 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5325 Richard Biener <rguenther@suse.de>
5328 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5329 for error or warning attributes if CALL_FROM_THUNK_P is set.
5332 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5335 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5336 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5339 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5340 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5342 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5343 instead of -1U in last predictors element's probability member.
5345 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5348 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5349 arguments if they are comparable.
5351 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5353 PR tree-optimization/84634
5354 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5355 masks and masked_loop_p with a single loop_masks, making sure it's
5356 null for bb vectorization.
5358 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5360 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5361 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5362 loop->force_vectorize to check whether there is no alias.
5364 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5367 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5369 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5371 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5372 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5373 instead of a loop around prev_real_insn.
5374 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5378 * config/i386/i386.c (ix86_print_operand): Use conditional
5379 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5382 2018-03-02 Richard Biener <rguenther@suse.de>
5384 PR tree-optimization/84427
5385 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5386 (bitmap_set_subtract_values): Rewrite to handle multiple
5389 (prune_clobbered_mems): Likewise.
5390 (phi_translate): Take edge instead of pred/phiblock.
5391 (phi_translate_1): Likewise.
5392 (phi_translate_set): Likewise. Insert all translated
5393 exprs for a value into the set, keeping possibly multiple
5394 expressions per value.
5395 (compute_antic_aux): Adjust for phi_translate changes.
5396 When intersecting union the expressions and prune those
5397 not in the final value set, keeping possibly multiple
5398 expressions per value. Do not use value-insertion
5399 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5400 all expressions. Add verification that the value-sets
5401 only shrink during iteration.
5402 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5403 (do_pre_regular_insertion): Likewise.
5404 (do_pre_partial_partial_insertion): Likewise.
5406 2018-03-02 Richard Biener <rguenther@suse.de>
5409 * config/darwin.c (saved_debug_info_level): New static global.
5410 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5411 (darwin_asm_lto_end): Restore debug info generation settings.
5413 2018-03-01 Martin Liska <mliska@suse.cz>
5416 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5419 2018-03-01 Richard Biener <rguenther@suse.de>
5422 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5423 type annotation with LTO when debug was disabled at compile-time.
5425 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5427 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5429 (mips_final_postscan_insn): Likewise.
5431 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5433 PR rtl-optimization/84528
5434 * alias.c (init_alias_target): Add commentary.
5435 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5436 a unique base value if the frame pointer is not eliminated
5437 to the stack pointer.
5439 2018-03-01 Tom de Vries <tom@codesourcery.com>
5441 PR rtl-optimization/83327
5442 * lra-int.h (hard_regs_spilled_into): Declare.
5443 * lra.c (hard_regs_spilled_into): Define.
5444 (init_reg_info): Init hard_regs_spilled_into.
5445 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5446 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5447 (process_bb_lives): Handle hard_regs_spilled_into.
5448 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5449 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5451 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5453 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5454 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5455 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5456 * config/rs6000/aix72.h: New file.
5458 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5460 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5461 instead of warning_at with conditional singular and plural messages
5465 * stor-layout.c (update_alignment_for_field): For
5466 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5467 && !DECL_PACKED (field), do the alignment update, just use
5468 only desired_align instead of MAX (type_align, desired_align)
5470 (place_field): Don't do known_align < desired_align handling
5471 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5472 is non-NULL, instead do it after rli->prev_field handling and
5473 only if not within a bitfield word. For DECL_PACKED (field)
5474 use type_align of BITS_PER_UNIT.
5476 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5478 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5479 superfluous parentheses and trailing spaces.
5481 2018-02-28 Richard Biener <rguenther@suse.de>
5483 PR tree-optimization/84584
5484 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5485 SCoPs with fake exit edge.
5487 2018-02-28 Martin Liska <mliska@suse.cz>
5490 * timevar.c (timer::print): Fix format to properly print 100%
5493 2018-02-28 Richard Biener <rguenther@suse.de>
5496 * genmatch.c (capture_info::walk_match): Do not mark
5497 captured expressions without operands as expr_p given
5498 they act more like predicates and should be subject to
5499 "lost tail" side-effect preserving.
5501 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5503 PR rtl-optimization/81611
5504 * auto-inc-dec.c (attempt_change): Move dead note from
5505 mem_insn if it's the next use of regno
5506 (find_address): Take address use of reg holding
5507 non-incremented value. Add parm to limit search to the named
5509 (merge_in_block): Attempt to use a mem insn that is the next
5510 use of the original regno.
5512 2018-02-27 Martin Sebor <msebor@redhat.com>
5515 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5516 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5518 2018-02-27 Martin Sebor <msebor@redhat.com>
5520 PR translation/84207
5521 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5522 n argument to unsigned HOST_WIDE_INT.
5523 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5524 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5525 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5526 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5528 2018-02-27 Richard Biener <rguenther@suse.de>
5530 PR tree-optimization/84512
5531 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5532 Do not use the estimate returned from record_stmt_cost for
5533 the scalar iteration cost but sum properly using add_stmt_cost.
5535 2018-02-27 Richard Biener <rguenther@suse.de>
5537 PR tree-optimization/84466
5538 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5539 Adjust last change to less strictly validate use operands.
5541 2018-02-27 Martin Liska <mliska@suse.cz>
5543 PR gcov-profile/84548
5544 * gcov.c (process_file): Allow partial overlap and consider it
5545 also as group functions.
5546 (output_lines): Properly calculate range of lines for a group.
5548 2018-02-27 Martin Liska <mliska@suse.cz>
5550 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5551 'ggc' suffixes. Change first column width.
5552 (timer::print): Fix formatting of the column.
5554 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5556 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5557 preserve inline entry blocks for the sake of debug inline
5558 entry point markers alone.
5559 (remove_unused_locals): Suggest in comments a better place to
5560 force the preservation of inline entry blocks that are
5561 otherwise unused, but do not preserve them.
5563 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5565 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5567 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5570 * config/i386/constraints.md (Bs): Replace
5571 ix86_indirect_branch_register with
5572 TARGET_INDIRECT_BRANCH_REGISTER.
5574 * config/i386/i386.md (indirect_jump): Likewise.
5575 (tablejump): Likewise.
5576 (*sibcall_memory): Likewise.
5577 (*sibcall_value_memory): Likewise.
5578 Peepholes of indirect call and jump via memory: Likewise.
5579 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5580 (*sibcall_value_GOT_32): Likewise.
5581 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5582 (GOT_memory_operand): Likewise.
5583 (call_insn_operand): Likewise.
5584 (sibcall_insn_operand): Likewise.
5585 (GOT32_symbol_operand): Likewise.
5586 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5588 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5590 PR rtl-optimization/83496
5591 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5592 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5593 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5594 redundant insn, if any.
5595 (relax_delay_slots): Likewise.
5596 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5598 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5600 PR tree-optimization/83965
5601 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5602 that grouped statements are part of a reduction chain. Return
5603 true if the statement is not marked as a reduction itself but
5605 (vect_recog_dot_prod_pattern): Don't check whether the statement
5606 is part of a group here.
5607 (vect_recog_sad_pattern): Likewise.
5608 (vect_recog_widen_sum_pattern): Likewise.
5610 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5613 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5615 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5618 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5620 (ix86_output_indirect_function_return): New prototype.
5621 (ix86_split_simple_return_pop_internal): Likewise.
5622 * config/i386/i386.c (indirect_return_via_cx): New.
5623 (indirect_return_via_cx_bnd): Likewise.
5624 (indirect_thunk_name): Handle return va CX_REG.
5625 (output_indirect_thunk_function): Create alias for
5626 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5627 (ix86_output_indirect_jmp): Remove the bool argument.
5628 (ix86_output_indirect_function_return): New function.
5629 (ix86_split_simple_return_pop_internal): Likewise.
5630 * config/i386/i386.md (*indirect_jump): Don't pass false
5631 to ix86_output_indirect_jmp.
5632 (*tablejump_1): Likewise.
5633 (simple_return_pop_internal): Change it to define_insn_and_split.
5634 Call ix86_split_simple_return_pop_internal to split it for
5636 (simple_return_indirect_internal): Call
5637 ix86_output_indirect_function_return instead of
5638 ix86_output_indirect_jmp.
5640 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5643 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5644 memset and value initialization afterwards.
5646 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5648 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5650 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5653 * common/config/aarch64/aarch64-common.c
5654 (aarch_option_optimization_table[]): Switch
5655 off fomit-frame-pointer
5657 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5658 Chung-Ju Wu <jasonwucj@gmail.com>
5660 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5662 (store_multiple): Ditto.
5664 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5666 * config.gcc: Add --with-cpu support for nds32 target.
5667 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5668 * config/nds32/nds32.opt: Add -mcpu= option.
5670 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5672 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5673 isel=yes): Warn for these deprecated options.
5675 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5677 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5678 ISA_2_5_MASKS_EMBEDDED.
5680 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5682 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5683 p->max as pointers rather than using iterative_hash_expr.
5685 2018-02-23 Carl Love <cel@us.ibm.com>
5687 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5688 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5690 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5691 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5692 P8V_BUILTIN_VEC_VUNSIGNED2.
5694 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5697 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5698 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5700 * lra-constraints.c (curr_insn_transform): Set up
5701 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5703 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5704 * lra-eliminations.c (spill_pseudos): Ditto.
5705 (process_insn_for_elimination): Ditto.
5706 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5707 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5708 LRA_NON_CLOBBERED_ALT.
5710 2018-02-22 Martin Sebor <msebor@redhat.com>
5712 PR tree-optimization/84480
5713 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5714 to maybe_diag_stxncpy_trunc. Call it.
5715 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5716 from gimple_fold_builtin_strcpy. Print inlining stack.
5717 (handle_builtin_stxncpy): Print inlining stack.
5718 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5720 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5723 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5724 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5725 and -fcheck-pointer-bounds are used together.
5726 (indirect_thunk_prefix): New enum.
5727 (indirect_thunk_need_prefix): New function.
5728 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5729 "_nt" instead of "_bnd" for NOTRACK prefix.
5730 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5731 (output_indirect_thunk_function): Likewise.
5733 (ix86_code_end): Update output_indirect_thunk_function calls.
5734 (ix86_output_indirect_branch_via_reg): Replace
5735 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5736 (ix86_output_indirect_branch_via_push): Likewise.
5737 (ix86_output_function_return): Likewise.
5738 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5739 incompatible with -fcf-protection=branch and
5740 -fcheck-pointer-bounds.
5742 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5745 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5746 Change gcc_assert call to output_operand_lossage.
5748 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5750 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5752 2018-02-22 DJ Delorie <dj@redhat.com>
5753 Sebastian Perta <sebastian.perta@renesas.com>
5754 Oleg Endo <olegendo@gcc.gnu.org>
5756 * config/rx/rx.c (rx_rtx_costs): New function.
5757 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5759 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5761 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5763 2018-02-22 Martin Liska <mliska@suse.cz>
5766 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5767 Add "native" as a possible value.
5769 2018-02-22 Martin Liska <mliska@suse.cz>
5772 * config/i386/i386.c (ix86_option_override_internal):
5773 Add "native" as a possible value for -march and -mtune.
5775 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5778 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5779 to all type variants.
5781 PR tree-optimization/84503
5782 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5783 width as info->bitpos + info->bitsize - start.
5784 (merged_store_group::merge_overlapping): Simplify width computation.
5785 (check_no_overlap): New function.
5786 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5787 start + width and last_order of the group, fail if check_no_overlap
5789 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5790 to group if check_no_overlap fails.
5792 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5794 * config/rs6000/altivec.md: Delete contraint arguments to
5795 define_expand, define_split, and define_peephole2, and in
5796 define_insn_and_split if always unused.
5797 * config/rs6000/darwin.md: Ditto.
5798 * config/rs6000/dfp.md: Ditto.
5799 * config/rs6000/rs6000.md: Ditto.
5800 * config/rs6000/sync.md: Ditto.
5801 * config/rs6000/vector.md: Ditto.
5802 * config/rs6000/vsx.md: Ditto.
5804 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5806 * config/rs6000/altivec.md: Write output control strings as braced
5807 blocks instead of double-quoted strings.
5808 * config/rs6000/darwin.md: Ditto.
5809 * config/rs6000/rs6000.md: Ditto.
5810 * config/rs6000/vector.md: Ditto.
5811 * config/rs6000/vsx.md: Ditto.
5813 2018-02-21 Jason Merrill <jason@redhat.com>
5815 PR c++/84314 - ICE with templates and fastcall attribute.
5816 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5818 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5820 * ipa-cp.c (determine_versionability): Fix comment typos.
5822 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5825 * ipa-cp.c (determine_versionability): Do not version functions caling
5828 2018-02-21 Martin Liska <mliska@suse.cz>
5831 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5832 Add "native" as a possible value.
5833 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5834 the macro when native cpu detection is available.
5836 2018-02-21 Martin Liska <mliska@suse.cz>
5839 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5840 Add "native" as a possible value.
5841 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5842 when native cpu detection is available.
5844 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5845 Martin Sebor <msebor@redhat.com>
5847 PR tree-optimization/84478
5848 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5850 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5851 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5852 support which is conservatively correct, for 2 only stay conservative
5853 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5854 argument to the 2 argument get_range_strlen, adjust 6 arg
5855 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5857 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5858 (gimple_fold_builtin_strlen): Pass true as last argument to
5861 2018-02-20 Martin Sebor <msebor@redhat.com>
5864 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5865 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5866 (builtin_memref::builtin_memref): Factor out parts into
5867 set_base_and_offset and call it.
5869 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5872 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5873 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5874 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5875 search at the associated MODE_INT.
5877 2018-02-20 Jeff Law <law@redhat.com>
5880 PR tree-optimization/81592
5882 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5883 for range data rather than using global data.
5884 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5885 range data rather than using global data.
5886 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5887 pass it to children as needed.
5888 (struct directive::fmtresult): Similarly.
5889 (struct directive::set_width): Similarly.
5890 (struct directive::set_precision): Similarly.
5891 (format_integer, format_directive, parse_directive): Similarly.
5892 (format_none): Accept unnamed vr_values parameter.
5893 (format_percent, format_floating, format_character): Similarly.
5894 (format_string, format_plain): Similarly.
5895 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5896 the EVRP range analyzer for range data rather than using global data.
5897 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5898 gimple-ssa-evrp-analyze.h
5899 (class sprintf_dom_walker): Add after_dom_children member function.
5900 Add evrp_range_analyzer member.
5901 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5902 range analyzer as needed.
5903 (sprintf_dom_walker::after_dom_children): New member function.
5904 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5906 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5907 (evrp_range_analyzer::pop_to_marker): Likewise.
5909 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5911 PR tree-optimization/84419
5912 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5913 with the required type if its current type is compatible but
5916 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5919 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5920 after vectorization.
5922 2018-02-20 Martin Liska <mliska@suse.cz>
5925 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5926 possible values if we don't have a hint.
5928 2018-02-20 Martin Liska <mliska@suse.cz>
5932 * final.c (shorten_branches): Build align_tab array with one
5934 * opts.c (finish_options): Add alignment option limit check.
5935 (MAX_CODE_ALIGN): Likewise.
5936 (MAX_CODE_ALIGN_VALUE): Likewise.
5937 * doc/invoke.texi: Document maximum allowed option value for
5938 all -falign-* options.
5940 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5943 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5944 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5945 * var-tracking.c (emit_note_insn_var_location): Remove all references
5946 to NOTE_INSN_CALL_ARG_LOCATION.
5947 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5948 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5949 Use copy_rtx_if_shared.
5950 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5951 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5952 (dwarf2out_var_location): Remove handling of
5953 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5955 * final.c (final_scan_insn): Remove all references to
5956 NOTE_INSN_CALL_ARG_LOCATION.
5957 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5958 before dumping final insns.
5959 * except.c (emit_note_eh_region_end): Remove all references to
5960 NOTE_INSN_CALL_ARG_LOCATION.
5961 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5962 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5963 * config/arc/arc.c (hwloop_optimize): Likewise.
5964 * config/arm/arm.c (create_fix_barrier): Likewise.
5965 * config/s390/s390.c (s390_chunkify_start): Likewise.
5966 * config/sh/sh.c (find_barrier): Likewise.
5967 * config/i386/i386.c (rest_of_insert_endbranch,
5968 ix86_seh_fixup_eh_fallthru): Likewise.
5969 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5970 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5971 * config/frv/frv.c (frv_function_prologue): Likewise.
5972 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5974 (note_outside_basic_block_p): Remove all references to
5975 NOTE_INSN_CALL_ARG_LOCATION.
5976 * gengtype.c (adjust_field_rtx_def): Likewise.
5977 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5979 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5980 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5983 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5986 PR tree-optimization/84452
5987 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5988 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5991 2018-02-19 Martin Liska <mliska@suse.cz>
5994 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5996 2018-02-19 Martin Liska <mliska@suse.cz>
5997 Richard Sandiford <richard.sandiford@linaro.org>
5999 PR tree-optimization/82491
6000 * gimple-fold.c (get_base_constructor): Make earlier bail out
6003 2018-02-19 Carl Love <cel@us.ibm.com>
6005 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6006 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6008 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6009 P8V_BUILTIN_VEC_NEG.
6011 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6013 * config/rl78/rl78.md (movdf): New define expand.
6015 2018-02-19 Martin Liska <mliska@suse.cz>
6018 * doc/invoke.texi: Fix typo.
6019 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6021 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6023 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6024 handle rs6000_single_float and rs6000_double_float specially for
6027 2018-02-16 Jeff Law <law@redhat.com>
6029 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6032 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6035 * ipa-inline.c (inline_small_functions): Fix a typo.
6037 2018-02-16 Nathan Sidwell <nathan@acm.org>
6039 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6041 2018-02-16 Carl Love <cel@us.ibm.com>
6043 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6044 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6045 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6046 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6047 expansion to P8V_BUILTIN_VEC_FLOAT2.
6049 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6051 PR rtl-optimization/70023
6052 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6053 src_regno into account.
6055 2018-02-16 Carl Love <cel@us.ibm.com>
6057 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6058 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6059 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6060 * config/rs6000/rs6000.c: Remove case statements for
6061 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6062 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6063 and P9V_BUILTIN_VEC_VINSERT4B.
6064 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6065 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6066 * config/rs6000/vsx.md:
6067 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6070 2018-02-16 Carl Love <cel@us.ibm.com>
6072 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6074 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6076 * config/rs6000/rs6000-c.c: Add the definitions for
6077 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6078 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6079 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6080 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6081 definition for insert4b and define insn *insert3b_internal.
6082 * doc/extend.texi: Add documentation for vec_extract4b.
6084 2018-02-16 Nathan Sidwell <nathan@acm.org>
6086 * doc/extend.texi (Backwards Compatibility): Mention friend
6087 injection. Note for-scope is deprecated.
6088 * doc/invoke.texi (-ffriend-injection): Deprecate.
6090 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6092 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6093 that moved to I2, also allow destinations that are a paradoxical
6094 subreg (instead of a normal reg).
6096 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6099 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6102 2018-02-16 Richard Biener <rguenther@suse.de>
6104 PR tree-optimization/84037
6105 PR tree-optimization/84016
6107 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6108 Adjust vec_construct for the fact we need additional higher latency
6109 128bit inserts for AVX256 and AVX512 vector builds.
6110 (ix86_add_stmt_cost): Scale vector construction cost for
6113 2018-02-16 Richard Biener <rguenther@suse.de>
6115 PR tree-optimization/84417
6116 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6117 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6118 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6120 2018-02-16 Martin Liska <mliska@suse.cz>
6123 * internal-fn.def (ASAN_CHECK): Set proper flags.
6124 (ASAN_MARK): Likewise.
6126 2018-02-16 Julia Koval <julia.koval@intel.com>
6128 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6129 from PTA_CANNONLAKE.
6131 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6134 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6135 Use ++iter rather than iter++ for std::list iterators.
6136 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6137 defer deleting them until all nodes in the forest are processed. Do
6138 free even leaf nodes. Change to_process into auto_vec.
6141 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6142 * vec.h (vec_default_construct): Use memset instead of placement new
6143 if BROKEN_VALUE_INITIALIZATION is defined.
6144 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6145 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6148 PR rtl-optimization/83723
6149 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6150 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6151 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6153 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6155 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6157 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6159 PR rtl-optimization/81443
6160 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6161 from inner REGs to paradoxical SUBREGs.
6163 2018-02-16 Richard Biener <rguenther@suse.de>
6165 PR tree-optimization/84399
6166 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6167 For operands we can analyze at their definition make sure we can
6168 analyze them at each use as well.
6170 2018-02-16 Richard Biener <rguenther@suse.de>
6172 PR tree-optimization/84190
6173 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6174 volatile accesses if the decl isn't volatile.
6176 2018-02-15 Jason Merrill <jason@redhat.com>
6178 PR c++/84314 - ICE with templates and fastcall attribute.
6179 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6180 TYPE_CANONICAL on an existing type.
6182 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6184 PR tree-optimization/84383
6185 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6186 dstoff nor call operand_equal_p if dstbase is NULL.
6188 PR tree-optimization/84334
6189 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6190 also a CONSTANT_CLASS_P, punt.
6192 2018-02-14 Jim Wilson <jimw@sifive.com>
6194 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6195 first SMALL_OPERAND check. New local min_second_step. Move assert
6196 to where locals are set. Add TARGET_RVC support.
6197 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6199 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6201 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6203 2018-02-14 Martin Sebor <msebor@redhat.com>
6205 PR tree-optimization/83698
6206 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6207 arrays constrain the offset range to their bounds.
6208 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6209 (builtin_access::overlap): Avoid setting the size of overlap if it's
6211 (maybe_diag_overlap): Also consider arrays when deciding what values
6212 of offsets to include in diagnostics.
6214 2018-02-14 Martin Sebor <msebor@redhat.com>
6217 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6218 that correspond to the kind of a declaration.
6220 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6223 * config/pa/pa.md: Load address of PIC label using the linkage table
6224 if the label is nonlocal.
6226 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6228 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6229 warning message if user requests -maltivec=be.
6230 * doc/invoke.texi: Document deprecation of -maltivec=be.
6232 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6235 * config/rs6000/rs6000-c.c: Update definitions for
6236 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6237 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6239 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6242 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6243 add _get_ssp intrinsics. Remove argument from
6244 __builtin_ia32_rdssp[d|q].
6245 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6246 * config/i386/i386-builtin.def: Remove argument from
6247 __builtin_ia32_rdssp[d|q].
6248 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6249 ix86_expand_special_args_builtin for _rdssp[d|q].
6250 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6251 Clear register before usage.
6252 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6253 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6255 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6257 PR tree-optimization/84357
6258 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6259 operand 1 of an ARRAY_REF too.
6261 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6264 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6265 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6267 (set_of_reg): New struct.
6268 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6269 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6270 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6272 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6273 Split into bitclr, bitset, bitinvert patterns if appropriate.
6274 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6275 use rx_fuse_in_memory_bitop.
6276 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6277 to named insn, correct maximum insn length.
6279 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6282 * machmode.def: Define a complex mode for PARTIAL_INT.
6283 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6285 * doc/rtl.texi: Document CSPImode.
6286 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6288 (msp430_hard_regno_nregs_with_padding): Likewise.
6290 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6293 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6295 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6297 PR rtl-optimization/84169
6298 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6299 we generated a parallel as new i3 and we split that to new i2 and i3
6300 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6301 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6302 those to i2, not i1. Partially rewrite this scan code.
6304 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6307 * stor-layout.c (place_field): For variable length fields, adjust
6308 offset_align afterwards not just based on the field's alignment,
6309 but also on the size.
6312 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6313 of exps and logs in the use_exp2 case.
6315 2018-02-13 Jeff Law <law@redhat.com>
6317 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6320 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6323 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6327 * common.opt (gas-loc-support, gas-locview-support): New.
6328 (ginline-points, ginternal-reset-location-views): New.
6329 * doc/invoke.texi: Document them. Use @itemx where intended.
6330 (gvariable-location-views): Adjust.
6331 * target.def (reset_location_view): New.
6332 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6333 (TARGET_RESET_LOCATION_VIEW): New.
6334 * doc/tm.texi: Rebuilt.
6335 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6336 (dwarf2out_default_as_locview_support): New.
6337 (output_asm_line_debug_info): Use option variables.
6338 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6339 (output_loc_list): Likewise.
6340 (add_high_low_attributes): Check option variables.
6341 Don't output entry view attribute in strict mode.
6342 (gen_inlined_subroutine_die): Check option variables.
6343 (dwarf2out_inline_entry): Likewise.
6344 (init_sections_and_labels): Likewise.
6345 (dwarf2out_early_finish): Likewise.
6346 (maybe_reset_location_view): New, from...
6347 (dwarf2out_var_location): ... here. Call it.
6348 * debug.h (dwarf2out_default_as_loc_support): Declare.
6349 (dwarf2out_default_as_locview_support): Declare.
6350 * hooks.c (hook_int_rtx_insn_0): New.
6351 * hooks.h (hook_int_rtx_insn_0): Declare.
6352 * toplev.c (process_options): Take -gas-loc-support and
6353 -gas-locview-support from dwarf2out. Enable
6354 -gvariable-location-views by default only with locview
6355 assembler support. Enable -ginternal-reset-location-views by
6356 default only if the target defines the corresponding hook.
6357 Enable -ginline-points by default if location views are
6358 enabled; force it disabled if statement frontiers are
6360 * tree-inline.c (expand_call_inline): Check option variables.
6361 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6363 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6365 PR tree-optimization/84321
6366 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6367 handling. Also check whether the anti-range contains any values
6368 that satisfy the mask; switch to a VR_RANGE if not.
6370 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6373 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6375 2018-02-13 Martin Jambor <mjambor@suse.cz>
6378 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6379 of call statements, also set location of a load to a temporary.
6381 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6383 * config/rl78/rl78.c (add_vector_labels): New function.
6384 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6385 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6386 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6387 which checks that no arguments are passed.
6388 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6389 * doc/extend.texi: Documentation for the new attribute.
6391 2018-02-13 Andreas Schwab <schwab@suse.de>
6393 * config/riscv/linux.h (CPP_SPEC): Define.
6395 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6398 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6399 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6400 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6401 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6402 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6403 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6404 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6405 temporarily for AES and PCLMUL builtins.
6407 PR tree-optimization/84339
6408 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6409 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6413 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6414 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6415 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6417 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6419 * omp-simd-clone.h: New file.
6420 * omp-simd-clone.c: Include omp-simd-clone.h.
6421 (expand_simd_clones): No longer static.
6422 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6423 cgraph.h and omp-simd-clone.h.
6424 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6425 (vect_recog_widen_shift_pattern): Formatting fix.
6426 (vect_pattern_recog_1): Don't check optab for calls.
6429 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6430 operands[2] into a REG before using gen_lowpart on it.
6432 2018-02-12 Jeff Law <law@redhat.com>
6435 * config/sh/sh.c (find_barrier): Consider a sibling call
6438 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6439 successfully back substituting a reg.
6441 2018-02-12 Richard Biener <rguenther@suse.de>
6443 PR tree-optimization/84037
6444 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6445 parameter, move visited init to caller.
6446 (vect_slp_analyze_operations): Separate cost from validity
6447 check, initialize visited once for all instances.
6448 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6450 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6452 (vect_model_promotion_demotion_cost): Likewise.
6453 (vectorizable_bswap): Guard cost modeling with !slp_node
6454 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6456 (vectorizable_call): Likewise.
6457 (vectorizable_conversion): Likewise.
6458 (vectorizable_assignment): Likewise.
6459 (vectorizable_shift): Likewise.
6460 (vectorizable_operation): Likewise.
6461 (vectorizable_store): Likewise.
6462 (vectorizable_load): Likewise.
6463 (vectorizable_condition): Likewise.
6464 (vectorizable_comparison): Likewise.
6466 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6469 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6470 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6471 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6473 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6476 * params.def (inline-min-speedup): Increase from 8 to 15.
6477 (max-inline-insns-auto): Decrease from 40 to 30.
6478 * ipa-split.c (consider_split): Add some buffer for function to
6479 be considered inlining candidate.
6480 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6483 2018-02-12 Richard Biener <rguenther@suse.de>
6485 PR tree-optimization/84037
6486 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6487 matched stmts if we cannot swap the non-matched ones.
6489 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6491 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6492 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6493 _mm_maskz_scalef_round_ss): New intrinsics.
6494 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6495 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6496 __builtin_ia32_scalefss_round): Remove.
6497 (__builtin_ia32_scalefsd_mask_round,
6498 __builtin_ia32_scalefss_mask_round): New intrinsics.
6499 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6500 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6501 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6502 "<round_constraint>")): Changed to ...
6503 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6504 "<round_scalar_constraint>")): ... this.
6505 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6506 %0, %1, %2<round_op3>}"): Changed to ...
6507 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6508 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6509 %2<round_scalar_mask_op3>}"): ... this.
6510 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6512 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6514 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6515 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6516 (_mm_maskz_sqrt_round_ss): New intrinsics.
6517 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6518 (__builtin_ia32_sqrtsd_mask_round)
6519 (__builtin_ia32_sqrtss_mask_round): New builtins.
6520 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6521 (__builtin_ia32_sqrtss_round): Remove.
6522 (__builtin_ia32_sqrtsd_mask_round)
6523 (__builtin_ia32_sqrtss_mask_round): New builtins.
6524 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6525 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6526 ((match_operand:VF_128 1 "vector_operand"
6527 "xBm,<round_constraint>")): Changed to ...
6528 ((match_operand:VF_128 1 "vector_operand"
6529 "xBm,<round_scalar_constraint>")): ... this.
6530 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6531 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6532 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6533 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6534 %<iptr>1<round_scalar_mask_op3>}): ... this.
6535 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6536 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6538 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6541 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6542 Cast vec_cmpeq result to correct type.
6543 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6544 Cast vec_cmpgt result to correct type.
6546 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6548 * final.c (final_scan_insn_1): Renamed from...
6549 (final_scan_insn): ... this. New wrapper, to recover
6550 seen from the outermost call in recursive ones.
6551 * config/sparc/sparc.c (output_return): Drop seen from call.
6552 (output_sibcall): Likewise.
6553 * config/visium/visium.c (output_branch): Likewise.
6555 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6557 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6560 2018-02-10 Alan Modra <amodra@gmail.com>
6563 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6564 Specify LR as an input.
6566 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6569 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6570 remove_member_access_dummy_vars): New functions.
6571 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6572 lower_omp_1, execute_lower_omp): Use them.
6574 PR rtl-optimization/84308
6575 * shrink-wrap.c (spread_components): Release todo vector.
6577 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6579 PR rtl-optimization/57193
6580 * ira-color.c (struct allocno_color_data): Add member
6581 conflict_allocno_hard_prefs.
6582 (update_conflict_allocno_hard_prefs): New.
6583 (bucket_allocno_compare_func): Add a preference based on
6584 conflict_allocno_hard_prefs.
6585 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6586 (color_allocnos): Remove a dead code. Initiate
6587 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6589 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6592 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6593 constraint from =wa to wa. Avoid a subreg on the output operand,
6594 instead use a pseudo and subreg it in a move.
6595 (p9_xxbrd_<mode>): Changed to ...
6596 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6597 (p9_xxbrd_v2df): New expander.
6598 (p9_xxbrw_<mode>): Changed to ...
6599 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6600 (p9_xxbrw_v4sf): New expander.
6602 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6604 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6606 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6609 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6610 multiply in 32-bit mode.
6611 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6612 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6615 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6617 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6618 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6619 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6620 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6622 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6625 * dwarf2out.c (is_trivial_indirect_ref): New function.
6626 (dwarf2out_late_global_decl): Do not generate a location
6627 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6628 and that are not defined in the current unit.
6630 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6632 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6633 instead of a libcall for UNORDERED.
6635 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6638 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6639 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6641 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6644 * config/s390/s390.c (s390_set_current_function): Invoke
6645 s390_indirect_branch_settings also if fndecl didn't change.
6647 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6649 * config/rs6000/rs6000.md (blockage): Set length to zero.
6651 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6653 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6655 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6658 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6659 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6663 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6664 PARALLEL incoming that failed vt_get_decl_and_offset check.
6667 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6668 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6670 (get_variable_section): For decls in named .bss* sections pass true as
6671 second argument to bss_initializer_p.
6673 2018-02-09 Marek Polacek <polacek@redhat.com>
6674 Jakub Jelinek <jakub@redhat.com>
6677 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6678 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6679 Sync some changes from cxx_fold_indirect_ref.
6681 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6683 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6685 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6686 (BLOCK_INLINE_ENTRY_LABEL): New.
6687 (dwarf2out_var_location): Disregard inline entry markers.
6688 (inline_entry_data): New struct.
6689 (inline_entry_data_hasher): New hashtable type.
6690 (inline_entry_data_hasher::hash): New.
6691 (inline_entry_data_hasher::equal): New.
6692 (inline_entry_data_table): New variable.
6693 (add_high_low_attributes): Add DW_AT_entry_pc and
6694 DW_AT_GNU_entry_view attributes if a pending entry is found
6695 in inline_entry_data_table. Add old entry_pc attribute only
6696 if debug nonbinding markers are disabled.
6697 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6698 markers are enabled.
6699 (block_within_block_p, dwarf2out_inline_entry): New.
6700 (dwarf2out_finish): Check that no entries remained in
6701 inline_entry_data_table.
6702 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6703 (final_scan_insn, notice_source_line): Likewise.
6704 (rest_of_clean_state): Skip inline entry markers.
6705 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6707 * gimple.c (gimple_build_debug_inline_entry): New.
6708 * gimple.h (enum gimple_debug_subcode): Add
6709 GIMPLE_DEBUG_INLINE_ENTRY.
6710 (gimple_build_debug_inline_entry): Declare.
6711 (gimple_debug_inline_entry_p): New.
6712 (gimple_debug_nonbind_marker_p): Adjust.
6713 * insn-notes.def (INLINE_ENTRY): New.
6714 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6715 inline entry marker notes.
6716 (print_insn): Likewise.
6717 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6718 (INSN_DEBUG_MARKER_KIND): Likewise.
6719 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6720 * tree-inline.c (expand_call_inline): Build and insert
6721 debug_inline_entry stmt.
6722 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6723 inline entry blocks early, if nonbind markers are enabled.
6724 (dump_scope_block): Dump fragment info.
6725 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6726 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6727 (gimple_build_debug_inline_entry): New.
6728 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6729 Enable/disable inline entry points too.
6730 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6731 (DEBUG_INSN): Describe inline entry markers.
6733 * common.opt (gvariable-location-views): New.
6734 (gvariable-location-views=incompat5): New.
6735 * config.in: Rebuilt.
6736 * configure: Rebuilt.
6737 * configure.ac: Test assembler for view support.
6738 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6739 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6740 * dwarf2out.c (var_loc_view): New typedef.
6741 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6742 (dwarf2out_locviews_in_attribute): New.
6743 (dwarf2out_locviews_in_loclist): New.
6744 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6745 (enum dw_line_info_opcode): Add LI_adv_address.
6746 (struct dw_line_info_table): Add view.
6747 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6748 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6749 (zero_view_p): New variable.
6750 (ZERO_VIEW_P): New macro.
6751 (output_asm_line_debug_info): New.
6752 (struct var_loc_node): Add view.
6753 (add_AT_view_list, AT_loc_list): New.
6754 (add_var_loc_to_decl): Add view param. Test it against last.
6755 (new_loc_list): Add view params. Record them.
6756 (AT_loc_list_ptr): Handle loc and view lists.
6757 (view_list_to_loc_list_val_node): New.
6758 (print_dw_val): Handle dw_val_class_view_list.
6759 (size_of_die): Likewise.
6760 (value_format): Likewise.
6761 (loc_list_has_views): New.
6762 (gen_llsym): Set vl_symbol too.
6763 (maybe_gen_llsym, skip_loc_list_entry): New.
6764 (dwarf2out_maybe_output_loclist_view_pair): New.
6765 (output_loc_list): Output view list or entries too.
6766 (output_view_list_offset): New.
6767 (output_die): Handle dw_val_class_view_list.
6768 (output_dwarf_version): New.
6769 (output_compilation_unit_header): Use it.
6770 (output_skeleton_debug_sections): Likewise.
6771 (output_rnglists, output_line_info): Likewise.
6772 (output_pubnames, output_aranges): Update version comments.
6773 (output_one_line_info_table): Output view numbers in asm comments.
6774 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6775 Call maybe_gen_llsym.
6776 (loc_list_from_tree_1): Adjust.
6777 (add_AT_location_description): Create view list attribute if
6778 needed, check it's absent otherwise.
6779 (convert_cfa_to_fb_loc_list): Adjust.
6780 (maybe_emit_file): Call output_asm_line_debug_info for test.
6781 (dwarf2out_var_location): Reset views as needed. Precompute
6782 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6783 attribute. Set view.
6784 (new_line_info_table): Reset next view.
6785 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6786 (dwarf2out_source_line): Likewise. Output view resets and labels to
6787 the assembler, or select appropriate line info opcodes.
6788 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6789 (optimize_string_length): Catch it. Adjust.
6790 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6791 dw_val_class_view_list, and remove it if no longer needed.
6792 (hash_loc_list): Hash view numbers.
6793 (loc_list_hasher::equal): Compare them.
6794 (optimize_location_lists): Check whether a view list symbol is
6795 needed, and whether the locview attribute is present, and
6796 whether they match. Remove the locview attribute if no longer
6798 (index_location_lists): Call skip_loc_list_entry for test.
6799 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6800 Use output_dwarf_version.
6801 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6802 (struct dw_val_node): Add val_view_list.
6803 * final.c (SEEN_NEXT_VIEW): New.
6804 (set_next_view_needed): New.
6805 (clear_next_view_needed): New.
6806 (maybe_output_next_view): New.
6807 (final_start_function): Rename to...
6808 (final_start_function_1): ... this. Take pointer to FIRST,
6809 add SEEN parameter. Emit param bindings in the initial view.
6810 (final_start_function): Reintroduce SEEN-less interface.
6811 (final): Rename to...
6812 (final_1): ... this. Take SEEN parameter. Output final pending
6813 next view at the end.
6814 (final): Reintroduce seen-less interface.
6815 (final_scan_insn): Output pending next view before switching
6816 sections or ending a block. Mark the next view as needed when
6817 outputting variable locations. Notify debug backend of section
6818 changes, and of location view changes.
6819 (rest_of_handle_final): Adjust.
6820 * toplev.c (process_options): Autodetect value for debug variable
6821 location views option. Warn on incompat5 without -gdwarf-5.
6822 * doc/invoke.texi (gvariable-location-views): New.
6823 (gvariable-location-views=incompat5): New.
6824 (gno-variable-location-views): New.
6826 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6828 PR tree-optimization/84136
6829 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6830 that the result of find_edge is non-NULL.
6832 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6835 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6836 storing integer register in SImode. Fix cost of 256 and 512
6837 byte aligned SSE register store.
6839 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6841 * config/i386/i386.c (ix86_multiplication_cost): Fix
6842 multiplication cost for TARGET_AVX512DQ.
6844 2018-02-08 Marek Polacek <polacek@redhat.com>
6846 PR tree-optimization/84238
6847 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6850 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6852 PR tree-optimization/84265
6853 * tree-vect-stmts.c (vectorizable_store): Don't treat
6854 VMAT_CONTIGUOUS accesses as grouped.
6855 (vectorizable_load): Likewise.
6857 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6859 PR tree-optimization/81635
6860 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6861 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6862 (test_round_for_mask): New functions.
6863 (wide_int_cc_tests): Call test_round_for_mask.
6864 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6865 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6866 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6867 range returned by get_range_info.
6869 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6872 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6873 * symtab.c: Include builtins.h
6874 (symtab_node::output_to_lto_symbol_table_p): Move here
6875 from lto-streamer-out.c:output_symbol_p.
6876 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6877 (output_symbol_p): Move all logic to symtab.c
6878 (produce_symtab): Update.
6880 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6882 * config/s390/s390-opts.h (enum indirect_branch): Define.
6883 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6884 (s390_indirect_branch_via_thunk)
6885 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6886 (enum s390_indirect_branch_type): Define.
6887 * config/s390/s390.c (struct s390_frame_layout, struct
6888 machine_function): Remove.
6889 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6890 (indirect_branch_table_label_no, indirect_branch_table_name):
6892 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6893 (enum s390_indirect_branch_option): Define.
6894 (s390_return_addr_from_memory): New function.
6895 (s390_handle_string_attribute): New function.
6896 (s390_attribute_table): Add new attribute handler.
6897 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6898 (s390_indirect_branch_via_thunk): New function.
6899 (s390_indirect_branch_via_inline_thunk): New function.
6900 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6901 sibling call optimization for non z10 compiles.
6902 (s390_emit_call): Force indirect branch target to be a single
6903 register. Add r1 clobber for non-z10 compiles.
6904 (s390_emit_epilogue): Emit return jump via return_use expander.
6905 (s390_reorg): Handle JUMP_INSNs as execute targets.
6906 (s390_option_override_internal): Perform validity checks for the
6907 new command line options.
6908 (s390_indirect_branch_attrvalue): New function.
6909 (s390_indirect_branch_settings): New function.
6910 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6911 (s390_output_indirect_thunk_function): New function.
6912 (s390_code_end): Implement target hook.
6913 (s390_case_values_threshold): Implement target hook.
6914 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6916 * config/s390/s390.h (struct s390_frame_layout)
6917 (struct machine_function): Move here from s390.c.
6918 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6919 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6920 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6921 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6922 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6923 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6924 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6925 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6926 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6927 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6928 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6929 (mnemonic attribute): Add values which aren't recognized
6931 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6932 pattern for branch conversion. Fix mnemonic attribute.
6933 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6934 indirect branch via thunk if requested.
6935 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6936 ("*indirect_jump"): Disable for branch conversion using out of
6938 ("indirect_jump_via_thunk<mode>_z10")
6939 ("indirect_jump_via_thunk<mode>")
6940 ("indirect_jump_via_inlinethunk<mode>_z10")
6941 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6942 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6943 ("casesi_jump_via_inlinethunk<mode>_z10")
6944 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6945 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6946 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6947 ("*indirect2_jump"): Disable for branch conversion.
6948 ("casesi_jump"): Turn into expander and expand patterns for branch
6950 ("return_use"): New expander.
6951 ("*return"): Emit return via thunk and rename it to ...
6952 ("*return<mode>"): ... this one.
6953 * config/s390/s390.opt: Add new options and and enum for the
6956 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6958 * lra-constraints.c (match_reload): Unconditionally use
6959 gen_lowpart_SUBREG, rather than selecting between that
6960 and equivalent gen_rtx_SUBREG code.
6962 2018-02-08 Richard Biener <rguenther@suse.de>
6964 PR tree-optimization/84233
6965 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6966 changed flag instead of boguously re-using phi_inserted.
6968 2018-02-08 Martin Jambor <mjambor@suse.cz>
6970 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6971 static local variables.
6973 2018-02-08 Richard Biener <rguenther@suse.de>
6975 PR tree-optimization/84278
6976 * tree-vect-stmts.c (vectorizable_store): When looking for
6977 smaller vector types to perform grouped strided loads/stores
6978 make sure the mode is supported by the target.
6979 (vectorizable_load): Likewise.
6981 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6983 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6984 Increase LDP/STP opportunities by adding adjacent callee-saves.
6986 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6988 PR rtl-optimization/84068
6989 PR rtl-optimization/83459
6990 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6992 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6994 PR tree-optimization/84224
6995 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6996 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6999 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7002 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7003 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7004 count, remove one USE.
7006 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7008 * doc/install.texi (Configuration): Document the
7009 --with-long-double-format={ibm,ieee} PowerPC configuration
7013 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7014 Convert from define_expand to be define_insn_and_split. Rework
7015 float/double/_Float128 conversions to QI/HI/SImode to work with
7016 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7017 conversions to QI/HImode types did a store and then a load to
7018 truncate the value. For conversions to VSX registers, don't split
7019 the insn, instead emit the code directly. Use the code iterator
7020 any_fix to combine signed and unsigned conversions.
7021 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7022 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7023 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7024 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7025 (fix_<mode>di2_hw): Likewise.
7026 (fixuns_<mode>di2_hw): Likewise.
7027 (fix_<mode>si2_hw): Likewise.
7028 (fixuns_<mode>si2_hw): Likewise.
7029 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7030 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7031 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7032 fix<uns>_trunc<SFDF:mode>si2_p8.
7033 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7035 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7036 (fix<uns>_<mode>_mem): Likewise.
7037 (fctiw<u>z_<mode>_mem): Likewise.
7038 (fix<uns>_<mode>_mem): Likewise.
7039 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7040 the register allocator from doing a direct move to the GPRs to do
7041 a store, and instead use the ISA 3.0 store byte/half-word from
7042 vector register instruction. For IEEE 128-bit floating point,
7043 also optimize stores of 32-bit ints.
7044 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7046 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7048 * genextract.c (push_pathstr_operand): New function to support
7050 (walk_rtx): Call push_pathstr_operand.
7051 (print_path): Support [a-zA-Z].
7053 2018-02-07 Richard Biener <rguenther@suse.de>
7055 PR tree-optimization/84037
7056 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7057 (cse_and_gimplify_to_preheader): Declare.
7058 (vect_get_place_in_interleaving_chain): Likewise.
7059 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7061 (_loop_vec_info::~_loop_vec_info): Delete it.
7062 (cse_and_gimplify_to_preheader): New function.
7063 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7064 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7065 (vectorizable_load): Likewise. For grouped stores always base
7066 the IV on the first element.
7067 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7068 condition before gimplifying.
7070 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7072 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7073 *DIV_EXPR and *MOD_EXPR.
7075 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7078 * config/i386/i386.c (ix86_option_override_internal): Mask out
7079 the CF_SET bit when checking -fcf-protection.
7081 2018-02-07 Tom de Vries <tom@codesourcery.com>
7084 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7087 2018-02-07 Richard Biener <rguenther@suse.de>
7089 PR tree-optimization/84204
7090 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7093 PR tree-optimization/84205
7094 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7095 special-case isl_ast_op_zdiv_r.
7097 PR tree-optimization/84223
7098 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7099 Only add conditions from within the region.
7100 (gather_bbs::after_dom_children): Adjust.
7102 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7105 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7106 * config/avr/avr.md: Only post-reload split REG-REG moves if
7107 either register is GENERAL_REG_P.
7109 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7111 PR tree-optimization/84235
7112 * tree-ssa-scopedtables.c
7113 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7114 if the subtraction is performed in floating point type where NaNs are
7115 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7116 build 1. Formatting fix.
7118 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7121 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7122 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7123 and skip it regardless of bb boundaries. Use CALL_P macro,
7124 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7125 unnecessarily, formatting fix.
7127 2018-02-06 Michael Collison <michael.collison@arm.com>
7129 * config/arm/thumb2.md:
7130 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7131 (*thumb_mov_notscc): Ditto.
7133 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7136 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7137 unsigned_fix, not "s".
7139 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7141 * configure.ac (gcc_fn_eh_frame_ro): New function.
7142 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7143 correct .eh_frame permissions.
7144 * configure: Regenerate.
7146 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7148 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7151 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7153 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7154 Display warning message for -mno-speculate-indirect-jumps.
7156 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7158 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7160 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7162 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7164 PR tree-optimization/84225
7165 * tree-eh.c (find_trapping_overflow): Only call
7166 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7168 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7171 * config/i386/i386.c: Reimplement the check of possible options
7172 -mibt/-mshstk conbination. Change error messages.
7173 * doc/invoke.texi: Fix a typo: remove extra '='.
7175 2018-02-06 Marek Polacek <polacek@redhat.com>
7177 PR tree-optimization/84228
7178 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7180 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7183 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7184 emitted arch directives.
7185 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7186 __ARM_FEATURE_COPROC before changing architectures.
7188 2018-02-06 Richard Biener <rguenther@suse.de>
7190 * config/i386/i386.c (print_reg): Fix typo.
7191 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7193 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7195 * configure: Regenerate.
7197 2018-02-05 Martin Sebor <msebor@redhat.com>
7199 PR tree-optimization/83369
7200 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7203 2018-02-05 Martin Liska <mliska@suse.cz>
7205 * doc/invoke.texi: Cherry-pick upstream r323995.
7207 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7209 * ira.c (ira_init_register_move_cost): Adjust comment.
7211 2018-02-05 Martin Liska <mliska@suse.cz>
7213 PR gcov-profile/84137
7214 * doc/gcov.texi: Fix typo in documentation.
7216 2018-02-05 Martin Liska <mliska@suse.cz>
7218 PR gcov-profile/83879
7219 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7220 using dlopen functionality.
7222 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7224 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7225 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7226 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7227 _mm_maskz_range_round_ss): New intrinsics.
7228 (__builtin_ia32_rangesd128_round)
7229 (__builtin_ia32_rangess128_round): Remove.
7230 (__builtin_ia32_rangesd128_mask_round,
7231 __builtin_ia32_rangess128_mask_round): New builtins.
7232 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7233 __builtin_ia32_rangess128_round): Remove.
7234 (__builtin_ia32_rangesd128_mask_round,
7235 __builtin_ia32_rangess128_mask_round): New builtins.
7236 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7237 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7238 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7239 "<round_saeonly_constraint>")): Changed to ...
7240 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7241 "<round_saeonly_scalar_constraint>")): ... this.
7242 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7243 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7244 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7245 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7246 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7248 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7250 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7252 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7253 Remove all values except native, 8540 and 8548.
7255 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7257 * config/i386/i386.c (ix86_output_function_return): Pass
7258 INVALID_REGNUM, instead of -1, as invalid register number to
7259 indirect_thunk_name and output_indirect_thunk.
7261 2018-02-02 Julia Koval <julia.koval@intel.com>
7263 * config.gcc: Add -march=icelake.
7264 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7265 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7266 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7267 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7268 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7269 (processor_target_table): Add icelake.
7270 (ix86_option_override_internal): Handle new PTAs.
7271 (get_builtin_code_for_version): Handle icelake.
7272 (M_INTEL_COREI7_ICELAKE): New.
7273 (fold_builtin_cpu): Handle icelake.
7274 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7275 * doc/invoke.texi: Add -march=icelake.
7277 2018-02-02 Julia Koval <julia.koval@intel.com>
7279 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7280 to wide_int_bitmask.
7281 * wide-int-bitmask.h: New.
7283 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7286 * config/i386/i386.md: Replace Pmode with word_mode in
7287 builtin_setjmp_setup and builtin_longjmp to support x32.
7289 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7293 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7295 (rs6000_supported_cpu_names): New static variable.
7296 (linux_cpu_translation_table): Likewise.
7297 (elf_platform) <cpu>: Define new static variable and use it.
7298 Translate kernel AT_PLATFORM name to canonical name if needed.
7299 Error if platform name is unknown.
7301 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7304 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7306 2018-02-01 Jeff Law <law@redhat.com>
7309 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7310 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7311 the scratch if RELEASE_VIA_POP is false.
7312 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7313 If we have to save a temporary register, decrement SIZE appropriately.
7314 Pass new arguments to release_scratch_register_on_entry.
7315 (ix86_adjust_stack_and_probe): Likewise.
7316 (ix86_emit_probe_stack_range): Pass new arguments to
7317 release_scratch_register_on_entry.
7319 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7321 PR rtl-optimization/84157
7322 * combine.c (change_zero_ext): Use REG_P predicate in
7323 front of HARD_REGISTER_P predicate.
7325 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7327 * config/avr/avr.c (avr_option_override): Move disabling of
7328 -fdelete-null-pointer-checks to...
7329 * common/config/avr/avr-common.c (avr_option_optimization_table):
7332 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7334 PR tree-optimization/81635
7335 * tree-data-ref.c (split_constant_offset_1): For types that
7336 wrap on overflow, try to use range info to prove that wrapping
7339 2018-02-01 Renlin Li <renlin.li@arm.com>
7342 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7344 (aarch64_register_move_cost): Likewise.
7345 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7347 (REG_CLASS_NAMES): Likewise.
7348 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7349 TAILCALL_ADDR_REGS. Remove IP registers.
7350 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7352 2018-02-01 Richard Biener <rguenther@suse.de>
7354 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7355 for specifying RPO order and allow NULL for that.
7356 * domwalk.c (dom_walker::dom_walker): Likewise.
7357 (dom_walker::walk): Handle NULL RPO order.
7358 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7360 (rewrite_update_dom_walker): Likewise.
7361 (mark_def_dom_walker): Likewise.
7363 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7365 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7366 (aarch64_maybe_expand_sve_subreg_move): Declare.
7367 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7368 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7370 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7371 that are semantically a reverse operation.
7372 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7373 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7374 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7376 (aarch64_can_change_mode_class): For big-endian, forbid changes
7377 between two SVE modes if they have different element sizes.
7379 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7381 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7382 the TImode handling for big-endian targets.
7384 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7386 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7387 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7389 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7390 Remove BSWAP handing for big-endian targets and use the form of
7391 LD1RQ appropariate for the mode.
7393 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7395 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7396 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7399 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7402 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7403 check for operands that need to go through aarch64_sve_reload_be.
7405 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7407 PR tree-optimization/81661
7408 PR tree-optimization/84117
7409 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7410 * tree-eh.c: Include gimplify.h.
7411 (find_trapping_overflow, replace_trapping_overflow,
7412 rewrite_to_non_trapping_overflow): New functions.
7413 * tree-vect-loop.c: Include tree-eh.h.
7414 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7415 * tree-data-ref.c: Include tree-eh.h.
7416 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7418 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7420 PR rtl-optimization/84123
7421 * combine.c (change_zero_ext): Check if hard register satisfies
7422 can_change_dest_mode before calling gen_lowpart_SUBREG.
7424 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7427 * ira.c (ira_init_register_move_cost): Remove assert.
7429 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7431 PR rtl-optimization/84071
7432 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7433 * doc/tm.texi: Regenerate.
7435 2018-01-31 Richard Biener <rguenther@suse.de>
7437 PR tree-optimization/84132
7438 * tree-data-ref.c (analyze_miv_subscript): Properly
7439 check whether evolution_function_is_affine_multivariate_p
7440 before calling gcd_of_steps_may_divide_p.
7442 2018-01-31 Julia Koval <julia.koval@intel.com>
7445 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7446 * config/i386/i386.md (rdpid_rex64) New.
7447 (rdpid): Make 32bit only.
7449 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7452 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7453 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7455 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7458 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7460 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7462 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7464 PR rtl-optimization/84071
7465 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7466 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7468 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7470 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7471 (arc_attribute_table): Add 'aux' attribute.
7472 (arc_in_small_data_p): Consider aux like variables.
7473 (arc_is_aux_reg_p): New function.
7474 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7475 (arc_get_aux_arg): New function.
7476 (prepare_move_operands): Handle aux-register access.
7477 (arc_handle_aux_attribute): New function.
7478 * doc/extend.texi (ARC Variable attributes): Add subsection.
7480 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7482 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7483 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7484 (arc_attribute_table): Add 'uncached' attribute.
7485 (arc_print_operand): Print '.di' flag for uncached memory
7487 (arc_in_small_data_p): Do not consider for small data the uncached
7489 (arc_is_uncached_mem_p): New function.
7490 * config/arc/predicates.md (compact_store_memory_operand): Check
7491 for uncached memory accesses.
7492 (nonvol_nonimm_operand): Likewise.
7493 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7495 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7498 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7499 falign-loops=): Add Optimization flag.
7501 2018-01-30 Jeff Law <law@redhat.com>
7504 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7505 INT_REGISTERS_SAVED. Check it prior to calling
7506 get_scratch_register_on_entry.
7507 (ix86_adjust_stack_and_probe): Similarly.
7508 (ix86_emit_probe_stack_range): Similarly.
7509 (ix86_expand_prologue): Corresponding changes.
7511 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7514 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7515 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7517 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7520 * lra-constraints.c (curr_insn_transform): Process AND in the
7523 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7525 PR rtl-optimization/83986
7526 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7527 dependence against last_pending_memory_flush in addition to
7530 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7532 PR tree-optimization/81611
7533 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7536 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7539 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7542 2018-01-30 Richard Biener <rguenther@suse.de>
7543 Jakub Jelinek <jakub@redhat.com>
7545 PR tree-optimization/84111
7546 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7547 inner loops added during recursion, as they don't have up-to-date
7550 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7553 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7554 (can_inline_edge_by_limits_p): ... here.
7555 (can_early_inline_edge_p, check_callers,
7556 update_caller_keys, update_callee_keys, recursive_inlining,
7557 add_new_edges_to_heap, speculation_useful_p,
7558 inline_small_functions,
7559 inline_small_functions, flatten_function,
7560 inline_to_all_callers_1): Update.
7562 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7564 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7567 2018-01-30 Richard Biener <rguenther@suse.de>
7569 PR tree-optimization/83008
7570 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7571 invariant and constant vector uses in stmts when they need
7574 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7577 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7578 * configure: Regenerate.
7580 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7582 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7584 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7585 Use gen_rtx_REG rather than gen_lowpart.
7587 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7589 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7590 rather than 0 when creating partial subregs.
7592 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7594 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7597 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7600 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7601 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7602 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7603 flags. This restores the settings used before the 2017-07-24.
7604 Turning off pre increment/decrement/modify allows IVOPTS to
7605 optimize DF/SF loops where the index is an int.
7607 2018-01-29 Richard Biener <rguenther@suse.de>
7608 Kelvin Nilsen <kelvin@gcc.gnu.org>
7611 * tree-vect-stmts.c (vectorizable_call): Don't call
7612 targetm.vectorize_builtin_md_vectorized_function if callee is
7615 2018-01-22 Carl Love <cel@us.ibm.com>
7617 * doc/extend.tex: Fix typo in second arg in
7618 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7620 2018-01-29 Richard Biener <rguenther@suse.de>
7622 PR tree-optimization/84086
7623 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7624 (flush_ssaname_freelist): When SSA names were released reset
7625 the SCEV hash table.
7627 2018-01-29 Richard Biener <rguenther@suse.de>
7629 PR tree-optimization/84057
7630 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7631 removed paths when removing edges.
7633 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7635 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7636 -mfunction-return=@var{choice}.
7638 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7641 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7643 (layout::print_source_line): Likewise.
7644 (test_get_line_width_without_trailing_whitespace): Add test cases.
7646 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7649 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7652 2018-01-26 Jim Wilson <jimw@sifive.com>
7654 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7656 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7659 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7661 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7662 and CMP + SUB-immediate -> SUBS.
7664 2018-01-26 Martin Sebor <msebor@redhat.com>
7666 PR tree-optimization/83896
7667 * tree-ssa-strlen.c (get_string_len): Rename...
7668 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7669 Avoid assuming length is constant.
7670 (handle_char_store): Use HOST_WIDE_INT for string length.
7672 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7675 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7676 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7678 2018-01-26 Richard Biener <rguenther@suse.de>
7680 PR rtl-optimization/84003
7681 * dse.c (record_store): Only record redundant stores when
7682 the earlier store aliases at least all accesses the later one does.
7684 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7686 PR rtl-optimization/83985
7687 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7688 REG_CFA_RESTORE insns.
7689 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7690 REG_CFA_RESTORE insns here.
7693 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7694 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7696 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7698 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7699 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7700 (arc_init): Likewise.
7701 (arc_override_options): Likewise.
7702 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7704 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7706 * config/arc/arc.h (TARGET_DBNZ): Define.
7707 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7708 properly set the tune attribute.
7709 (dbnz): Use TARGET_DBNZ guard.
7710 * config/arc/arc.opt (mtune): Add core3 option.
7712 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7714 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7715 recognize new pic like addresses.
7716 (arc_delegitimize_address): Clean up.
7718 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7720 * config/arc/arc-arches.def: Option mrf16 valid for all
7722 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7723 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7724 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7725 * config/arc/arc-tables.opt: Regenerate.
7726 * config/arc/arc.c (arc_conditional_register_usage): Handle
7727 reduced register file case.
7728 (arc_file_start): Set must have build attributes.
7729 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7731 * config/arc/arc.opt (mrf16): Add new option.
7732 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7733 * config/arc/genmultilib.awk: Handle new mrf16 option.
7734 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7735 * config/arc/t-multilib: Regenerate.
7736 * doc/invoke.texi (ARC Options): Document mrf16 option.
7738 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7740 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7741 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7742 (arc_attribute_table): Add 'secure_call' attribute.
7743 (arc_print_operand): Print secure call operand.
7744 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7746 (arc_is_secure_call_p): New function. * config/arc/arc.md
7747 (call_i): Add support for sjli instruction.
7748 (call_value_i): Likewise.
7749 * config/arc/constraints.md (Csc): New constraint.
7751 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7752 John Eric Martin <John.Martin@emmicro-us.com>
7754 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7755 * config/arc/arc.c (_arc_jli_section): New struct.
7756 (arc_jli_section): New type.
7757 (rc_jli_sections): New static variable.
7758 (arc_handle_jli_attribute): New function.
7759 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7760 (arc_file_end): New function.
7761 (TARGET_ASM_FILE_END): Define.
7762 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7763 (arc_add_jli_section): New function.
7764 (jli_call_scan): Likewise.
7765 (arc_reorg): Call jli_call_scan.
7766 (arc_output_addsi): Remove 'S' from printing asm operand.
7767 (arc_is_jli_call_p): New function.
7768 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7770 (movhi_insn): Likewise.
7771 (movsi_insn): Likewise.
7772 (movsi_set_cc_insn): Likewise.
7773 (loadqi_update): Likewise.
7774 (load_zeroextendqisi_update): Likewise.
7775 (load_signextendqisi_update): Likewise.
7776 (loadhi_update): Likewise.
7777 (load_zeroextendhisi_update): Likewise.
7778 (load_signextendhisi_update): Likewise.
7779 (loadsi_update): Likewise.
7780 (loadsf_update): Likewise.
7781 (movsicc_insn): Likewise.
7782 (bset_insn): Likewise.
7783 (bxor_insn): Likewise.
7784 (bclr_insn): Likewise.
7785 (bmsk_insn): Likewise.
7786 (bicsi3_insn): Likewise.
7787 (cmpsi_cc_c_insn): Likewise.
7788 (movsi_ne): Likewise.
7789 (movsi_cond_exec): Likewise.
7790 (clrsbsi2): Likewise.
7801 (call_i): Remove 'S' asm letter, add jli instruction.
7802 (call_value_i): Likewise.
7803 * config/arc/arc.op (mjli-always): New option.
7804 * config/arc/constraints.md (Cji): New constraint.
7805 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7807 (subsf3_fpx): Likewise.
7808 (mulsf3_fpx): Likewise.
7809 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7811 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7813 * doc/invoke.texi (ARC): Document mjli-always option.
7815 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7817 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7818 avoid addition with 0 and use incw and decw where possible.
7820 2018-01-26 Richard Biener <rguenther@suse.de>
7822 PR tree-optimization/81082
7823 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7824 association if it requires casting to unsigned.
7825 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7826 from fold_plusminus_mult_expr to catch important cases late when
7827 range info is available.
7829 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7831 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7832 * configure.ac (hidden_linkonce): New test.
7833 * configure: Regenerate.
7834 * config.in: Regenerate.
7836 2018-01-26 Julia Koval <julia.koval@intel.com>
7838 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7839 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7840 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7841 _mm_mask_bitshuffle_epi64_mask): Fix type.
7842 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7843 USI_FTYPE_V4DI_V4DI_USI): Remove.
7844 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7845 __builtin_ia32_vpshufbitqmb256_mask,
7846 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7847 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7848 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7850 2018-01-26 Alan Modra <amodra@gmail.com>
7853 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7854 UNSPEC_VBPERMQ. Sort other unspecs.
7856 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7858 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7860 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7863 * predict.c (drop_profile): Do not push/pop cfun; update also
7865 (handle_missing_profiles): Fix logic looking for zero profiles.
7867 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7870 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7871 on functions with #pragma omp declare simd or functions with simd
7873 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7874 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7875 Remove trailing \n from warning_at calls.
7877 2018-01-25 Tom de Vries <tom@codesourcery.com>
7880 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7881 for neutered workers.
7883 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7886 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7887 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7889 2018-01-24 Jeff Law <law@redhat.com>
7892 * i386.c (get_probe_interval): Move to earlier point.
7893 (ix86_compute_frame_layout): If -fstack-clash-protection and
7894 the frame is larger than the probe interval, then use pushes
7895 to save registers rather than reg->mem moves.
7896 (ix86_expand_prologue): Remove conditional for int_registers_saved
7899 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7902 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7903 min/max for never referenced object.
7905 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7908 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7910 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7911 attributes from DECL_ATTRIBUTES (decl) without affecting
7912 DECL_ATTRIBUTES (current_function_decl).
7913 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7914 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7916 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7918 PR tree-optimization/83979
7919 * fold-const.c (fold_comparison): Use constant_boolean_node
7920 instead of boolean_{true,false}_node.
7922 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7924 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7927 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7929 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7930 Simplify the clause that sets the length attribute.
7931 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7932 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7933 clause that sets the length attribute.
7934 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7936 2018-01-24 Tom de Vries <tom@codesourcery.com>
7939 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7940 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7941 Add strict parameter.
7942 (prevent_branch_around_nothing): Insert dummy insn between branch to
7943 label and label with no ptx insn inbetween.
7944 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7946 2018-01-24 Tom de Vries <tom@codesourcery.com>
7949 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7950 for neutered threads in warp.
7951 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7953 2018-01-24 Richard Biener <rguenther@suse.de>
7955 PR tree-optimization/83176
7956 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7959 2018-01-24 Richard Biener <rguenther@suse.de>
7961 PR tree-optimization/82819
7962 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7963 code generating pluses that are no-ops in the target precision.
7965 2018-01-24 Richard Biener <rguenther@suse.de>
7968 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7970 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7972 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7973 to merge probabilities.
7974 * predict.c (probably_never_executed): Also mark as cold functions
7975 with global 0 profile and guessed local profile.
7976 * profile-count.c (profile_probability::combine_with_count): New
7978 * profile-count.h (profile_probability::operator*,
7979 profile_probability::operator*=, profile_probability::operator/,
7980 profile_probability::operator/=): Reduce precision to adjusted
7981 and set value to guessed on contradictory divisions.
7982 (profile_probability::combine_with_freq): Remove.
7983 (profile_probability::combine_wiht_count): Declare.
7984 (profile_count::force_nonzero):: Set to adjusted.
7985 (profile_count::probability_in):: Set quality to adjusted.
7986 * tree-ssa-tail-merge.c (replace_block_by): Use
7989 2018-01-23 Andrew Waterman <andrew@sifive.com>
7990 Jim Wilson <jimw@sifive.com>
7992 * config/riscv/riscv.c (riscv_stack_boundary): New.
7993 (riscv_option_override): Set riscv_stack_boundary. Handle
7994 riscv_preferred_stack_boundary_arg.
7995 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7996 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7997 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7998 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7999 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8000 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8002 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8005 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8006 of struct ix86_frame.
8007 (ix86_expand_epilogue): Likewise. Add a local variable for
8008 the reg_save_offset field in struct ix86_frame.
8010 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8012 PR tree-optimization/82604
8013 * tree-loop-distribution.c (enum partition_kind): New enum item
8014 PKIND_PARTIAL_MEMSET.
8015 (partition_builtin_p): Support above new enum item.
8016 (generate_code_for_partition): Ditto.
8017 (compute_access_range): Differentiate cases that equality can be
8018 proven at all loops, the innermost loops or no loops.
8019 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8020 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8021 (finalize_partitions, distribute_loop): Don't fuse partition of
8022 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8023 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8026 2018-01-23 Martin Liska <mliska@suse.cz>
8028 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8029 order to ignore the predictor.
8030 (PRED_POLYMORPHIC_CALL): Likewise.
8031 (PRED_RECURSIVE_CALL): Likewise.
8033 2018-01-23 Martin Liska <mliska@suse.cz>
8035 * tree-profile.c (tree_profiling): Print function header to
8036 aware reader which function we are working on.
8037 * value-prof.c (gimple_find_values_to_profile): Do not print
8038 not interesting value histograms.
8040 2018-01-23 Martin Liska <mliska@suse.cz>
8042 * profile-count.h (enum profile_quality): Add
8043 profile_uninitialized as the first value. Do not number values
8044 as they are zero based.
8045 (profile_count::verify): Update sanity check.
8046 (profile_probability::verify): Likewise.
8048 2018-01-23 Nathan Sidwell <nathan@acm.org>
8050 * doc/invoke.texi (ffor-scope): Deprecate.
8052 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8054 PR tree-optimization/83510
8055 * domwalk.c (set_all_edges_as_executable): New function.
8056 (dom_walker::dom_walker): Convert bool param
8057 "skip_unreachable_blocks" to enum reachability. Move setup of
8058 edge flags to set_all_edges_as_executable and only do it when
8059 reachability is REACHABLE_BLOCKS.
8060 * domwalk.h (enum dom_walker::reachability): New enum.
8061 (dom_walker::dom_walker): Convert bool param
8062 "skip_unreachable_blocks" to enum reachability.
8063 (set_all_edges_as_executable): New decl.
8064 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8065 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8067 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8068 but converting true to REACHABLE_BLOCKS.
8069 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8071 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8072 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8073 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8075 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8076 if check_all_array_refs will be called.
8078 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8080 * tree.c (selftest::test_location_wrappers): Add more test
8083 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8085 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8086 (selftest::test_bit_in_range): Likewise.
8088 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8091 * doc/sourcebuild.texi (vect_float): Say that the selector
8092 only describes the situation when -funsafe-math-optimizations is on.
8093 (vect_float_strict): Document.
8095 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8097 PR tree-optimization/83965
8098 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8099 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8100 instead of checking only for a reduction.
8101 (vect_recog_widen_sum_pattern): Likewise.
8103 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8105 * predict.c (probably_never_executed): Only use precise profile info.
8106 (compute_function_frequency): Skip after inlining hack since we now
8107 have quality checking.
8109 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8111 * profile-count.h (profile_probability::very_unlikely,
8112 profile_probability::unlikely, profile_probability::even): Set
8113 precision to guessed.
8115 2018-01-23 Richard Biener <rguenther@suse.de>
8117 PR tree-optimization/83963
8118 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8119 Properly terminate dominator walk when crossing the exit edge not
8120 when visiting its source block.
8122 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8125 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8126 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8128 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8130 PR tree-optimization/83957
8131 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8132 semicolon after for body surrounded by braces.
8134 PR tree-optimization/83081
8135 * profile-count.h (profile_probability::split): New method.
8136 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8137 Use profile_probability::split.
8138 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8139 when splitting a single conditional jump into 2.
8141 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8143 PR tree-optimization/69452
8144 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8147 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8149 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8150 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8151 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8153 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8155 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8156 * config/rl78/rl78.md (movdi): New define_expand.
8157 * config/rl78/rl78.c (rl78_split_movdi): New function.
8159 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8162 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8164 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8165 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8166 128-bit to produce an UNSPEC move to get the double word with the
8167 signbit and then a shift directly to do signbit.
8168 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8169 implementation with a new version that just does either a direct
8170 move or a regular move. Move memory interface to separate insns.
8171 Move insns so they are next to the expander.
8172 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8173 with signbit move. Split big and little endian case.
8174 (signbit<mode>2_dm_mem_le): Likewise.
8175 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8176 (signbit<mode>2_dm2): Likewise.
8178 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8180 * config/rl78/rl78.md (anddi3): New define_expand.
8182 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8184 * config/rl78/rl78.md (umindi3): New define_expand.
8186 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8188 * config/rl78/rl78.md (smindi3): New define_expand.
8190 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8192 * config/rl78/rl78.md (smaxdi3): New define_expand.
8194 2018-01-22 Carl Love <cel@us.ibm.com>
8196 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8197 LVX_V1TI): Add macro expansion.
8198 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8199 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8200 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8201 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8202 Change check to determine if the instruction is a byte reversing
8203 entry. Fix typo in comment.
8204 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8205 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8206 Add def_builtin calls for new builtins.
8207 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8208 Add define_insn expansion.
8210 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8212 * config/rl78/rl78.md (umaxdi3): New define_expand.
8214 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8216 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8217 for non-QImode registers.
8219 2018-01-22 Richard Biener <rguenther@suse.de>
8221 PR tree-optimization/83963
8222 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8223 including the loop exit block.
8224 (scop_detection::merge_sese): Likewise.
8225 (scop_detection::add_scop): Do it here instead.
8227 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8229 * doc/sourcebuild.texi (arm_softfloat): Document.
8231 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8234 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8235 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8236 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8238 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8239 David Edelsohn <dje.gcc@gmail.com>
8242 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8243 Change "crset eq" to "crset 2".
8244 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8245 (*call_indirect_aix<mode>_nospec): Likewise.
8246 (*call_value_indirect_aix<mode>_nospec): Likewise.
8247 (*call_indirect_elfv2<mode>_nospec): Likewise.
8248 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8249 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8250 change assembly output from . to $.
8251 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8252 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8253 (*tablejump<mode>_internal1_nospec): Likewise.
8255 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8258 * config/sh/sh_optimize_sett_clrt.cc:
8259 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8261 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8263 PR tree-optimization/83940
8264 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8265 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8266 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8267 use it to pass back the definition type.
8268 (vect_check_store_rhs): Likewise rhs_dt_out.
8269 (vect_build_gather_load_calls): Add a mask_dt argument and use
8270 it instead of a call to vect_is_simple_use.
8271 (vectorizable_store): Update calls to vect_check_load_store_mask
8272 and vect_check_store_rhs. Use the dt returned by the latter instead
8273 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8274 instead of calls to vect_is_simple_use. Pass the scalar rather
8275 than the vector operand to vect_is_simple_use when handling
8276 second and subsequent copies of an rhs value.
8277 (vectorizable_load): Update calls to vect_check_load_store_mask
8278 and vect_build_gather_load_calls. Use the cached mask_dt and
8279 gs_info.offset_dt instead of calls to vect_is_simple_use.
8281 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8284 * tree-emutls.c: Include gimplify.h.
8285 (lower_emutls_2): New function.
8286 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8287 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8288 it before further processing.
8291 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8292 UINTVAL (trueop1) instead of INTVAL (op1).
8294 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8298 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8299 INCOMING_FRAME_SP_OFFSET if not defined.
8300 (scan_trace): Add ENTRY argument. If true and
8301 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8302 emit a note to adjust the CFA offset.
8303 (create_cfi_notes): Adjust scan_trace callers.
8304 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8305 INCOMING_FRAME_SP_OFFSET in the CIE.
8306 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8307 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8309 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8310 * doc/tm.texi: Regenerated.
8312 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8314 PR rtl-optimization/83147
8315 * lra-constraints.c (remove_inheritance_pseudos): Use
8316 lra_substitute_pseudo_within_insn.
8318 2018-01-19 Tom de Vries <tom@codesourcery.com>
8319 Cesar Philippidis <cesar@codesourcery.com>
8322 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8324 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8327 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8328 spaces for function labels.
8330 2018-01-19 Martin Liska <mliska@suse.cz>
8332 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8333 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8334 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8335 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8336 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8337 (PRED_CONST_RETURN): Change from 69 to 65.
8338 (PRED_NULL_RETURN): Change from 91 to 71.
8339 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8340 (PRED_LOOP_GUARD): Change from 66 to 73.
8342 2018-01-19 Martin Liska <mliska@suse.cz>
8344 * predict.c (predict_insn_def): Add new assert.
8345 (struct branch_predictor): Change type to signed integer.
8346 (test_prediction_value_range): Amend test to cover
8348 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8349 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8350 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8351 (PRED_LOOP_IV_COMPARE): Likewise.
8352 * predict.h (PROB_UNINITIALIZED): Define new constant.
8354 2018-01-19 Martin Liska <mliska@suse.cz>
8356 * predict.c (dump_prediction): Add new format for
8357 analyze_brprob.py script which is enabled with -details
8359 * profile-count.h (precise_p): New function.
8361 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8363 PR tree-optimization/83922
8364 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8365 there are no statements that need masking.
8366 (vect_active_double_reduction_p): New function.
8367 (vect_analyze_loop_operations): Use it when handling phis that
8368 are not in the loop header.
8370 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8372 PR tree-optimization/83914
8373 * tree-vect-loop.c (vectorizable_induction): Don't convert
8374 init_expr or apply the peeling adjustment for inductions
8375 that are nested within the vectorized loop.
8377 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8379 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8382 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8386 * function.h (gimplify_parameters): Add gimple_seq * argument.
8387 * function.c: Include gimple.h and options.h.
8388 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8389 for the added local temporaries if needed.
8390 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8391 if there are any parameter cleanups, wrap whole body into a
8392 try/finally with the cleanups.
8394 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8397 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8398 Use GET_MODE_CLASS for scalar floating point.
8400 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8404 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8405 Fix call of call_cgraph_insertion_hooks.
8407 2018-01-18 Martin Sebor <msebor@redhat.com>
8409 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8411 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8414 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8417 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8420 * common.opt: (-ffile-prefix-map): New option.
8421 * opts.c (common_handle_option): Defer it.
8422 * opts-global.c (handle_common_deferred_options): Handle it.
8423 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8424 * file-prefix-map.h: New file.
8425 (remap_debug_filename, add_debug_prefix_map): ...here.
8426 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8427 * final.c (debug_prefix_map, add_debug_prefix_map
8428 remap_debug_filename): Move to...
8429 * file-prefix-map.c: New file.
8430 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8431 generalize, get rid of alloca(), use strrchr() instead of strchr().
8432 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8433 Implement in terms of add_prefix_map().
8434 (remap_macro_filename, remap_debug_filename): Implement in term of
8436 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8437 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8438 * dbxout.c: Include file-prefix-map.h.
8439 * varasm.c: Likewise.
8440 * vmsdbgout.c: Likewise.
8441 * xcoffout.c: Likewise.
8442 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8443 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8444 * doc/invoke.texi (-ffile-prefix-map): Document.
8445 (-fdebug-prefix-map): Update description.
8447 2018-01-18 Martin Liska <mliska@suse.cz>
8449 * config/i386/i386.c (indirect_thunk_name): Document that also
8451 (output_indirect_thunk): Document why both instructions
8452 (pause and lfence) are generated.
8454 2018-01-18 Richard Biener <rguenther@suse.de>
8456 PR tree-optimization/83887
8457 * graphite-scop-detection.c
8458 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8459 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8460 (scop_detection::merge_sese): Re-implement with a flood-fill
8461 algorithm that properly finds a SESE region if it exists.
8463 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8466 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8467 pointer_diff optimizations use view_convert instead of convert.
8469 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8471 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8472 Generate different code for -mno-speculate-indirect-jumps.
8473 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8474 (*call_indirect_aix<mode>): Disable for
8475 -mno-speculate-indirect-jumps.
8476 (*call_indirect_aix<mode>_nospec): New define_insn.
8477 (*call_value_indirect_aix<mode>): Disable for
8478 -mno-speculate-indirect-jumps.
8479 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8480 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8481 -mno-speculate-indirect-jumps.
8482 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8484 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8486 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8487 long double type, set the flags for noting the default long double
8488 type, even if we don't pass or return a long double type.
8490 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8493 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8496 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8498 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8499 support for merge[hl].
8500 (fold_mergehl_helper): New helper function.
8501 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8502 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8503 (altivec_vmrglw_direct): Add xxmrglw insn.
8505 2018-01-17 Andrew Waterman <andrew@sifive.com>
8507 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8508 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8510 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8513 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8514 call the lto_location_cache before reading the
8515 DECL_SOURCE_LOCATION of the types.
8517 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8518 Richard Sandiford <richard.sandiford@linaro.org>
8520 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8521 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8522 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8523 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8524 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8526 * config/aarch64/constraints.md (aarch64_movti_operand):
8528 * config/aarch64/predicates.md (Uti): Add new constraint.
8530 2018-01-17 Carl Love <cel@us.ibm.com>
8532 * config/rs6000/vsx.md (define_expand xl_len_r,
8533 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8534 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8536 (define_expand, define_insn): Move the shift left from the
8537 define_insn to the define_expand for lxvl and stxvl instructions.
8538 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8539 and XL_LEN_R definitions to PURE.
8541 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8543 * config/i386/i386.c (indirect_thunk_name): Declare regno
8544 as unsigned int. Compare regno with INVALID_REGNUM.
8545 (output_indirect_thunk): Ditto.
8546 (output_indirect_thunk_function): Ditto.
8547 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8548 in the call to output_indirect_thunk_function.
8550 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8553 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8554 rather than the size of inner_type to determine the stack slot size
8555 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8557 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8560 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8563 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8565 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8566 endian Linux systems to optionally enable multilibs for selecting
8567 the long double type if the user configured an explicit type.
8568 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8569 have no long double multilibs if not defined.
8570 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8571 warn if the user used -mabi={ieee,ibm}longdouble and we built
8572 multilibs for long double.
8573 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8574 appropriate multilib option.
8575 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8577 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8578 for building long double multilibs.
8579 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8581 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8583 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8586 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8588 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8591 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8594 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8597 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8599 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8600 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8602 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8604 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8605 different rtl trees depending on TARGET_64BIT.
8606 (rs6000_gen_lvx): Likewise.
8608 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8610 * config/visium/visium.md (nop): Tweak comment.
8611 (hazard_nop): Likewise.
8613 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8615 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8616 -mspeculate-indirect-jumps.
8617 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8618 for -mno-speculate-indirect-jumps.
8619 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8620 (*call_value_indirect_elfv2<mode>): Disable for
8621 -mno-speculate-indirect-jumps.
8622 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8623 (indirect_jump): Emit different RTL for
8624 -mno-speculate-indirect-jumps.
8625 (*indirect_jump<mode>): Disable for
8626 -mno-speculate-indirect-jumps.
8627 (*indirect_jump<mode>_nospec): New define_insn.
8628 (tablejump): Emit different RTL for
8629 -mno-speculate-indirect-jumps.
8630 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8631 (tablejumpsi_nospec): New define_expand.
8632 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8633 (tablejumpdi_nospec): New define_expand.
8634 (*tablejump<mode>_internal1): Disable for
8635 -mno-speculate-indirect-jumps.
8636 (*tablejump<mode>_internal1_nospec): New define_insn.
8637 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8640 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8642 * caller-save.c (insert_save): Drop unnecessary parameter. All
8645 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8646 Richard Biener <rguenth@suse.de>
8649 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8650 return early, inline manually is_gimple_sizepos. Make sure if we
8651 call gimplify_expr we don't end up with a gimple constant.
8652 * tree.c (variably_modified_type_p): Don't return true for
8653 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8654 * gimplify.h (is_gimple_sizepos): Remove.
8656 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8658 PR tree-optimization/83857
8659 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8660 vectorizable_live_operation for pure SLP statements.
8661 (vectorizable_live_operation): Handle PHIs.
8663 2018-01-16 Richard Biener <rguenther@suse.de>
8665 PR tree-optimization/83867
8666 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8667 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8669 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8672 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8673 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8674 If off is not INTEGER_CST, issue a may not be aligned warning
8675 rather than isn't aligned. Use isn%'t rather than isn't.
8676 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8678 <case MULT_EXPR>: Improve the case when bottom and one of the
8679 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8680 operand, in that case check if the other operand is multiple of
8681 bottom divided by the INTEGER_CST operand.
8683 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8686 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8687 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8688 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8689 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8690 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8691 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8692 (pa_function_arg_size): New function.
8694 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8696 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8697 in a separate statement.
8699 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8701 PR tree-optimization/83847
8702 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8703 group gathers and scatters.
8705 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8707 PR rtl-optimization/86620
8708 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8710 PR rtl-optimization/83213
8711 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8712 to last if both are JUMP_INSNs.
8714 PR tree-optimization/83843
8715 * gimple-ssa-store-merging.c
8716 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8717 store_immediate_info for bswap/nop orig_stores.
8719 2018-01-15 Andrew Waterman <andrew@sifive.com>
8721 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8723 <UDIV>: Increase cost if !TARGET_DIV.
8725 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8727 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8728 (define_attr "cr_logical_3op"): New.
8729 (cceq_ior_compare): Adjust.
8730 (cceq_ior_compare_complement): Adjust.
8731 (*cceq_rev_compare): Adjust.
8732 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8733 (is_cracked_insn): Adjust.
8734 (insn_must_be_first_in_group): Adjust.
8735 * config/rs6000/40x.md: Adjust.
8736 * config/rs6000/440.md: Adjust.
8737 * config/rs6000/476.md: Adjust.
8738 * config/rs6000/601.md: Adjust.
8739 * config/rs6000/603.md: Adjust.
8740 * config/rs6000/6xx.md: Adjust.
8741 * config/rs6000/7450.md: Adjust.
8742 * config/rs6000/7xx.md: Adjust.
8743 * config/rs6000/8540.md: Adjust.
8744 * config/rs6000/cell.md: Adjust.
8745 * config/rs6000/e300c2c3.md: Adjust.
8746 * config/rs6000/e500mc.md: Adjust.
8747 * config/rs6000/e500mc64.md: Adjust.
8748 * config/rs6000/e5500.md: Adjust.
8749 * config/rs6000/e6500.md: Adjust.
8750 * config/rs6000/mpc.md: Adjust.
8751 * config/rs6000/power4.md: Adjust.
8752 * config/rs6000/power5.md: Adjust.
8753 * config/rs6000/power6.md: Adjust.
8754 * config/rs6000/power7.md: Adjust.
8755 * config/rs6000/power8.md: Adjust.
8756 * config/rs6000/power9.md: Adjust.
8757 * config/rs6000/rs64.md: Adjust.
8758 * config/rs6000/titan.md: Adjust.
8760 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8762 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8763 ix86_indirect_branch_register logic.
8765 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8767 * config/i386/constraints.md (Bs): Update
8768 ix86_indirect_branch_register check. Don't check
8769 ix86_indirect_branch_register with GOT_memory_operand.
8771 * config/i386/predicates.md (GOT_memory_operand): Don't check
8772 ix86_indirect_branch_register here.
8773 (GOT32_symbol_operand): Likewise.
8775 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8777 * config/i386/predicates.md (constant_call_address_operand):
8778 Rewrite ix86_indirect_branch_register logic.
8779 (sibcall_insn_operand): Likewise.
8781 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8783 * config/i386/constraints.md (Bs): Replace
8784 ix86_indirect_branch_thunk_register with
8785 ix86_indirect_branch_register.
8787 * config/i386/i386.md (indirect_jump): Likewise.
8788 (tablejump): Likewise.
8789 (*sibcall_memory): Likewise.
8790 (*sibcall_value_memory): Likewise.
8791 Peepholes of indirect call and jump via memory: Likewise.
8792 * config/i386/i386.opt: Likewise.
8793 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8794 (GOT_memory_operand): Likewise.
8795 (call_insn_operand): Likewise.
8796 (sibcall_insn_operand): Likewise.
8797 (GOT32_symbol_operand): Likewise.
8799 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8802 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8803 type rather than type addr's type points to.
8804 (expand_omp_atomic_mutex): Likewise.
8805 (expand_omp_atomic): Likewise.
8807 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8810 * config/i386/i386.c (output_indirect_thunk_function): Use
8811 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8812 for __x86_return_thunk.
8814 2018-01-15 Richard Biener <rguenther@suse.de>
8817 * expmed.c (extract_bit_field_1): Fix typo.
8819 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8822 * config/arm/iterators.md (VF): New mode iterator.
8823 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8824 Remove integer-related logic from pattern.
8825 (neon_vabd<mode>_3): Likewise.
8827 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8830 * common.opt (fstrict-overflow): No longer an alias.
8831 (fwrapv-pointer): New option.
8832 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8833 also for pointer types based on flag_wrapv_pointer.
8834 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8835 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8836 opts->x_flag_wrapv got set.
8837 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8838 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8839 POINTER_TYPE_OVERFLOW_UNDEFINED.
8840 * match.pd: Likewise in address comparison pattern.
8841 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8843 2018-01-15 Richard Biener <rguenther@suse.de>
8846 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8847 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8848 Reset type names to their identifier if their TYPE_DECL doesn't
8849 have linkage (and thus is used for ODR and devirt).
8850 (save_debug_info_for_decl): Remove.
8851 (save_debug_info_for_type): Likewise.
8852 (add_tree_to_fld_list): Adjust.
8853 * tree-pretty-print.c (dump_generic_node): Make dumping of
8854 type names more robust.
8856 2018-01-15 Richard Biener <rguenther@suse.de>
8858 * BASE-VER: Bump to 8.0.1.
8860 2018-01-14 Martin Sebor <msebor@redhat.com>
8863 * builtins.c (check_access): Avoid warning when the no-warning bit
8866 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8868 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8869 * ira-color (allocno_hard_regs_compare): Likewise.
8871 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8874 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8875 Use .pushsection/.popsection.
8877 2018-01-14 Martin Sebor <msebor@redhat.com>
8880 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8882 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8884 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8885 entry from extra_headers.
8886 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8887 extra_headers, make the list bitwise identical to the i?86-*-* one.
8889 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8891 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8892 -mcmodel=large with -mindirect-branch=thunk,
8893 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8894 -mfunction-return=thunk-extern.
8895 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8896 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8897 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8899 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8901 * config/i386/i386.c (print_reg): Print the name of the full
8902 integer register without '%'.
8903 (ix86_print_operand): Handle 'V'.
8904 * doc/extend.texi: Document 'V' modifier.
8906 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8908 * config/i386/constraints.md (Bs): Disallow memory operand for
8909 -mindirect-branch-register.
8911 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8912 (GOT_memory_operand): Likewise.
8913 (call_insn_operand): Likewise.
8914 (sibcall_insn_operand): Likewise.
8915 (GOT32_symbol_operand): Likewise.
8916 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8917 for -mindirect-branch-register.
8918 (tablejump): Likewise.
8919 (*sibcall_memory): Likewise.
8920 (*sibcall_value_memory): Likewise.
8921 Disallow peepholes of indirect call and jump via memory for
8922 -mindirect-branch-register.
8923 (*call_pop): Replace m with Bw.
8924 (*call_value_pop): Likewise.
8925 (*sibcall_pop_memory): Replace m with Bs.
8926 * config/i386/i386.opt (mindirect-branch-register): New option.
8927 * doc/invoke.texi: Document -mindirect-branch-register option.
8929 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8931 * config/i386/i386-protos.h (ix86_output_function_return): New.
8932 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8933 set function_return_type.
8934 (indirect_thunk_name): Add ret_p to indicate thunk for function
8936 (output_indirect_thunk_function): Pass false to
8937 indirect_thunk_name.
8938 (ix86_output_indirect_branch_via_reg): Likewise.
8939 (ix86_output_indirect_branch_via_push): Likewise.
8940 (output_indirect_thunk_function): Create alias for function
8941 return thunk if regno < 0.
8942 (ix86_output_function_return): New function.
8943 (ix86_handle_fndecl_attribute): Handle function_return.
8944 (ix86_attribute_table): Add function_return.
8945 * config/i386/i386.h (machine_function): Add
8946 function_return_type.
8947 * config/i386/i386.md (simple_return_internal): Use
8948 ix86_output_function_return.
8949 (simple_return_internal_long): Likewise.
8950 * config/i386/i386.opt (mfunction-return=): New option.
8951 (indirect_branch): Mention -mfunction-return=.
8952 * doc/extend.texi: Document function_return function attribute.
8953 * doc/invoke.texi: Document -mfunction-return= option.
8955 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8957 * config/i386/i386-opts.h (indirect_branch): New.
8958 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8959 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8960 with local indirect jump when converting indirect call and jump.
8961 (ix86_set_indirect_branch_type): New.
8962 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8963 (indirectlabelno): New.
8964 (indirect_thunk_needed): Likewise.
8965 (indirect_thunk_bnd_needed): Likewise.
8966 (indirect_thunks_used): Likewise.
8967 (indirect_thunks_bnd_used): Likewise.
8968 (INDIRECT_LABEL): Likewise.
8969 (indirect_thunk_name): Likewise.
8970 (output_indirect_thunk): Likewise.
8971 (output_indirect_thunk_function): Likewise.
8972 (ix86_output_indirect_branch_via_reg): Likewise.
8973 (ix86_output_indirect_branch_via_push): Likewise.
8974 (ix86_output_indirect_branch): Likewise.
8975 (ix86_output_indirect_jmp): Likewise.
8976 (ix86_code_end): Call output_indirect_thunk_function if needed.
8977 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8979 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8980 (ix86_attribute_table): Add indirect_branch.
8981 * config/i386/i386.h (machine_function): Add indirect_branch_type
8982 and has_local_indirect_jump.
8983 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8985 (tablejump): Likewise.
8986 (*indirect_jump): Use ix86_output_indirect_jmp.
8987 (*tablejump_1): Likewise.
8988 (simple_return_indirect_internal): Likewise.
8989 * config/i386/i386.opt (mindirect-branch=): New option.
8990 (indirect_branch): New.
8993 (thunk-inline): Likewise.
8994 (thunk-extern): Likewise.
8995 * doc/extend.texi: Document indirect_branch function attribute.
8996 * doc/invoke.texi: Document -mindirect-branch= option.
8998 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9001 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9003 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9005 * ipa-inline.c (want_inline_small_function_p): Return false if
9006 inlining has already failed with CIF_FINAL_ERROR.
9007 (update_caller_keys): Call want_inline_small_function_p before
9009 (update_callee_keys): Likewise.
9011 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9013 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9015 (rs6000_quadword_masked_address_p): Likewise.
9016 (quad_aligned_load_p): Likewise.
9017 (quad_aligned_store_p): Likewise.
9018 (const_load_sequence_p): Add comment to describe the outer-most loop.
9019 (mimic_memory_attributes_and_flags): New function.
9020 (rs6000_gen_stvx): Likewise.
9021 (replace_swapped_aligned_store): Likewise.
9022 (rs6000_gen_lvx): Likewise.
9023 (replace_swapped_aligned_load): Likewise.
9024 (replace_swapped_load_constant): Capitalize argument name in
9025 comment describing this function.
9026 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9027 and stores that access quad-word aligned addresses and replace
9028 with stvx or lvx instructions when appropriate.
9029 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9030 New function prototype.
9031 (rs6000_quadword_masked_address_p): Likewise.
9032 (rs6000_gen_lvx): Likewise.
9033 (rs6000_gen_stvx): Likewise.
9034 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9035 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9036 when memory address is aligned.
9037 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9038 this split to select lvx instruction when memory address is aligned.
9039 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9040 instruction when memory address is aligned.
9041 (*vsx_le_perm_load_v16qi): Likewise.
9042 (four unnamed splitters): Modify to select the stvx instruction
9043 when memory is aligned.
9045 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9047 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9048 which appears in the queue multiple times.
9050 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9051 Alan Hayward <alan.hayward@arm.com>
9052 David Sherwood <david.sherwood@arm.com>
9054 * tree-vectorizer.h (vec_lower_bound): New structure.
9055 (_loop_vec_info): Add check_nonzero and lower_bounds.
9056 (LOOP_VINFO_CHECK_NONZERO): New macro.
9057 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9058 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9059 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9060 fields. Make seg_len the distance travelled, not including the
9062 (dr_direction_indicator): Declare.
9063 (dr_zero_step_indicator): Likewise.
9064 (dr_known_forward_stride_p): Likewise.
9065 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9067 (runtime_alias_check_p): Allow runtime alias checks with
9069 (operator ==): Compare access_size and align.
9070 (prune_runtime_alias_test_list): Rework for new distinction between
9071 the access_size and seg_len.
9072 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9074 (get_segment_min_max): New function.
9075 (create_intersect_range_checks): Use it.
9076 (dr_step_indicator): New function.
9077 (dr_direction_indicator): Likewise.
9078 (dr_zero_step_indicator): Likewise.
9079 (dr_known_forward_stride_p): Likewise.
9080 * tree-loop-distribution.c (data_ref_segment_size): Return
9081 DR_STEP * (niters - 1).
9082 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9084 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9085 (vect_preserves_scalar_order_p): New function, split out from...
9086 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9087 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9088 (vect_vfa_access_size): New function.
9089 (vect_vfa_align): Likewise.
9090 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9091 (dump_lower_bound): New function.
9092 (vect_check_lower_bound): Likewise.
9093 (vect_small_gap_p): Likewise.
9094 (vectorizable_with_step_bound_p): Likewise.
9095 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9096 depencies if the vectorization factor is 1. Convert the checks
9097 for nonzero steps into checks on the bounds of DR_STEP. Try using
9098 a bunds check for variable steps if the minimum required step is
9099 relatively small. Update calls to the dr_with_seg_len
9100 constructor and to vect_compile_time_alias.
9101 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9103 (vect_loop_versioning): Call it.
9104 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9106 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9108 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9109 Alan Hayward <alan.hayward@arm.com>
9110 David Sherwood <david.sherwood@arm.com>
9112 * doc/sourcebuild.texi (vect_scatter_store): Document.
9113 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9115 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9117 * genopinit.c (main): Add supports_vec_scatter_store and
9118 supports_vec_scatter_store_cached to target_optabs.
9119 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9120 IFN_MASK_SCATTER_STORE.
9121 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9123 * internal-fn.h (internal_store_fn_p): Declare.
9124 (internal_fn_stored_value_index): Likewise.
9125 * internal-fn.c (scatter_store_direct): New macro.
9126 (expand_scatter_store_optab_fn): New function.
9127 (direct_scatter_store_optab_supported_p): New macro.
9128 (internal_store_fn_p): New function.
9129 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9130 IFN_MASK_SCATTER_STORE.
9131 (internal_fn_mask_index): Likewise.
9132 (internal_fn_stored_value_index): New function.
9133 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9135 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9136 * optabs-query.c (supports_vec_scatter_store_p): New function.
9137 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9138 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9139 true for scatter stores.
9140 (vect_gather_scatter_fn_p): Handle scatter stores too.
9141 (vect_check_gather_scatter): Consider using scatter stores if
9142 supports_vec_scatter_store_p.
9143 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9145 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9146 internal_fn_stored_value_index.
9147 (check_load_store_masking): Handle scatter stores too.
9148 (vect_get_store_rhs): Make public.
9149 (vectorizable_call): Use internal_store_fn_p.
9150 (vectorizable_store): Handle scatter store internal functions.
9151 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9152 when deciding whether the end of the group has been reached.
9153 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9154 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9155 (mask_scatter_store<mode>): New insns.
9157 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9158 Alan Hayward <alan.hayward@arm.com>
9159 David Sherwood <david.sherwood@arm.com>
9161 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9162 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9163 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9165 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9166 Use vect_truncate_gather_scatter_offset if we can't treat the
9167 operation as a normal gather load or scatter store.
9168 (get_group_load_store_type): Take the gather_scatter_info
9169 as argument. Try using a gather load or scatter store for
9170 single-element groups.
9171 (get_load_store_type): Update calls to get_group_load_store_type
9172 and vect_use_strided_gather_scatters_p.
9174 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9175 Alan Hayward <alan.hayward@arm.com>
9176 David Sherwood <david.sherwood@arm.com>
9178 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9179 optional tree argument.
9180 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9182 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9183 but continue to use the current value as a fallback.
9184 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9185 to compare the updates.
9186 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9187 (get_load_store_type): Use it when handling a strided access.
9188 (vect_get_strided_load_store_ops): New function.
9189 (vect_get_data_ptr_increment): Likewise.
9190 (vectorizable_load): Handle strided gather loads. Always pass
9191 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9193 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9194 Alan Hayward <alan.hayward@arm.com>
9195 David Sherwood <david.sherwood@arm.com>
9197 * doc/md.texi (gather_load@var{m}): Document.
9198 (mask_gather_load@var{m}): Likewise.
9199 * genopinit.c (main): Add supports_vec_gather_load and
9200 supports_vec_gather_load_cached to target_optabs.
9201 * optabs-tree.c (init_tree_optimization_optabs): Use
9202 ggc_cleared_alloc to allocate target_optabs.
9203 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9204 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9206 * internal-fn.h (internal_load_fn_p): Declare.
9207 (internal_gather_scatter_fn_p): Likewise.
9208 (internal_fn_mask_index): Likewise.
9209 (internal_gather_scatter_fn_supported_p): Likewise.
9210 * internal-fn.c (gather_load_direct): New macro.
9211 (expand_gather_load_optab_fn): New function.
9212 (direct_gather_load_optab_supported_p): New macro.
9213 (direct_internal_fn_optab): New function.
9214 (internal_load_fn_p): Likewise.
9215 (internal_gather_scatter_fn_p): Likewise.
9216 (internal_fn_mask_index): Likewise.
9217 (internal_gather_scatter_fn_supported_p): Likewise.
9218 * optabs-query.c (supports_at_least_one_mode_p): New function.
9219 (supports_vec_gather_load_p): Likewise.
9220 * optabs-query.h (supports_vec_gather_load_p): Declare.
9221 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9222 and memory_type field.
9223 (NUM_PATTERNS): Bump to 15.
9224 * tree-vect-data-refs.c: Include internal-fn.h.
9225 (vect_gather_scatter_fn_p): New function.
9226 (vect_describe_gather_scatter_call): Likewise.
9227 (vect_check_gather_scatter): Try using internal functions for
9228 gather loads. Recognize existing calls to a gather load function.
9229 (vect_analyze_data_refs): Consider using gather loads if
9230 supports_vec_gather_load_p.
9231 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9232 (vect_get_gather_scatter_offset_type): Likewise.
9233 (vect_convert_mask_for_vectype): Likewise.
9234 (vect_add_conversion_to_patterm): Likewise.
9235 (vect_try_gather_scatter_pattern): Likewise.
9236 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9237 (vect_vect_recog_func_ptrs): Add it.
9238 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9239 internal_fn_mask_index and internal_gather_scatter_fn_p.
9240 (check_load_store_masking): Take the gather_scatter_info as an
9241 argument and handle gather loads.
9242 (vect_get_gather_scatter_ops): New function.
9243 (vectorizable_call): Check internal_load_fn_p.
9244 (vectorizable_load): Likewise. Handle gather load internal
9246 (vectorizable_store): Update call to check_load_store_masking.
9247 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9248 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9249 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9250 (aarch64_gather_scale_operand_d): New predicates.
9251 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9252 (mask_gather_load<mode>): New insns.
9254 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9255 Alan Hayward <alan.hayward@arm.com>
9256 David Sherwood <david.sherwood@arm.com>
9258 * optabs.def (fold_left_plus_optab): New optab.
9259 * doc/md.texi (fold_left_plus_@var{m}): Document.
9260 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9261 * internal-fn.c (fold_left_direct): Define.
9262 (expand_fold_left_optab_fn): Likewise.
9263 (direct_fold_left_optab_supported_p): Likewise.
9264 * fold-const-call.c (fold_const_fold_left): New function.
9265 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9266 * tree-parloops.c (valid_reduction_p): New function.
9267 (gather_scalar_reductions): Use it.
9268 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9269 (vect_finish_replace_stmt): Declare.
9270 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9271 (needs_fold_left_reduction_p): New function, split out from...
9272 (vect_is_simple_reduction): ...here. Accept reductions that
9273 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9274 (vect_force_simple_reduction): Also store the reduction type in
9275 the assignment's STMT_VINFO_REDUC_TYPE.
9276 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9277 (merge_with_identity): New function.
9278 (vect_expand_fold_left): Likewise.
9279 (vectorize_fold_left_reduction): Likewise.
9280 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9281 scalar phi in place for it. Check for target support and reject
9282 cases that would reassociate the operation. Defer the transform
9283 phase to vectorize_fold_left_reduction.
9284 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9285 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9286 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9288 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9290 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9291 call to ifc_temp_var.
9293 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9294 Alan Hayward <alan.hayward@arm.com>
9295 David Sherwood <david.sherwood@arm.com>
9297 * target.def (legitimize_address_displacement): Take the original
9298 offset as a poly_int.
9299 * targhooks.h (default_legitimize_address_displacement): Update
9301 * targhooks.c (default_legitimize_address_displacement): Likewise.
9302 * doc/tm.texi: Regenerate.
9303 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9304 as an argument, moving assert of ad->disp == ad->disp_term to...
9305 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9306 Try calling targetm.legitimize_address_displacement before expanding
9307 the address rather than afterwards, and adjust for the new interface.
9308 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9309 Match the new hook interface. Handle SVE addresses.
9310 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9313 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9315 * Makefile.in (OBJS): Add early-remat.o.
9316 * target.def (select_early_remat_modes): New hook.
9317 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9318 * doc/tm.texi: Regenerate.
9319 * targhooks.h (default_select_early_remat_modes): Declare.
9320 * targhooks.c (default_select_early_remat_modes): New function.
9321 * timevar.def (TV_EARLY_REMAT): New timevar.
9322 * passes.def (pass_early_remat): New pass.
9323 * tree-pass.h (make_pass_early_remat): Declare.
9324 * early-remat.c: New file.
9325 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9327 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9329 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9330 Alan Hayward <alan.hayward@arm.com>
9331 David Sherwood <david.sherwood@arm.com>
9333 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9334 vfm1 with a bound_epilog parameter.
9335 (vect_do_peeling): Update calls accordingly, and move the prologue
9336 call earlier in the function. Treat the base bound_epilog as 0 for
9337 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9338 this base when peeling for gaps.
9339 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9340 with fully-masked loops.
9341 (vect_estimate_min_profitable_iters): Handle the single peeled
9342 iteration in that case.
9344 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9345 Alan Hayward <alan.hayward@arm.com>
9346 David Sherwood <david.sherwood@arm.com>
9348 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9349 single-element interleaving even if the size is not a power of 2.
9350 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9351 accesses for single-element interleaving if the group size is
9354 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9355 Alan Hayward <alan.hayward@arm.com>
9356 David Sherwood <david.sherwood@arm.com>
9358 * doc/md.texi (fold_extract_last_@var{m}): Document.
9359 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9360 * optabs.def (fold_extract_last_optab): New optab.
9361 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9362 * internal-fn.c (fold_extract_direct): New macro.
9363 (expand_fold_extract_optab_fn): Likewise.
9364 (direct_fold_extract_optab_supported_p): Likewise.
9365 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9366 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9367 EXTRACT_LAST_REDUCTION.
9368 (get_initial_def_for_reduction): Do not create an initial vector
9369 for EXTRACT_LAST_REDUCTION reductions.
9370 (vectorizable_reduction): Leave the scalar phi in place for
9371 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9372 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9373 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9374 transform phase to vectorizable_condition.
9375 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9377 (vect_finish_stmt_generation): ...here.
9378 (vect_finish_replace_stmt): New function.
9379 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9380 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9382 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9384 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9385 Alan Hayward <alan.hayward@arm.com>
9386 David Sherwood <david.sherwood@arm.com>
9388 * doc/md.texi (extract_last_@var{m}): Document.
9389 * optabs.def (extract_last_optab): New optab.
9390 * internal-fn.def (EXTRACT_LAST): New internal function.
9391 * internal-fn.c (cond_unary_direct): New macro.
9392 (expand_cond_unary_optab_fn): Likewise.
9393 (direct_cond_unary_optab_supported_p): Likewise.
9394 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9395 loops using EXTRACT_LAST.
9396 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9397 (extract_last_<mode>): ...this optab.
9398 (vec_extract<mode><Vel>): Update accordingly.
9400 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9401 Alan Hayward <alan.hayward@arm.com>
9402 David Sherwood <david.sherwood@arm.com>
9404 * target.def (empty_mask_is_expensive): New hook.
9405 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9406 * doc/tm.texi: Regenerate.
9407 * targhooks.h (default_empty_mask_is_expensive): Declare.
9408 * targhooks.c (default_empty_mask_is_expensive): New function.
9409 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9410 if the target says that empty masks are expensive.
9411 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9413 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9415 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9416 Alan Hayward <alan.hayward@arm.com>
9417 David Sherwood <david.sherwood@arm.com>
9419 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9420 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9421 (vect_use_loop_mask_for_alignment_p): New function.
9422 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9423 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9424 niters_skip argument. Make sure that the first niters_skip elements
9425 of the first iteration are inactive.
9426 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9427 Update call to vect_set_loop_masks_directly.
9428 (get_misalign_in_elems): New function, split out from...
9429 (vect_gen_prolog_loop_niters): ...here.
9430 (vect_update_init_of_dr): Take a code argument that specifies whether
9431 the adjustment should be added or subtracted.
9432 (vect_update_init_of_drs): Likewise.
9433 (vect_prepare_for_masked_peels): New function.
9434 (vect_do_peeling): Skip prologue peeling if we're using a mask
9435 instead. Update call to vect_update_inits_of_drs.
9436 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9438 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9439 alignment. Do not include the number of peeled iterations in
9440 the minimum threshold in that case.
9441 (vectorizable_induction): Adjust the start value down by
9442 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9443 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9444 Take the number of skipped iterations into account when calculating
9446 * tree-vect-stmts.c (vect_gen_while_not): New function.
9448 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9449 Alan Hayward <alan.hayward@arm.com>
9450 David Sherwood <david.sherwood@arm.com>
9452 * doc/sourcebuild.texi (vect_fully_masked): Document.
9453 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9455 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9457 (vect_analyze_loop_2): ...here. Don't check the vectorization
9458 factor against the number of loop iterations if the loop is
9461 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9462 Alan Hayward <alan.hayward@arm.com>
9463 David Sherwood <david.sherwood@arm.com>
9465 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9466 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9467 (dump_groups): Update accordingly.
9468 (iv_use::mem_type): New member variable.
9469 (address_p): New function.
9470 (record_use): Add a mem_type argument and initialize the new
9472 (record_group_use): Add a mem_type argument. Use address_p.
9473 Remove obsolete null checks of base_object. Update call to record_use.
9474 (find_interesting_uses_op): Update call to record_group_use.
9475 (find_interesting_uses_cond): Likewise.
9476 (find_interesting_uses_address): Likewise.
9477 (get_mem_type_for_internal_fn): New function.
9478 (find_address_like_use): Likewise.
9479 (find_interesting_uses_stmt): Try find_address_like_use before
9480 calling find_interesting_uses_op.
9481 (addr_offset_valid_p): Use the iv mem_type field as the type
9482 of the addressed memory.
9483 (add_autoinc_candidates): Likewise.
9484 (get_address_cost): Likewise.
9485 (split_small_address_groups_p): Use address_p.
9486 (split_address_groups): Likewise.
9487 (add_iv_candidate_for_use): Likewise.
9488 (autoinc_possible_for_pair): Likewise.
9489 (rewrite_groups): Likewise.
9490 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9491 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9492 (get_alias_ptr_type_for_ptr_address): New function.
9493 (rewrite_use_address): Rewrite address uses in calls that were
9494 identified by find_address_like_use.
9496 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9497 Alan Hayward <alan.hayward@arm.com>
9498 David Sherwood <david.sherwood@arm.com>
9500 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9502 * gimple-expr.h (is_gimple_addressable: Likewise.
9503 * gimple-expr.c (is_gimple_address): Likewise.
9504 * internal-fn.c (expand_call_mem_ref): New function.
9505 (expand_mask_load_optab_fn): Use it.
9506 (expand_mask_store_optab_fn): Likewise.
9508 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9509 Alan Hayward <alan.hayward@arm.com>
9510 David Sherwood <david.sherwood@arm.com>
9512 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9513 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9514 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9515 (cond_umax@var{mode}): Document.
9516 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9517 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9518 (cond_umin_optab, cond_umax_optab): New optabs.
9519 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9520 (COND_IOR, COND_XOR): New internal functions.
9521 * internal-fn.h (get_conditional_internal_fn): Declare.
9522 * internal-fn.c (cond_binary_direct): New macro.
9523 (expand_cond_binary_optab_fn): Likewise.
9524 (direct_cond_binary_optab_supported_p): Likewise.
9525 (get_conditional_internal_fn): New function.
9526 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9527 Cope with reduction statements that are vectorized as calls rather
9529 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9530 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9531 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9532 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9533 (UNSPEC_COND_EOR): New unspecs.
9534 (optab): Add mappings for them.
9535 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9536 (sve_int_op, sve_fp_op): New int attributes.
9538 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9539 Alan Hayward <alan.hayward@arm.com>
9540 David Sherwood <david.sherwood@arm.com>
9542 * optabs.def (while_ult_optab): New optab.
9543 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9544 * internal-fn.def (WHILE_ULT): New internal function.
9545 * internal-fn.h (direct_internal_fn_supported_p): New override
9546 that takes two types as argument.
9547 * internal-fn.c (while_direct): New macro.
9548 (expand_while_optab_fn): New function.
9549 (convert_optab_supported_p): Likewise.
9550 (direct_while_optab_supported_p): New macro.
9551 * wide-int.h (wi::udiv_ceil): New function.
9552 * tree-vectorizer.h (rgroup_masks): New structure.
9553 (vec_loop_masks): New typedef.
9554 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9556 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9557 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9558 (vect_max_vf): New function.
9559 (slpeel_make_loop_iterate_ntimes): Delete.
9560 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9561 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9562 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9563 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9564 internal-fn.h, stor-layout.h and optabs-query.h.
9565 (vect_set_loop_mask): New function.
9566 (add_preheader_seq): Likewise.
9567 (add_header_seq): Likewise.
9568 (interleave_supported_p): Likewise.
9569 (vect_maybe_permute_loop_masks): Likewise.
9570 (vect_set_loop_masks_directly): Likewise.
9571 (vect_set_loop_condition_masked): Likewise.
9572 (vect_set_loop_condition_unmasked): New function, split out from
9573 slpeel_make_loop_iterate_ntimes.
9574 (slpeel_make_loop_iterate_ntimes): Rename to..
9575 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9576 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9577 (vect_do_peeling): Update call accordingly.
9578 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9580 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9581 mask_compare_type, can_fully_mask_p and fully_masked_p.
9582 (release_vec_loop_masks): New function.
9583 (_loop_vec_info): Use it to free the loop masks.
9584 (can_produce_all_loop_masks_p): New function.
9585 (vect_get_max_nscalars_per_iter): Likewise.
9586 (vect_verify_full_masking): Likewise.
9587 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9588 retries, and free the mask rgroups before retrying. Check loop-wide
9589 reasons for disallowing fully-masked loops. Make the final decision
9590 about whether use a fully-masked loop or not.
9591 (vect_estimate_min_profitable_iters): Do not assume that peeling
9592 for the number of iterations will be needed for fully-masked loops.
9593 (vectorizable_reduction): Disable fully-masked loops.
9594 (vectorizable_live_operation): Likewise.
9595 (vect_halve_mask_nunits): New function.
9596 (vect_double_mask_nunits): Likewise.
9597 (vect_record_loop_mask): Likewise.
9598 (vect_get_loop_mask): Likewise.
9599 (vect_transform_loop): Handle the case in which the final loop
9600 iteration might handle a partial vector. Call vect_set_loop_condition
9601 instead of slpeel_make_loop_iterate_ntimes.
9602 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9603 (check_load_store_masking): New function.
9604 (prepare_load_store_mask): Likewise.
9605 (vectorizable_store): Handle fully-masked loops.
9606 (vectorizable_load): Likewise.
9607 (supportable_widening_operation): Use vect_halve_mask_nunits for
9609 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9610 (vect_gen_while): New function.
9611 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9612 (aarch64_uqdec<mode>): New insn.
9614 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9615 Alan Hayward <alan.hayward@arm.com>
9616 David Sherwood <david.sherwood@arm.com>
9618 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9619 (reduc_xor_scal_optab): New optabs.
9620 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9621 (reduc_xor_scal_@var{m}): Document.
9622 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9623 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9625 * fold-const-call.c (fold_const_call): Handle them.
9626 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9627 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9628 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9629 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9630 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9631 (UNSPEC_XORV): New unspecs.
9632 (optab): Add entries for them.
9633 (BITWISEV): New int iterator.
9634 (bit_reduc_op): New int attributes.
9636 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9637 Alan Hayward <alan.hayward@arm.com>
9638 David Sherwood <david.sherwood@arm.com>
9640 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9641 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9642 * optabs.def (vec_shl_insert_optab): New optab.
9643 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9644 (duplicate_and_interleave): Likewise.
9645 * tree-vect-loop.c: Include internal-fn.h.
9646 (neutral_op_for_slp_reduction): New function, split out from
9647 get_initial_defs_for_reduction.
9648 (get_initial_def_for_reduction): Handle option 2 for variable-length
9649 vectors by loading the neutral value into a vector and then shifting
9650 the initial value into element 0.
9651 (get_initial_defs_for_reduction): Replace the code argument with
9652 the neutral value calculated by neutral_op_for_slp_reduction.
9653 Use gimple_build_vector for constant-length vectors.
9654 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9655 but the first group_size elements have a neutral value.
9656 Use duplicate_and_interleave otherwise.
9657 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9658 Update call to get_initial_defs_for_reduction. Handle SLP
9659 reductions for variable-length vectors by creating one vector
9660 result for each scalar result, with the elements associated
9661 with other scalar results stubbed out with the neutral value.
9662 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9663 Require IFN_VEC_SHL_INSERT for double reductions on
9664 variable-length vectors, or SLP reductions that have
9665 a neutral value. Require can_duplicate_and_interleave_p
9666 support for variable-length unchained SLP reductions if there
9667 is no neutral value, such as for MIN/MAX reductions. Also require
9668 the number of vector elements to be a multiple of the number of
9669 SLP statements when doing variable-length unchained SLP reductions.
9670 Update call to vect_create_epilog_for_reduction.
9671 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9672 and remove initial values.
9673 (duplicate_and_interleave): Make public.
9674 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9675 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9677 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9678 Alan Hayward <alan.hayward@arm.com>
9679 David Sherwood <david.sherwood@arm.com>
9681 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9682 (can_duplicate_and_interleave_p): New function.
9683 (vect_get_and_check_slp_defs): Take the vector of statements
9684 rather than just the current one. Remove excess parentheses.
9685 Restriction rejectinon of vect_constant_def and vect_external_def
9686 for variable-length vectors to boolean types, or types for which
9687 can_duplicate_and_interleave_p is false.
9688 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9689 (duplicate_and_interleave): New function.
9690 (vect_get_constant_vectors): Use gimple_build_vector for
9691 constant-length vectors and suitable variable-length constant
9692 vectors. Use duplicate_and_interleave for other variable-length
9693 vectors. Don't defer the update when inserting new statements.
9695 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9696 Alan Hayward <alan.hayward@arm.com>
9697 David Sherwood <david.sherwood@arm.com>
9699 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9700 min_profitable_iters doesn't go negative.
9702 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9703 Alan Hayward <alan.hayward@arm.com>
9704 David Sherwood <david.sherwood@arm.com>
9706 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9707 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9708 * optabs.def (vec_mask_load_lanes_optab): New optab.
9709 (vec_mask_store_lanes_optab): Likewise.
9710 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9711 (MASK_STORE_LANES): Likewise.
9712 * internal-fn.c (mask_load_lanes_direct): New macro.
9713 (mask_store_lanes_direct): Likewise.
9714 (expand_mask_load_optab_fn): Handle masked operations.
9715 (expand_mask_load_lanes_optab_fn): New macro.
9716 (expand_mask_store_optab_fn): Handle masked operations.
9717 (expand_mask_store_lanes_optab_fn): New macro.
9718 (direct_mask_load_lanes_optab_supported_p): Likewise.
9719 (direct_mask_store_lanes_optab_supported_p): Likewise.
9720 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9722 (vect_load_lanes_supported): Likewise.
9723 * tree-vect-data-refs.c (strip_conversion): New function.
9724 (can_group_stmts_p): Likewise.
9725 (vect_analyze_data_ref_accesses): Use it instead of checking
9726 for a pair of assignments.
9727 (vect_store_lanes_supported): Take a masked_p parameter.
9728 (vect_load_lanes_supported): Likewise.
9729 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9730 vect_store_lanes_supported and vect_load_lanes_supported.
9731 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9732 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9733 parameter. Don't allow gaps for masked accesses.
9734 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9735 and vect_load_lanes_supported.
9736 (get_load_store_type): Take a masked_p parameter and update
9737 call to get_group_load_store_type.
9738 (vectorizable_store): Update call to get_load_store_type.
9739 Handle IFN_MASK_STORE_LANES.
9740 (vectorizable_load): Update call to get_load_store_type.
9741 Handle IFN_MASK_LOAD_LANES.
9743 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9744 Alan Hayward <alan.hayward@arm.com>
9745 David Sherwood <david.sherwood@arm.com>
9747 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9749 * config/aarch64/aarch64-protos.h
9750 (aarch64_sve_struct_memory_operand_p): Declare.
9751 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9752 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9753 (VPRED, vpred): Handle SVE structure modes.
9754 * config/aarch64/constraints.md (Utx): New constraint.
9755 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9756 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9757 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9758 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9759 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9760 structure modes. Split into pieces after RA.
9761 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9762 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9764 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9765 SVE structure modes.
9766 (aarch64_classify_address): Likewise.
9767 (sizetochar): Move earlier in file.
9768 (aarch64_print_operand): Handle SVE register lists.
9769 (aarch64_array_mode): New function.
9770 (aarch64_sve_struct_memory_operand_p): Likewise.
9771 (TARGET_ARRAY_MODE): Redefine.
9773 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9774 Alan Hayward <alan.hayward@arm.com>
9775 David Sherwood <david.sherwood@arm.com>
9777 * target.def (array_mode): New target hook.
9778 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9779 * doc/tm.texi: Regenerate.
9780 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9781 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9782 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9784 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9787 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9788 Alan Hayward <alan.hayward@arm.com>
9789 David Sherwood <david.sherwood@arm.com>
9791 * fold-const.c (fold_binary_loc): Check the argument types
9792 rather than the result type when testing for a vector operation.
9794 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9796 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9797 * doc/tm.texi: Regenerate.
9799 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9800 Alan Hayward <alan.hayward@arm.com>
9801 David Sherwood <david.sherwood@arm.com>
9803 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9804 (sve): Document new AArch64 extension.
9805 * doc/md.texi (w): Extend the description of the AArch64
9806 constraint to include SVE vectors.
9807 (Upl, Upa): Document new AArch64 predicate constraints.
9808 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9810 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9811 (msve-vector-bits=): New option.
9812 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9813 SVE when these are disabled.
9814 (sve): New extension.
9815 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9816 modes. Adjust their number of units based on aarch64_sve_vg.
9817 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9818 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9819 aarch64_addr_query_type.
9820 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9821 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9822 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9823 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9824 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9825 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9826 (aarch64_simd_imm_zero_p): Delete.
9827 (aarch64_check_zero_based_sve_index_immediate): Declare.
9828 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9829 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9830 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9831 (aarch64_sve_float_mul_immediate_p): Likewise.
9832 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9834 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9835 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9836 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9837 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9838 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9839 (aarch64_regmode_natural_size): Likewise.
9840 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9841 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9843 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9844 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9845 for VG and the SVE predicate registers.
9846 (V_ALIASES): Add a "z"-prefixed alias.
9847 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9848 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9849 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9850 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9851 (REG_CLASS_NAMES): Add entries for them.
9852 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9853 and the predicate registers.
9854 (aarch64_sve_vg): Declare.
9855 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9856 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9857 (REGMODE_NATURAL_SIZE): Define.
9858 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9860 * config/aarch64/aarch64.c: Include cfgrtl.h.
9861 (simd_immediate_info): Add a constructor for series vectors,
9862 and an associated step field.
9863 (aarch64_sve_vg): New variable.
9864 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9865 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9866 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9867 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9868 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9869 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9870 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9871 (aarch64_get_mask_mode): New functions.
9872 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9873 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9874 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9875 predicate modes and predicate registers. Explicitly restrict
9876 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9877 to store a vector mode if it is recognized by
9878 aarch64_classify_vector_mode.
9879 (aarch64_regmode_natural_size): New function.
9880 (aarch64_hard_regno_caller_save_mode): Return the original mode
9882 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9883 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9884 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9885 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9887 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9888 does not overlap dest if the function is frame-related. Handle
9890 (aarch64_split_add_offset): New function.
9891 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9892 them aarch64_add_offset.
9893 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9894 and update call to aarch64_sub_sp.
9895 (aarch64_add_cfa_expression): New function.
9896 (aarch64_expand_prologue): Pass extra temporary registers to the
9897 functions above. Handle the case in which we need to emit new
9898 DW_CFA_expressions for registers that were originally saved
9899 relative to the stack pointer, but now have to be expressed
9900 relative to the frame pointer.
9901 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9903 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9904 IP0 and IP1 values for SVE frames.
9905 (aarch64_expand_vec_series): New function.
9906 (aarch64_expand_sve_widened_duplicate): Likewise.
9907 (aarch64_expand_sve_const_vector): Likewise.
9908 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9909 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9910 into the register, rather than emitting a SET directly.
9911 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9912 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9913 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9914 (offset_9bit_signed_scaled_p): New functions.
9915 (aarch64_replicate_bitmask_imm): New function.
9916 (aarch64_bitmask_imm): Use it.
9917 (aarch64_cannot_force_const_mem): Reject expressions involving
9918 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9919 (aarch64_classify_index): Handle SVE indices, by requiring
9920 a plain register index with a scale that matches the element size.
9921 (aarch64_classify_address): Handle SVE addresses. Assert that
9922 the mode of the address is VOIDmode or an integer mode.
9923 Update call to aarch64_classify_symbol.
9924 (aarch64_classify_symbolic_expression): Update call to
9925 aarch64_classify_symbol.
9926 (aarch64_const_vec_all_in_range_p): New function.
9927 (aarch64_print_vector_float_operand): Likewise.
9928 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9929 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9930 and the FP immediates 1.0 and 0.5.
9931 (aarch64_print_address_internal): Handle SVE addresses.
9932 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9933 (aarch64_regno_regclass): Handle predicate registers.
9934 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9936 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9937 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9938 (aarch64_convert_sve_vector_bits): New function.
9939 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9940 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9942 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9943 Handle SVE vector and predicate modes. Accept VL-based constants
9944 that need only one temporary register, and VL offsets that require
9945 no temporary registers.
9946 (aarch64_conditional_register_usage): Mark the predicate registers
9947 as fixed if SVE isn't available.
9948 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9949 Return true for SVE vector and predicate modes.
9950 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9951 rather than an unsigned int. Handle SVE modes.
9952 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9954 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9956 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9957 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9958 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9959 (aarch64_sve_float_mul_immediate_p): New functions.
9960 (aarch64_sve_valid_immediate): New function.
9961 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9962 Explicitly reject structure modes. Check for INDEX constants.
9963 Handle PTRUE and PFALSE constants.
9964 (aarch64_check_zero_based_sve_index_immediate): New function.
9965 (aarch64_simd_imm_zero_p): Delete.
9966 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9967 vector modes. Accept constants in the range of CNT[BHWD].
9968 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9969 ask for an Advanced SIMD mode.
9970 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9971 (aarch64_simd_vector_alignment): Handle SVE predicates.
9972 (aarch64_vectorize_preferred_vector_alignment): New function.
9973 (aarch64_simd_vector_alignment_reachable): Use it instead of
9975 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9976 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9978 (MAX_VECT_LEN): Delete.
9979 (expand_vec_perm_d): Add a vec_flags field.
9980 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9981 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9982 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9984 (aarch64_evpc_rev): Rename to...
9985 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9986 (aarch64_evpc_rev_global): New function.
9987 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9988 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9990 (aarch64_evpc_sve_tbl): New function.
9991 (aarch64_expand_vec_perm_const_1): Update after rename of
9992 aarch64_evpc_rev. Handle SVE permutes too, trying
9993 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9994 than aarch64_evpc_tbl.
9995 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9996 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9997 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9998 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9999 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10000 (aarch64_expand_sve_vcond): New functions.
10001 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10002 of aarch64_vector_mode_p.
10003 (aarch64_dwarf_poly_indeterminate_value): New function.
10004 (aarch64_compute_pressure_classes): Likewise.
10005 (aarch64_can_change_mode_class): Likewise.
10006 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10007 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10008 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10009 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10010 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10011 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10012 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10013 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10015 (Dn, Dl, Dr): Accept const as well as const_vector.
10016 (Dz): Likewise. Compare against CONST0_RTX.
10017 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10018 of "vector" where appropriate.
10019 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10020 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10021 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10022 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10023 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10024 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10025 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10026 (v_int_equiv): Extend to SVE modes.
10027 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10029 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10030 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10031 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10032 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10033 (SVE_COND_FP_CMP): New int iterators.
10034 (perm_hilo): Handle the new unpack unspecs.
10035 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10037 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10038 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10039 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10040 (aarch64_equality_operator, aarch64_constant_vector_operand)
10041 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10042 (aarch64_sve_nonimmediate_operand): Likewise.
10043 (aarch64_sve_general_operand): Likewise.
10044 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10045 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10046 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10047 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10048 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10049 (aarch64_sve_float_arith_immediate): Likewise.
10050 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10051 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10052 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10053 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10054 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10055 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10056 (aarch64_sve_float_arith_operand): Likewise.
10057 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10058 (aarch64_sve_float_mul_operand): Likewise.
10059 (aarch64_sve_vec_perm_operand): Likewise.
10060 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10061 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10062 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10063 as well as const_vector.
10064 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10065 in file. Use CONST0_RTX and CONSTM1_RTX.
10066 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10067 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10068 Use aarch64_simd_imm_zero.
10069 * config/aarch64/aarch64-sve.md: New file.
10070 * config/aarch64/aarch64.md: Include it.
10071 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10072 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10073 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10074 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10075 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10076 (sve): New attribute.
10077 (enabled): Disable instructions with the sve attribute unless
10079 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10080 aarch64_expand_mov_immediate.
10081 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10082 CNT[BHSD] immediates.
10083 (movti): Split CONST_POLY_INT moves into two halves.
10084 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10085 Split additions that need a temporary here if the destination
10086 is the stack pointer.
10087 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10088 (*add<mode>3_poly_1): New instruction.
10089 (set_clobber_cc): New expander.
10091 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10093 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10094 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10095 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10096 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10097 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10098 Change innermode from fixed_mode_size to machine_mode.
10099 (simplify_subreg): Update call accordingly. Handle a constant-sized
10100 subreg of a variable-length CONST_VECTOR.
10102 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10103 Alan Hayward <alan.hayward@arm.com>
10104 David Sherwood <david.sherwood@arm.com>
10106 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10107 (add_offset_to_base): New function, split out from...
10108 (create_mem_ref): ...here. When handling a scale other than 1,
10109 check first whether the address is valid without the offset.
10110 Add it into the base if so, leaving the index and scale as-is.
10112 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10115 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10116 fold_for_warn before checking if arg2 is INTEGER_CST.
10118 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10120 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10121 (store_multiple_operation): Delete.
10122 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10123 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10124 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10125 guarded by TARGET_STRING.
10126 (rs6000_output_load_multiple): Delete.
10127 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10128 OPTION_MASK_STRING / TARGET_STRING handling.
10129 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10130 (const rs6000_opt_masks) <"string">: Change mask to 0.
10131 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10132 (MASK_STRING): Delete.
10133 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10135 (load_multiple): Delete.
10142 (store_multiple): Delete.
10149 (movmemsi_8reg): Delete.
10150 (corresponding unnamed define_insn): Delete.
10151 (movmemsi_6reg): Delete.
10152 (corresponding unnamed define_insn): Delete.
10153 (movmemsi_4reg): Delete.
10154 (corresponding unnamed define_insn): Delete.
10155 (movmemsi_2reg): Delete.
10156 (corresponding unnamed define_insn): Delete.
10157 (movmemsi_1reg): Delete.
10158 (corresponding unnamed define_insn): Delete.
10159 * config/rs6000/rs6000.opt (mno-string): New.
10160 (mstring): Replace by deprecation warning stub.
10161 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10163 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10165 * regrename.c (regrename_do_replace): If replacing the same
10166 reg multiple times, try to reuse last created gen_raw_REG.
10169 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10170 main to workaround a bug in GDB.
10172 2018-01-12 Tom de Vries <tom@codesourcery.com>
10175 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10177 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10179 PR rtl-optimization/80481
10180 * ira-color.c (get_cap_member): New function.
10181 (allocnos_conflict_by_live_ranges_p): Use it.
10182 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10183 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10185 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10188 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10189 (*saddl_se_1): Ditto.
10190 (*ssubsi_1): Ditto.
10191 (*ssubl_se_1): Ditto.
10193 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10195 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10196 rather than wi::to_widest for DR_INITs.
10197 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10198 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10199 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10201 (vect_analyze_group_access_1): Note that here.
10203 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10205 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10206 polynomial type sizes.
10208 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10210 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10211 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10212 (gimple_add_tmp_var): Likewise.
10214 2018-01-12 Martin Liska <mliska@suse.cz>
10216 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10217 (gimple_alloc_sizes): Likewise.
10218 (dump_gimple_statistics): Use PRIu64 in printf format.
10219 * gimple.h: Change uint64_t to int.
10221 2018-01-12 Martin Liska <mliska@suse.cz>
10223 * tree-core.h: Use uint64_t instead of int.
10224 * tree.c (tree_node_counts): Likewise.
10225 (tree_node_sizes): Likewise.
10226 (dump_tree_statistics): Use PRIu64 in printf format.
10228 2018-01-12 Martin Liska <mliska@suse.cz>
10230 * Makefile.in: As qsort_chk is implemented in vec.c, add
10231 vec.o to linkage of gencfn-macros.
10232 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10233 passing the info to record_node_allocation_statistics.
10234 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10236 * ggc-common.c (struct ggc_usage): Add operator== and use
10237 it in operator< and compare function.
10238 * mem-stats.h (struct mem_usage): Likewise.
10239 * vec.c (struct vec_usage): Remove operator< and compare
10240 function. Can be simply inherited.
10242 2018-01-12 Martin Jambor <mjambor@suse.cz>
10245 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10246 * tree-ssa-math-opts.c: Include domwalk.h.
10247 (convert_mult_to_fma_1): New function.
10248 (fma_transformation_info): New type.
10249 (fma_deferring_state): Likewise.
10250 (cancel_fma_deferring): New function.
10251 (result_of_phi): Likewise.
10252 (last_fma_candidate_feeds_initial_phi): Likewise.
10253 (convert_mult_to_fma): Added deferring logic, split actual
10254 transformation to convert_mult_to_fma_1.
10255 (math_opts_dom_walker): New type.
10256 (math_opts_dom_walker::after_dom_children): New method, body moved
10257 here from pass_optimize_widening_mul::execute, added deferring logic
10259 (pass_optimize_widening_mul::execute): Moved most of code to
10260 math_opts_dom_walker::after_dom_children.
10261 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10262 * config/i386/i386.c (ix86_option_override_internal): Added
10263 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10265 2018-01-12 Richard Biener <rguenther@suse.de>
10268 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10269 inline instance vars.
10271 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10274 * config/rx/rx.c (rx_is_restricted_memory_address):
10275 Handle SUBREG case.
10277 2018-01-12 Richard Biener <rguenther@suse.de>
10279 PR tree-optimization/80846
10280 * target.def (split_reduction): New target hook.
10281 * targhooks.c (default_split_reduction): New function.
10282 * targhooks.h (default_split_reduction): Declare.
10283 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10284 target requests first reduce vectors by combining low and high
10286 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10287 (get_vectype_for_scalar_type_and_size): Export.
10288 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10289 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10290 * doc/tm.texi: Regenerate.
10291 * config/i386/i386.c (ix86_split_reduction): Implement
10292 TARGET_VECTORIZE_SPLIT_REDUCTION.
10294 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10297 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10298 in PIC mode except for TARGET_VXWORKS_RTP.
10299 * config/sparc/sparc.c: Include cfgrtl.h.
10300 (TARGET_INIT_PIC_REG): Define.
10301 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10302 (sparc_pic_register_p): New predicate.
10303 (sparc_legitimate_address_p): Use it.
10304 (sparc_legitimize_pic_address): Likewise.
10305 (sparc_delegitimize_address): Likewise.
10306 (sparc_mode_dependent_address_p): Likewise.
10307 (gen_load_pcrel_sym): Remove 4th parameter.
10308 (load_got_register): Adjust call to above. Remove obsolete stuff.
10309 (sparc_expand_prologue): Do not call load_got_register here.
10310 (sparc_flat_expand_prologue): Likewise.
10311 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10312 (sparc_use_pseudo_pic_reg): New function.
10313 (sparc_init_pic_reg): Likewise.
10314 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10315 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10317 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10319 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10320 Add item for branch_cost.
10322 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10324 PR rtl-optimization/83565
10325 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10326 not extend the result to a larger mode for rotate operations.
10327 (num_sign_bit_copies1): Likewise.
10329 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10332 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10334 Use values-Xc.o for -pedantic.
10335 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10337 2018-01-12 Martin Liska <mliska@suse.cz>
10340 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10342 (possible_polymorphic_call_targets): Use it.
10343 (ipa_devirt): Likewise.
10345 2018-01-12 Martin Liska <mliska@suse.cz>
10347 * profile-count.h (enum profile_quality): Use 0 as invalid
10348 enum value of profile_quality.
10350 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10352 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10353 -mext-string options.
10355 2018-01-12 Richard Biener <rguenther@suse.de>
10357 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10358 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10359 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10361 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10363 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10365 * configure.ac (--with-long-double-format): Add support for the
10366 configuration option to change the default long double format on
10368 * config.gcc (powerpc*-linux*-*): Likewise.
10369 * configure: Regenerate.
10370 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10371 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10372 used without modification.
10374 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10376 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10377 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10378 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10379 MISC_BUILTIN_SPEC_BARRIER.
10380 (rs6000_init_builtins): Likewise.
10381 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10383 (speculation_barrier): New define_insn.
10384 * doc/extend.texi: Document __builtin_speculation_barrier.
10386 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10389 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10390 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10391 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10393 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10394 integral modes instead of "ss" and "sd".
10395 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10396 vectors with 32-bit and 64-bit elements.
10397 (vecdupssescalarmodesuffix): New mode attribute.
10398 (vec_dup<mode>): Use it.
10400 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10403 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10404 frame if argument is passed on stack.
10406 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10409 * ree.c (combine_reaching_defs): Optimize also
10410 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10411 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10413 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10415 PR middle-end/83189
10416 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10418 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10420 PR middle-end/83718
10421 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10422 after they are computed.
10424 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10426 PR tree-optimization/83695
10427 * gimple-loop-linterchange.cc
10428 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10429 reset cached scev information after interchange.
10430 (pass_linterchange::execute): Remove call to scev_reset_htab.
10432 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10434 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10435 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10436 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10437 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10438 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10439 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10440 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10441 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10442 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10443 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10444 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10445 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10446 (V_lane_reg): Likewise.
10447 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10449 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10450 (vfmal_lane_low<mode>_intrinsic,
10451 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10452 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10453 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10454 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10455 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10456 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10458 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10460 * config/arm/arm-cpus.in (fp16fml): New feature.
10461 (ALL_SIMD): Add fp16fml.
10462 (armv8.2-a): Add fp16fml as an option.
10463 (armv8.3-a): Likewise.
10464 (armv8.4-a): Add fp16fml as part of fp16.
10465 * config/arm/arm.h (TARGET_FP16FML): Define.
10466 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10468 * config/arm/arm-modes.def (V2HF): Define.
10469 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10470 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10471 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10472 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10473 vfmsl_low, vfmsl_high): New set of builtins.
10474 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10475 (vfml_op): New code attribute.
10476 (VFMLHALVES): New int iterator.
10477 (VFML, VFMLSEL): New mode attributes.
10478 (V_reg): Define mapping for V2HF.
10479 (V_hi, V_lo): New mode attributes.
10480 (VF_constraint): Likewise.
10481 (vfml_half, vfml_half_selector): New int attributes.
10482 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10484 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10485 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10487 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10488 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10489 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10490 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10492 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10493 Document new effective target and option set.
10495 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10497 * config/arm/arm-cpus.in (armv8_4): New feature.
10498 (ARMv8_4a): New fgroup.
10499 (armv8.4-a): New arch.
10500 * config/arm/arm-tables.opt: Regenerate.
10501 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10502 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10503 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10504 Add matching rules for -march=armv8.4-a and extensions.
10505 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10507 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10510 * config/rx/rx.md (BW): New mode attribute.
10511 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10513 2018-01-11 Richard Biener <rguenther@suse.de>
10515 PR tree-optimization/83435
10516 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10517 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10518 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10520 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10521 Alan Hayward <alan.hayward@arm.com>
10522 David Sherwood <david.sherwood@arm.com>
10524 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10526 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10527 (aarch64_print_address_internal): Use it to check for a zero offset.
10529 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10530 Alan Hayward <alan.hayward@arm.com>
10531 David Sherwood <david.sherwood@arm.com>
10533 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10534 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10535 Return a poly_int64 rather than a HOST_WIDE_INT.
10536 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10537 rather than a HOST_WIDE_INT.
10538 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10539 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10540 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10541 final_offset from HOST_WIDE_INT to poly_int64.
10542 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10543 to_constant when getting the number of units in an Advanced SIMD
10545 (aarch64_builtin_vectorized_function): Check for a constant number
10547 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10549 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10550 attribute instead of GET_MODE_NUNITS.
10551 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10552 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10553 GET_MODE_SIZE for fixed-size registers.
10554 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10555 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10556 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10557 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10558 (aarch64_print_operand, aarch64_print_address_internal)
10559 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10560 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10561 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10562 Handle polynomial GET_MODE_SIZE.
10563 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10564 wider than SImode without modification.
10565 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10566 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10567 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10568 passing and returning SVE modes.
10569 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10570 rather than GEN_INT.
10571 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10572 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10573 (aarch64_allocate_and_probe_stack_space): Likewise.
10574 (aarch64_layout_frame): Cope with polynomial offsets.
10575 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10576 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10577 polynomial offsets.
10578 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10579 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10580 poly_int64 rather than a HOST_WIDE_INT.
10581 (aarch64_get_separate_components, aarch64_process_components)
10582 (aarch64_expand_prologue, aarch64_expand_epilogue)
10583 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10584 (aarch64_anchor_offset): New function, split out from...
10585 (aarch64_legitimize_address): ...here.
10586 (aarch64_builtin_vectorization_cost): Handle polynomial
10587 TYPE_VECTOR_SUBPARTS.
10588 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10590 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10591 number of elements from the PARALLEL rather than the mode.
10592 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10593 rather than GET_MODE_BITSIZE.
10594 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10595 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10596 (aarch64_expand_vec_perm_const_1): Handle polynomial
10597 d->perm.length () and d->perm elements.
10598 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10599 Apply to_constant to d->perm elements.
10600 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10601 polynomial CONST_VECTOR_NUNITS.
10602 (aarch64_move_pointer): Take amount as a poly_int64 rather
10604 (aarch64_progress_pointer): Avoid temporary variable.
10605 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10606 the mode attribute instead of GET_MODE.
10608 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10609 Alan Hayward <alan.hayward@arm.com>
10610 David Sherwood <david.sherwood@arm.com>
10612 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10613 x exists before using it.
10614 (aarch64_add_constant_internal): Rename to...
10615 (aarch64_add_offset_1): ...this. Replace regnum with separate
10616 src and dest rtxes. Handle the case in which they're different,
10617 including when the offset is zero. Replace scratchreg with an rtx.
10618 Use 2 additions if there is no spare register into which we can
10619 move a 16-bit constant.
10620 (aarch64_add_constant): Delete.
10621 (aarch64_add_offset): Replace reg with separate src and dest
10622 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10623 Use aarch64_add_offset_1.
10624 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10625 an rtx rather than an int. Take the delta as a poly_int64
10626 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10627 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10628 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10629 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10630 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10631 and aarch64_add_sp.
10632 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10633 aarch64_add_constant.
10635 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10637 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10638 Use scalar_float_mode.
10640 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10642 * config/aarch64/aarch64-simd.md
10643 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10644 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10645 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10646 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10647 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10648 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10649 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10650 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10651 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10652 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10654 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10657 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10658 targ_options->x_arm_arch_string is non NULL.
10660 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10662 * config/aarch64/aarch64.h
10663 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10665 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10668 * expmed.c (emit_store_flag_force): Swap if const op0
10669 and change VOIDmode to mode of op0.
10671 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10673 PR rtl-optimization/83761
10674 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10675 than bytes to mode_for_size.
10677 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10679 PR middle-end/83189
10680 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10681 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10684 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10686 PR middle-end/83575
10687 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10688 when in layout mode.
10689 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10690 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10693 2018-01-10 Michael Collison <michael.collison@arm.com>
10695 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10696 * config/aarch64/aarch64-option-extension.def: Add
10697 AARCH64_OPT_EXTENSION of 'fp16fml'.
10698 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10699 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10700 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10701 * config/aarch64/constraints.md (Ui7): New constraint.
10702 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10703 (VFMLA_SEL_W): Ditto.
10706 (VFMLA16_LOW): New int iterator.
10707 (VFMLA16_HIGH): Ditto.
10708 (UNSPEC_FMLAL): New unspec.
10709 (UNSPEC_FMLSL): Ditto.
10710 (UNSPEC_FMLAL2): Ditto.
10711 (UNSPEC_FMLSL2): Ditto.
10712 (f16mac): New code attribute.
10713 * config/aarch64/aarch64-simd-builtins.def
10714 (aarch64_fmlal_lowv2sf): Ditto.
10715 (aarch64_fmlsl_lowv2sf): Ditto.
10716 (aarch64_fmlalq_lowv4sf): Ditto.
10717 (aarch64_fmlslq_lowv4sf): Ditto.
10718 (aarch64_fmlal_highv2sf): Ditto.
10719 (aarch64_fmlsl_highv2sf): Ditto.
10720 (aarch64_fmlalq_highv4sf): Ditto.
10721 (aarch64_fmlslq_highv4sf): Ditto.
10722 (aarch64_fmlal_lane_lowv2sf): Ditto.
10723 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10724 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10725 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10726 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10727 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10728 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10729 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10730 (aarch64_fmlal_lane_highv2sf): Ditto.
10731 (aarch64_fmlsl_lane_highv2sf): Ditto.
10732 (aarch64_fmlal_laneq_highv2sf): Ditto.
10733 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10734 (aarch64_fmlalq_lane_highv4sf): Ditto.
10735 (aarch64_fmlsl_lane_highv4sf): Ditto.
10736 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10737 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10738 * config/aarch64/aarch64-simd.md:
10739 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10740 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10741 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10742 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10743 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10744 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10745 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10746 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10747 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10748 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10749 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10750 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10751 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10752 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10753 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10754 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10755 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10756 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10757 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10758 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10759 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10760 (vfmlsl_low_u32): Ditto.
10761 (vfmlalq_low_u32): Ditto.
10762 (vfmlslq_low_u32): Ditto.
10763 (vfmlal_high_u32): Ditto.
10764 (vfmlsl_high_u32): Ditto.
10765 (vfmlalq_high_u32): Ditto.
10766 (vfmlslq_high_u32): Ditto.
10767 (vfmlal_lane_low_u32): Ditto.
10768 (vfmlsl_lane_low_u32): Ditto.
10769 (vfmlal_laneq_low_u32): Ditto.
10770 (vfmlsl_laneq_low_u32): Ditto.
10771 (vfmlalq_lane_low_u32): Ditto.
10772 (vfmlslq_lane_low_u32): Ditto.
10773 (vfmlalq_laneq_low_u32): Ditto.
10774 (vfmlslq_laneq_low_u32): Ditto.
10775 (vfmlal_lane_high_u32): Ditto.
10776 (vfmlsl_lane_high_u32): Ditto.
10777 (vfmlal_laneq_high_u32): Ditto.
10778 (vfmlsl_laneq_high_u32): Ditto.
10779 (vfmlalq_lane_high_u32): Ditto.
10780 (vfmlslq_lane_high_u32): Ditto.
10781 (vfmlalq_laneq_high_u32): Ditto.
10782 (vfmlslq_laneq_high_u32): Ditto.
10783 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10784 (AARCH64_FL_FOR_ARCH8_4): New.
10785 (AARCH64_ISA_F16FML): New ISA flag.
10786 (TARGET_F16FML): New feature flag for fp16fml.
10787 (doc/invoke.texi): Document new fp16fml option.
10789 2018-01-10 Michael Collison <michael.collison@arm.com>
10791 * config/aarch64/aarch64-builtins.c:
10792 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10793 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10794 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10795 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10796 (AARCH64_ISA_SHA3): New ISA flag.
10797 (TARGET_SHA3): New feature flag for sha3.
10798 * config/aarch64/iterators.md (sha512_op): New int attribute.
10799 (CRYPTO_SHA512): New int iterator.
10800 (UNSPEC_SHA512H): New unspec.
10801 (UNSPEC_SHA512H2): Ditto.
10802 (UNSPEC_SHA512SU0): Ditto.
10803 (UNSPEC_SHA512SU1): Ditto.
10804 * config/aarch64/aarch64-simd-builtins.def
10805 (aarch64_crypto_sha512hqv2di): New builtin.
10806 (aarch64_crypto_sha512h2qv2di): Ditto.
10807 (aarch64_crypto_sha512su0qv2di): Ditto.
10808 (aarch64_crypto_sha512su1qv2di): Ditto.
10809 (aarch64_eor3qv8hi): Ditto.
10810 (aarch64_rax1qv2di): Ditto.
10811 (aarch64_xarqv2di): Ditto.
10812 (aarch64_bcaxqv8hi): Ditto.
10813 * config/aarch64/aarch64-simd.md:
10814 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10815 (aarch64_crypto_sha512su0qv2di): Ditto.
10816 (aarch64_crypto_sha512su1qv2di): Ditto.
10817 (aarch64_eor3qv8hi): Ditto.
10818 (aarch64_rax1qv2di): Ditto.
10819 (aarch64_xarqv2di): Ditto.
10820 (aarch64_bcaxqv8hi): Ditto.
10821 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10822 (vsha512h2q_u64): Ditto.
10823 (vsha512su0q_u64): Ditto.
10824 (vsha512su1q_u64): Ditto.
10825 (veor3q_u16): Ditto.
10826 (vrax1q_u64): Ditto.
10827 (vxarq_u64): Ditto.
10828 (vbcaxq_u16): Ditto.
10829 * config/arm/types.md (crypto_sha512): New type attribute.
10830 (crypto_sha3): Ditto.
10831 (doc/invoke.texi): Document new sha3 option.
10833 2018-01-10 Michael Collison <michael.collison@arm.com>
10835 * config/aarch64/aarch64-builtins.c:
10836 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10837 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10838 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10839 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10840 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10841 (AARCH64_ISA_SM4): New ISA flag.
10842 (TARGET_SM4): New feature flag for sm4.
10843 * config/aarch64/aarch64-simd-builtins.def
10844 (aarch64_sm3ss1qv4si): Ditto.
10845 (aarch64_sm3tt1aq4si): Ditto.
10846 (aarch64_sm3tt1bq4si): Ditto.
10847 (aarch64_sm3tt2aq4si): Ditto.
10848 (aarch64_sm3tt2bq4si): Ditto.
10849 (aarch64_sm3partw1qv4si): Ditto.
10850 (aarch64_sm3partw2qv4si): Ditto.
10851 (aarch64_sm4eqv4si): Ditto.
10852 (aarch64_sm4ekeyqv4si): Ditto.
10853 * config/aarch64/aarch64-simd.md:
10854 (aarch64_sm3ss1qv4si): Ditto.
10855 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10856 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10857 (aarch64_sm4eqv4si): Ditto.
10858 (aarch64_sm4ekeyqv4si): Ditto.
10859 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10860 (sm3part_op): Ditto.
10861 (CRYPTO_SM3TT): Ditto.
10862 (CRYPTO_SM3PART): Ditto.
10863 (UNSPEC_SM3SS1): New unspec.
10864 (UNSPEC_SM3TT1A): Ditto.
10865 (UNSPEC_SM3TT1B): Ditto.
10866 (UNSPEC_SM3TT2A): Ditto.
10867 (UNSPEC_SM3TT2B): Ditto.
10868 (UNSPEC_SM3PARTW1): Ditto.
10869 (UNSPEC_SM3PARTW2): Ditto.
10870 (UNSPEC_SM4E): Ditto.
10871 (UNSPEC_SM4EKEY): Ditto.
10872 * config/aarch64/constraints.md (Ui2): New constraint.
10873 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10874 * config/arm/types.md (crypto_sm3): New type attribute.
10875 (crypto_sm4): Ditto.
10876 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10877 (vsm3tt1aq_u32): Ditto.
10878 (vsm3tt1bq_u32): Ditto.
10879 (vsm3tt2aq_u32): Ditto.
10880 (vsm3tt2bq_u32): Ditto.
10881 (vsm3partw1q_u32): Ditto.
10882 (vsm3partw2q_u32): Ditto.
10883 (vsm4eq_u32): Ditto.
10884 (vsm4ekeyq_u32): Ditto.
10885 (doc/invoke.texi): Document new sm4 option.
10887 2018-01-10 Michael Collison <michael.collison@arm.com>
10889 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10890 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10891 (AARCH64_FL_FOR_ARCH8_4): New.
10892 (AARCH64_FL_V8_4): New flag.
10893 (doc/invoke.texi): Document new armv8.4-a option.
10895 2018-01-10 Michael Collison <michael.collison@arm.com>
10897 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10898 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10899 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10900 * config/aarch64/aarch64-option-extension.def: Add
10901 AARCH64_OPT_EXTENSION of 'sha2'.
10902 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10903 (crypto): Disable sha2 and aes if crypto disabled.
10904 (crypto): Enable aes and sha2 if enabled.
10905 (simd): Disable sha2 and aes if simd disabled.
10906 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10908 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10909 (TARGET_SHA2): New feature flag for sha2.
10910 (TARGET_AES): New feature flag for aes.
10911 * config/aarch64/aarch64-simd.md:
10912 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10913 conditional on TARGET_AES.
10914 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10915 (aarch64_crypto_sha1hsi): Make pattern conditional
10917 (aarch64_crypto_sha1hv4si): Ditto.
10918 (aarch64_be_crypto_sha1hv4si): Ditto.
10919 (aarch64_crypto_sha1su1v4si): Ditto.
10920 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10921 (aarch64_crypto_sha1su0v4si): Ditto.
10922 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10923 (aarch64_crypto_sha256su0v4si): Ditto.
10924 (aarch64_crypto_sha256su1v4si): Ditto.
10925 (doc/invoke.texi): Document new aes and sha2 options.
10927 2018-01-10 Martin Sebor <msebor@redhat.com>
10929 PR tree-optimization/83781
10930 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10933 2018-01-11 Martin Sebor <msebor@gmail.com>
10934 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10936 PR tree-optimization/83501
10937 PR tree-optimization/81703
10939 * tree-ssa-strlen.c (get_string_cst): Rename...
10940 (get_string_len): ...to this. Handle global constants.
10941 (handle_char_store): Adjust.
10943 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10944 Jim Wilson <jimw@sifive.com>
10946 * config/riscv/riscv-protos.h (riscv_output_return): New.
10947 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10948 (riscv_attribute_table, riscv_output_return),
10949 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10950 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10951 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10953 (riscv_expand_prologue): Add early return for naked function.
10954 (riscv_expand_epilogue): Likewise.
10955 (riscv_function_ok_for_sibcall): Return false for naked function.
10956 (riscv_set_current_function): New.
10957 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10958 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10959 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10960 * doc/extend.texi (RISC-V Function Attributes): New.
10962 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10964 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10965 check for 128-bit long double before checking TCmode.
10966 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10967 128-bit long doubles before checking TFmode or TCmode.
10968 (FLOAT128_IBM_P): Likewise.
10970 2018-01-10 Martin Sebor <msebor@redhat.com>
10972 PR tree-optimization/83671
10973 * builtins.c (c_strlen): Unconditionally return zero for the empty
10975 Use -Warray-bounds for warnings.
10976 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10977 for non-constant array indices with COMPONENT_REF, arrays of
10978 arrays, and pointers to arrays.
10979 (gimple_fold_builtin_strlen): Determine and set length range for
10980 non-constant character arrays.
10982 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10984 PR middle-end/81897
10985 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10988 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10990 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10992 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10995 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10996 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10997 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10998 indexed_or_indirect_operand predicate.
10999 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11000 (*vsx_le_perm_load_v8hi): Likewise.
11001 (*vsx_le_perm_load_v16qi): Likewise.
11002 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11003 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11004 (*vsx_le_perm_store_v8hi): Likewise.
11005 (*vsx_le_perm_store_v16qi): Likewise.
11006 (eight unnamed splitters): Likewise.
11008 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11010 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11011 * config/rs6000/emmintrin.h: Likewise.
11012 * config/rs6000/mmintrin.h: Likewise.
11013 * config/rs6000/xmmintrin.h: Likewise.
11015 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11018 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11020 * tree.c (tree_nop_conversion): Return true for location wrapper
11022 (maybe_wrap_with_location): New function.
11023 (selftest::check_strip_nops): New function.
11024 (selftest::test_location_wrappers): New function.
11025 (selftest::tree_c_tests): Call it.
11026 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11027 (maybe_wrap_with_location): New decl.
11028 (EXPR_LOCATION_WRAPPER_P): New macro.
11029 (location_wrapper_p): New inline function.
11030 (tree_strip_any_location_wrapper): New inline function.
11032 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11035 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11036 stack_realign_offset for the largest alignment of stack slot
11038 (ix86_find_max_used_stack_alignment): New function.
11039 (ix86_finalize_stack_frame_flags): Use it. Set
11040 max_used_stack_alignment if we don't realign stack.
11041 * config/i386/i386.h (machine_function): Add
11042 max_used_stack_alignment.
11044 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11046 * config/arm/arm.opt (-mbranch-cost): New option.
11047 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11050 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11053 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11054 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11056 2018-01-10 Richard Biener <rguenther@suse.de>
11059 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11060 early out so it also covers the case where we have a non-NULL
11063 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11065 PR tree-optimization/83753
11066 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11067 for non-strided grouped accesses if the number of elements is 1.
11069 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11072 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11073 * i386.h (TARGET_USE_GATHER): Define.
11074 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11076 2018-01-10 Martin Liska <mliska@suse.cz>
11079 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11080 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11082 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11083 CLEANUP_NO_PARTITIONING is not set.
11085 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11087 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11088 for vectors, as a partial revert of r254296.
11089 * rtl.h (const_vec_p): Delete.
11090 (const_vec_duplicate_p): Don't test for vector CONSTs.
11091 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11092 * expmed.c (make_tree): Likewise.
11095 * common.md (E, F): Use CONSTANT_P instead of checking for
11097 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11098 checking for CONST_VECTOR.
11100 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11102 PR middle-end/83575
11103 * predict.c (force_edge_cold): Handle in more sane way edges
11104 with no prediction.
11106 2018-01-09 Carl Love <cel@us.ibm.com>
11108 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11110 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11111 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11112 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11113 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11114 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11115 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11116 * config/rs6000/rs6000-protos.h: Add extern defition for
11117 rs6000_generate_float2_double_code.
11118 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11120 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11121 (float2_v2df): Add define_expand.
11123 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11126 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11127 op_mode in the force_to_mode call.
11129 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11131 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11132 instead of checking each element individually.
11133 (aarch64_evpc_uzp): Likewise.
11134 (aarch64_evpc_zip): Likewise.
11135 (aarch64_evpc_ext): Likewise.
11136 (aarch64_evpc_rev): Likewise.
11137 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11138 instead of checking each element individually. Return true without
11140 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11141 whether all selected elements come from the same input, instead of
11142 checking each element individually. Remove calls to gen_rtx_REG,
11143 start_sequence and end_sequence and instead assert that no rtl is
11146 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11148 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11149 order of HIGH and CONST checks.
11151 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11153 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11154 if the destination isn't an SSA_NAME.
11156 2018-01-09 Richard Biener <rguenther@suse.de>
11158 PR tree-optimization/83668
11159 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11161 (canonicalize_loop_form): ... here, renamed from ...
11162 (canonicalize_loop_closed_ssa_form): ... this and amended to
11163 swap successor edges for loop exit blocks to make us use
11164 the RPO order we need for initial schedule generation.
11166 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11168 PR tree-optimization/64811
11169 * match.pd: When optimizing comparisons with Inf, avoid
11170 introducing or losing exceptions from comparisons with NaN.
11172 2018-01-09 Martin Liska <mliska@suse.cz>
11175 * asan.c (shadow_mem_size): Add gcc_assert.
11177 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11179 Don't save registers in main().
11182 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11183 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11184 * config/avr/avr.c (avr_set_current_function): Don't error if
11185 naked, OS_task or OS_main are specified at the same time.
11186 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11188 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11190 * common/config/avr/avr-common.c (avr_option_optimization_table):
11191 Switch on -mmain-is-OS_task for optimizing compilations.
11193 2018-01-09 Richard Biener <rguenther@suse.de>
11195 PR tree-optimization/83572
11196 * graphite.c: Include cfganal.h.
11197 (graphite_transform_loops): Connect infinite loops to exit
11198 and remove fake edges at the end.
11200 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11202 * ipa-inline.c (edge_badness): Revert accidental checkin.
11204 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11207 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11208 symbols; not inline clones.
11210 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11213 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11214 hard registers. Formatting fixes.
11216 PR preprocessor/83722
11217 * gcc.c (try_generate_repro): Pass
11218 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11219 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11222 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11223 Kito Cheng <kito.cheng@gmail.com>
11225 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11226 (riscv_leaf_function_p): Delete.
11227 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11229 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11231 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11233 (do_ifelse): New function.
11234 (do_isel): New function.
11235 (do_sub3): New function.
11236 (do_add3): New function.
11237 (do_load_mask_compare): New function.
11238 (do_overlap_load_compare): New function.
11239 (expand_compare_loop): New function.
11240 (expand_block_compare): Call expand_compare_loop() when appropriate.
11241 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11242 option description.
11243 (-mblock-compare-inline-loop-limit): New option.
11245 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11248 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11249 Reverse order of second and third operands in first alternative.
11250 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11251 of first and second elements in UNSPEC_VPERMR vector.
11252 (altivec_expand_vec_perm_le): Likewise.
11254 2018-01-08 Jeff Law <law@redhat.com>
11256 PR rtl-optimizatin/81308
11257 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11258 (process_switch): If group_case_labels makes a change, then set
11260 (pass_convert_switch::execute): If a switch is converted, then
11261 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11263 PR rtl-optimization/81308
11264 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11267 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11269 PR target/83663 - Revert r255946
11270 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11271 generation for cases where splatting a value is not useful.
11272 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11273 across a vec_duplicate and a paradoxical subreg forming a vector
11274 mode to a vec_concat.
11276 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11278 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11279 -march=armv8.3-a variants.
11280 * config/arm/t-multilib: Likewise.
11281 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11283 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11285 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11287 (cceq_ior_compare_complement): Give it a name so I can use it, and
11288 change boolean_or_operator predicate to boolean_operator so it can
11289 be used to generate a crand.
11290 (eqne): New code iterator.
11291 (bd/bd_neg): New code_attrs.
11292 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11293 a single define_insn.
11294 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11295 decrement (bdnzt/bdnzf/bdzt/bdzf).
11296 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11297 with the new names of the branch decrement patterns, and added the
11298 names of the branch decrement conditional patterns.
11300 2018-01-08 Richard Biener <rguenther@suse.de>
11302 PR tree-optimization/83563
11303 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11306 2018-01-08 Richard Biener <rguenther@suse.de>
11308 PR middle-end/83713
11309 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11311 2018-01-08 Richard Biener <rguenther@suse.de>
11313 PR tree-optimization/83685
11314 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11315 references to abnormals.
11317 2018-01-08 Richard Biener <rguenther@suse.de>
11320 * dwarf2out.c (output_indirect_strings): Handle empty
11321 skeleton_debug_str_hash.
11322 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11324 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11326 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11327 (emit_store_direct): Likewise.
11328 (arc_trampoline_adjust_address): Likewise.
11329 (arc_asm_trampoline_template): New function.
11330 (arc_initialize_trampoline): Use asm_trampoline_template.
11331 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11332 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11333 * config/arc/arc.md (flush_icache): Delete pattern.
11335 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11337 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11338 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11341 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11344 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11345 by not USED_FOR_TARGET.
11346 (make_pass_resolve_sw_modes): Likewise.
11348 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11350 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11353 2018-01-08 Richard Biener <rguenther@suse.de>
11355 PR middle-end/83580
11356 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11358 2018-01-08 Richard Biener <rguenther@suse.de>
11360 PR middle-end/83517
11361 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11363 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11365 PR middle-end/81897
11366 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11367 basic blocks with a small number of successors.
11368 (convert_control_dep_chain_into_preds): Improve handling of
11370 (dump_predicates): Split apart into...
11371 (dump_pred_chain): ...here...
11372 (dump_pred_info): ...and here.
11373 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11374 (can_chain_union_be_invalidated_p): Improve check for invalidation
11376 (uninit_uses_cannot_happen): Avoid unnecessary if
11377 convert_control_dep_chain_into_preds yielded nothing.
11379 2018-01-06 Martin Sebor <msebor@redhat.com>
11381 PR tree-optimization/83640
11382 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11383 subtracting negative offset from size.
11384 (builtin_access::overlap): Adjust offset bounds of the access to fall
11385 within the size of the object if possible.
11387 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11389 PR rtl-optimization/83699
11390 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11391 extract_bit_field_as_subreg to cases in which the extracted
11392 value is also a vector.
11394 * lra-constraints.c (process_alt_operands): Test for the equivalence
11395 substitutions when detecting a possible reload cycle.
11397 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11400 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11401 by default if flag_selective_schedling{,2}. Formatting fixes.
11403 PR rtl-optimization/83682
11404 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11405 if it has non-VECTOR_MODE element mode.
11406 (vec_duplicate_p): Likewise.
11408 PR middle-end/83694
11409 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11410 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11412 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11415 * config/i386/i386-builtin.def
11416 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11417 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11418 Require also OPTION_MASK_ISA_AVX512F in addition to
11419 OPTION_MASK_ISA_GFNI.
11420 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11421 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11422 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11423 to OPTION_MASK_ISA_GFNI.
11424 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11425 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11426 OPTION_MASK_ISA_AVX512BW.
11427 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11428 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11429 addition to OPTION_MASK_ISA_GFNI.
11430 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11431 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11432 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11433 to OPTION_MASK_ISA_GFNI.
11434 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11435 a requirement for all ISAs rather than any of them with a few
11437 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11439 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11440 bitmasks to be enabled with 3 exceptions, instead of requiring any
11441 enabled ISA with lots of exceptions.
11442 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11443 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11444 Change avx512bw in isa attribute to avx512f.
11445 * config/i386/sgxintrin.h: Add license boilerplate.
11446 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11447 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11448 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11449 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11451 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11452 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11453 temporarily sse2 rather than sse if not enabled already.
11456 * config/i386/sse.md (VI248_VLBW): Rename to ...
11457 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11458 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11459 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11460 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11461 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11462 mode iterator instead of VI248_VLBW.
11464 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11466 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11467 (record_modified): Skip clobbers; add debug output.
11468 (param_change_prob): Use sreal frequencies.
11470 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11472 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11473 punt for user-aligned variables.
11475 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11477 * tree-chrec.c (chrec_contains_symbols): Return true for
11480 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11483 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11484 of (x|y) == x for BICS pattern.
11486 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11488 PR tree-optimization/83605
11489 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11490 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11493 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11495 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11496 * config/epiphany/rtems.h: New file.
11498 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11499 Uros Bizjak <ubizjak@gmail.com>
11502 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11503 QIreg_operand instead of register_operand predicate.
11504 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11505 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11506 comments instead of -fmitigate[-_]rop.
11508 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11511 * cgraphunit.c (symbol_table::compile): Switch to text_section
11512 before calling assembly_start debug hook.
11513 * run-rtl-passes.c (run_rtl_passes): Likewise.
11516 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11518 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11519 range_int_cst_p rather than !symbolic_range_p before calling
11520 extract_range_from_multiplicative_op_1.
11522 2018-01-04 Jeff Law <law@redhat.com>
11524 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11525 redundant test in assertion.
11527 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11529 * doc/rtl.texi: Document machine_mode wrapper classes.
11531 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11533 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11534 using tree_to_uhwi.
11536 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11538 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11539 the VEC_PERM_EXPR fold to fail.
11541 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11544 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11545 to switched_sections.
11547 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11550 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11551 test for d.testing.
11553 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11556 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11557 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11559 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11562 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11563 is BLKmode and bitpos not zero or mode change is needed.
11565 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11568 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11571 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11574 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11575 instead of MULT rtx. Update all corresponding splitters.
11576 (*saddl_se): Ditto.
11577 (*ssub<modesuffix>): Ditto.
11578 (*ssubl_se): Ditto.
11579 (*cmp_sadd_di): Update split patterns.
11580 (*cmp_sadd_si): Ditto.
11581 (*cmp_sadd_sidi): Ditto.
11582 (*cmp_ssub_di): Ditto.
11583 (*cmp_ssub_si): Ditto.
11584 (*cmp_ssub_sidi): Ditto.
11585 * config/alpha/predicates.md (const23_operand): New predicate.
11586 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11587 Look for ASHIFT, not MULT inner operand.
11588 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11590 2018-01-04 Martin Liska <mliska@suse.cz>
11592 PR gcov-profile/83669
11593 * gcov.c (output_intermediate_file): Add version to intermediate
11595 * doc/gcov.texi: Document new field 'version' in intermediate
11596 file format. Fix location of '-k' option of gcov command.
11598 2018-01-04 Martin Liska <mliska@suse.cz>
11601 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11603 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11605 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11607 2018-01-03 Martin Sebor <msebor@redhat.com>
11609 PR tree-optimization/83655
11610 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11611 checking calls with invalid arguments.
11613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11615 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11616 (vectorizable_mask_load_store): Delete.
11617 (vectorizable_call): Return false for masked loads and stores.
11618 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11619 instead of gimple_assign_rhs1.
11620 (vectorizable_load): Handle IFN_MASK_LOAD.
11621 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11625 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11627 (vectorizable_mask_load_store): ...here.
11628 (vectorizable_load): ...and here.
11630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11632 * tree-vect-stmts.c (vect_build_all_ones_mask)
11633 (vect_build_zero_merge_argument): New functions, split out from...
11634 (vectorizable_load): ...here.
11636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11638 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11640 (vectorizable_mask_load_store): ...here.
11641 (vectorizable_store): ...and here.
11643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11645 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11647 (vectorizable_mask_load_store): ...here.
11649 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11651 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11652 (vect_model_store_cost): Take a vec_load_store_type instead of a
11654 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11655 (vect_model_store_cost): Take a vec_load_store_type instead of a
11657 (vectorizable_mask_load_store): Update accordingly.
11658 (vectorizable_store): Likewise.
11659 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11661 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11663 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11664 IFN_MASK_LOAD calls here rather than...
11665 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11668 Alan Hayward <alan.hayward@arm.com>
11669 David Sherwood <david.sherwood@arm.com>
11671 * expmed.c (extract_bit_field_1): For vector extracts,
11672 fall back to extract_bit_field_as_subreg if vec_extract
11675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11676 Alan Hayward <alan.hayward@arm.com>
11677 David Sherwood <david.sherwood@arm.com>
11679 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11680 they are variable or constant sized.
11681 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11682 slots for constant-sized data.
11684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11685 Alan Hayward <alan.hayward@arm.com>
11686 David Sherwood <david.sherwood@arm.com>
11688 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11689 handling COND_EXPRs with boolean comparisons, try to find a better
11690 basis for the mask type than the boolean itself.
11692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11694 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11695 is calculated and how it can be overridden.
11696 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11697 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11699 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11702 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11703 Alan Hayward <alan.hayward@arm.com>
11704 David Sherwood <david.sherwood@arm.com>
11706 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11707 Remove the mode argument.
11708 (aarch64_simd_valid_immediate): Remove the mode and inverse
11710 * config/aarch64/iterators.md (bitsize): New iterator.
11711 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11712 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11713 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11714 aarch64_simd_valid_immediate.
11715 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11716 (aarch64_reg_or_bic_imm): Likewise.
11717 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11718 with an insn_type enum and msl with a modifier_type enum.
11719 Replace element_width with a scalar_mode. Change the shift
11720 to unsigned int. Add constructors for scalar_float_mode and
11721 scalar_int_mode elements.
11722 (aarch64_vect_float_const_representable_p): Delete.
11723 (aarch64_can_const_movi_rtx_p)
11724 (aarch64_simd_scalar_immediate_valid_for_move)
11725 (aarch64_simd_make_constant): Update call to
11726 aarch64_simd_valid_immediate.
11727 (aarch64_advsimd_valid_immediate_hs): New function.
11728 (aarch64_advsimd_valid_immediate): Likewise.
11729 (aarch64_simd_valid_immediate): Remove mode and inverse
11730 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11731 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11732 and aarch64_float_const_representable_p on the result.
11733 (aarch64_output_simd_mov_immediate): Remove mode argument.
11734 Update call to aarch64_simd_valid_immediate and use of
11735 simd_immediate_info.
11736 (aarch64_output_scalar_simd_mov_immediate): Update call
11739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11740 Alan Hayward <alan.hayward@arm.com>
11741 David Sherwood <david.sherwood@arm.com>
11743 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11744 (mode_nunits): Likewise CONST_MODE_NUNITS.
11745 * machmode.def (ADJUST_NUNITS): Document.
11746 * genmodes.c (mode_data::need_nunits_adj): New field.
11747 (blank_mode): Update accordingly.
11748 (adj_nunits): New variable.
11749 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11751 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11752 listed in adj_nunits.
11753 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11754 listed in adj_nunits. Don't emit case statements for such modes.
11755 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11756 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11757 nothing if adj_nunits is nonnull.
11758 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11759 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11760 (emit_mode_fbit): Update use of print_maybe_const_decl.
11761 (emit_move_size): Likewise. Treat the array as non-const
11763 (emit_mode_adjustments): Handle adj_nunits.
11765 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11767 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11768 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11769 (VECTOR_MODES): Use it.
11770 (make_vector_modes): Take the prefix as an argument.
11772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11773 Alan Hayward <alan.hayward@arm.com>
11774 David Sherwood <david.sherwood@arm.com>
11776 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11777 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11778 for MODE_VECTOR_BOOL.
11779 * machmode.def (VECTOR_BOOL_MODE): Document.
11780 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11781 (make_vector_bool_mode): New function.
11782 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11784 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11785 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11787 * stor-layout.c (int_mode_for_mode): Likewise.
11788 * tree.c (build_vector_type_for_mode): Likewise.
11789 * varasm.c (output_constant_pool_2): Likewise.
11790 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11791 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11792 for MODE_VECTOR_BOOL.
11793 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11794 of mode class checks.
11795 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11796 instead of a list of mode class checks.
11797 (expand_vector_scalar_condition): Likewise.
11798 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11801 Alan Hayward <alan.hayward@arm.com>
11802 David Sherwood <david.sherwood@arm.com>
11804 * machmode.h (mode_size): Change from unsigned short to
11806 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11807 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11808 or if measurement_type is not polynomial.
11809 (fixed_size_mode::includes_p): Check for constant-sized modes.
11810 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11811 return a poly_uint16 rather than an unsigned short.
11812 (emit_mode_size): Change the type of mode_size from unsigned short
11813 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11814 (emit_mode_adjustments): Cope with polynomial vector sizes.
11815 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11817 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11819 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11820 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11821 * caller-save.c (setup_save_areas): Likewise.
11822 (replace_reg_with_saved_mem): Likewise.
11823 * calls.c (emit_library_call_value_1): Likewise.
11824 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11825 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11826 (gen_lowpart_for_combine): Likewise.
11827 * convert.c (convert_to_integer_1): Likewise.
11828 * cse.c (equiv_constant, cse_insn): Likewise.
11829 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11830 (cselib_subst_to_values): Likewise.
11831 * dce.c (word_dce_process_block): Likewise.
11832 * df-problems.c (df_word_lr_mark_ref): Likewise.
11833 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11834 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11835 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11836 (rtl_for_decl_location): Likewise.
11837 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11838 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11839 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11840 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11841 (expand_expr_real_1): Likewise.
11842 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11843 (pad_below): Likewise.
11844 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11845 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11846 * ira.c (get_subreg_tracking_sizes): Likewise.
11847 * ira-build.c (ira_create_allocno_objects): Likewise.
11848 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11849 (ira_sort_regnos_for_alter_reg): Likewise.
11850 * ira-costs.c (record_operand_costs): Likewise.
11851 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11852 (resolve_simple_move): Likewise.
11853 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11854 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11855 (lra_constraints): Likewise.
11856 (CONST_POOL_OK_P): Reject variable-sized modes.
11857 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11858 (add_pseudo_to_slot, lra_spill): Likewise.
11859 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11860 * optabs-query.c (get_best_extraction_insn): Likewise.
11861 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11862 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11863 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11864 * recog.c (offsettable_address_addr_space_p): Likewise.
11865 * regcprop.c (maybe_mode_change): Likewise.
11866 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11867 * regrename.c (build_def_use): Likewise.
11868 * regstat.c (dump_reg_info): Likewise.
11869 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11870 (find_reloads, find_reloads_subreg_address): Likewise.
11871 * reload1.c (eliminate_regs_1): Likewise.
11872 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11873 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11874 (simplify_binary_operation_1, simplify_subreg): Likewise.
11875 * targhooks.c (default_function_arg_padding): Likewise.
11876 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11877 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11878 (verify_gimple_assign_ternary): Likewise.
11879 * tree-inline.c (estimate_move_cost): Likewise.
11880 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11881 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11882 (get_address_cost_ainc): Likewise.
11883 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11884 (vect_supportable_dr_alignment): Likewise.
11885 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11886 (vectorizable_reduction): Likewise.
11887 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11888 (vectorizable_operation, vectorizable_load): Likewise.
11889 * tree.c (build_same_sized_truth_vector_type): Likewise.
11890 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11891 * var-tracking.c (emit_note_insn_var_location): Likewise.
11892 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11893 (ADDR_VEC_ALIGN): Likewise.
11895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11896 Alan Hayward <alan.hayward@arm.com>
11897 David Sherwood <david.sherwood@arm.com>
11899 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11901 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11902 or if measurement_type is polynomial.
11903 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11904 * combine.c (make_extraction): Likewise.
11905 * dse.c (find_shift_sequence): Likewise.
11906 * dwarf2out.c (mem_loc_descriptor): Likewise.
11907 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11908 (extract_bit_field, extract_low_bits): Likewise.
11909 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11910 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11911 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11912 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11913 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11914 * reload.c (find_reloads): Likewise.
11915 * reload1.c (alter_reg): Likewise.
11916 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11917 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11918 * tree-if-conv.c (predicate_mem_writes): Likewise.
11919 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11920 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11921 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11922 * valtrack.c (dead_debug_insert_temp): Likewise.
11923 * varasm.c (mergeable_constant_section): Likewise.
11924 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11927 Alan Hayward <alan.hayward@arm.com>
11928 David Sherwood <david.sherwood@arm.com>
11930 * expr.c (expand_assignment): Cope with polynomial mode sizes
11931 when assigning to a CONCAT.
11933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11934 Alan Hayward <alan.hayward@arm.com>
11935 David Sherwood <david.sherwood@arm.com>
11937 * machmode.h (mode_precision): Change from unsigned short to
11939 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11941 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11942 or if measurement_type is not polynomial.
11943 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11944 in which the mode is already known to be a scalar_int_mode.
11945 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11946 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11948 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11949 for GET_MODE_PRECISION.
11950 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11951 for GET_MODE_PRECISION.
11952 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11954 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11955 (expand_field_assignment, make_extraction): Likewise.
11956 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11957 (get_last_value): Likewise.
11958 * convert.c (convert_to_integer_1): Likewise.
11959 * cse.c (cse_insn): Likewise.
11960 * expr.c (expand_expr_real_1): Likewise.
11961 * lra-constraints.c (simplify_operand_subreg): Likewise.
11962 * optabs-query.c (can_atomic_load_p): Likewise.
11963 * optabs.c (expand_atomic_load): Likewise.
11964 (expand_atomic_store): Likewise.
11965 * ree.c (combine_reaching_defs): Likewise.
11966 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11967 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11968 * tree.h (type_has_mode_precision_p): Likewise.
11969 * ubsan.c (instrument_si_overflow): Likewise.
11971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11972 Alan Hayward <alan.hayward@arm.com>
11973 David Sherwood <david.sherwood@arm.com>
11975 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11976 polynomial numbers of units.
11977 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11978 (valid_vector_subparts_p): New function.
11979 (build_vector_type): Remove temporary shim and take the number
11980 of units as a poly_uint64 rather than an int.
11981 (build_opaque_vector_type): Take the number of units as a
11982 poly_uint64 rather than an int.
11983 * tree.c (build_vector_from_ctor): Handle polynomial
11984 TYPE_VECTOR_SUBPARTS.
11985 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11986 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11987 (build_vector_from_val): If the number of units is variable,
11988 use build_vec_duplicate_cst for constant operands and
11989 VEC_DUPLICATE_EXPR otherwise.
11990 (make_vector_type): Remove temporary is_constant ().
11991 (build_vector_type, build_opaque_vector_type): Take the number of
11992 units as a poly_uint64 rather than an int.
11993 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11995 * cfgexpand.c (expand_debug_expr): Likewise.
11996 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11997 (store_constructor, expand_expr_real_1): Likewise.
11998 (const_scalar_mask_from_tree): Likewise.
11999 * fold-const-call.c (fold_const_reduction): Likewise.
12000 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12001 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12002 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12003 (fold_relational_const): Likewise.
12004 (native_interpret_vector): Likewise. Change the size from an
12005 int to an unsigned int.
12006 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12007 TYPE_VECTOR_SUBPARTS.
12008 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12009 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12010 duplicating a non-constant operand into a variable-length vector.
12011 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12012 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12013 * ipa-icf.c (sem_variable::equals): Likewise.
12014 * match.pd: Likewise.
12015 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12016 * print-tree.c (print_node): Likewise.
12017 * stor-layout.c (layout_type): Likewise.
12018 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12019 * tree-cfg.c (verify_gimple_comparison): Likewise.
12020 (verify_gimple_assign_binary): Likewise.
12021 (verify_gimple_assign_ternary): Likewise.
12022 (verify_gimple_assign_single): Likewise.
12023 * tree-pretty-print.c (dump_generic_node): Likewise.
12024 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12025 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12026 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12027 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12028 (vect_shift_permute_load_chain): Likewise.
12029 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12030 (expand_vector_condition, optimize_vector_constructor): Likewise.
12031 (lower_vec_perm, get_compute_type): Likewise.
12032 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12033 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12034 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12035 (vect_recog_mask_conversion_pattern): Likewise.
12036 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12037 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12038 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12039 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12040 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12041 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12042 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12043 (supportable_widening_operation): Likewise.
12044 (supportable_narrowing_operation): Likewise.
12045 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12047 * varasm.c (output_constant): Likewise.
12049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12050 Alan Hayward <alan.hayward@arm.com>
12051 David Sherwood <david.sherwood@arm.com>
12053 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12054 so that both the length == 3 and length != 3 cases set up their
12055 own permute vectors. Add comments explaining why we know the
12056 number of elements is constant.
12057 (vect_permute_load_chain): Likewise.
12059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12060 Alan Hayward <alan.hayward@arm.com>
12061 David Sherwood <david.sherwood@arm.com>
12063 * machmode.h (mode_nunits): Change from unsigned char to
12065 (ONLY_FIXED_SIZE_MODES): New macro.
12066 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12067 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12068 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12070 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12071 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12072 or if measurement_type is not polynomial.
12073 * genmodes.c (ZERO_COEFFS): New macro.
12074 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12076 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12077 Use ZERO_COEFFS when emitting initializers.
12078 * data-streamer.h (bp_pack_poly_value): New function.
12079 (bp_unpack_poly_value): Likewise.
12080 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12081 for GET_MODE_NUNITS.
12082 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12083 for GET_MODE_NUNITS.
12084 * tree.c (make_vector_type): Remove temporary shim and make
12085 the real function take the number of units as a poly_uint64
12086 rather than an int.
12087 (build_vector_type_for_mode): Handle polynomial nunits.
12088 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12089 * emit-rtl.c (const_vec_series_p_1): Likewise.
12090 (gen_rtx_CONST_VECTOR): Likewise.
12091 * fold-const.c (test_vec_duplicate_folding): Likewise.
12092 * genrecog.c (validate_pattern): Likewise.
12093 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12094 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12095 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12096 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12097 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12098 * rtlanal.c (subreg_get_info): Likewise.
12099 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12100 (vect_grouped_load_supported): Likewise.
12101 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12102 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12103 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12104 (simplify_const_unary_operation, simplify_binary_operation_1)
12105 (simplify_const_binary_operation, simplify_ternary_operation)
12106 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12107 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12108 instead of CONST_VECTOR_NUNITS.
12109 * varasm.c (output_constant_pool_2): Likewise.
12110 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12111 explicit-encoded elements in the XVEC for variable-length vectors.
12113 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12115 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12117 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12118 Alan Hayward <alan.hayward@arm.com>
12119 David Sherwood <david.sherwood@arm.com>
12121 * coretypes.h (fixed_size_mode): Declare.
12122 (fixed_size_mode_pod): New typedef.
12123 * builtins.h (target_builtins::x_apply_args_mode)
12124 (target_builtins::x_apply_result_mode): Change type to
12125 fixed_size_mode_pod.
12126 * builtins.c (apply_args_size, apply_result_size, result_vector)
12127 (expand_builtin_apply_args_1, expand_builtin_apply)
12128 (expand_builtin_return): Update accordingly.
12130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12132 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12133 * cselib.c (cselib_hash_rtx): Likewise.
12134 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12135 CONST_VECTOR encoding.
12137 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12138 Jeff Law <law@redhat.com>
12141 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12142 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12143 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12144 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12147 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12148 explicitly probe *sp in a noreturn function if there were any callee
12149 register saves or frame pointer is needed.
12151 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12154 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12155 BLKmode for ternary, binary or unary expressions.
12158 * var-tracking.c (delete_vta_debug_insn): New inline function.
12159 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12160 insns from get_insns () to NULL instead of each bb separately.
12161 Use delete_vta_debug_insn. No longer static.
12162 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12163 delete_vta_debug_insns callers.
12164 * rtl.h (delete_vta_debug_insns): Declare.
12165 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12166 instead of variable_tracking_main.
12168 2018-01-03 Martin Sebor <msebor@redhat.com>
12170 PR tree-optimization/83603
12171 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12172 arguments past the endof the argument list in functions declared
12173 without a prototype.
12174 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12175 Avoid checking when arguments are null.
12177 2018-01-03 Martin Sebor <msebor@redhat.com>
12180 * doc/extend.texi (attribute const): Fix a typo.
12181 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12182 issuing -Wsuggest-attribute for void functions.
12184 2018-01-03 Martin Sebor <msebor@redhat.com>
12186 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12187 offset_int::from instead of wide_int::to_shwi.
12188 (maybe_diag_overlap): Remove assertion.
12189 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12190 * gimple-ssa-sprintf.c (format_directive): Same.
12191 (parse_directive): Same.
12192 (sprintf_dom_walker::compute_format_length): Same.
12193 (try_substitute_return_value): Same.
12195 2018-01-03 Jeff Law <law@redhat.com>
12197 PR middle-end/83654
12198 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12199 non-constant residual for zero at runtime and avoid probing in
12200 that case. Reorganize code for trailing problem to mirror handling
12203 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12205 PR tree-optimization/83501
12206 * tree-ssa-strlen.c (get_string_cst): New.
12207 (handle_char_store): Call get_string_cst.
12209 2018-01-03 Martin Liska <mliska@suse.cz>
12211 PR tree-optimization/83593
12212 * tree-ssa-strlen.c: Include tree-cfg.h.
12213 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12214 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12215 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12217 (strlen_dom_walker::before_dom_children): Call
12218 gimple_purge_dead_eh_edges. Dump tranformation with details
12220 (strlen_dom_walker::before_dom_children): Update call by adding
12221 new argument cleanup_eh.
12222 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12224 2018-01-03 Martin Liska <mliska@suse.cz>
12227 * cif-code.def (VARIADIC_THUNK): New enum value.
12228 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12231 2018-01-03 Jan Beulich <jbeulich@suse.com>
12233 * sse.md (mov<mode>_internal): Tighten condition for when to use
12234 vmovdqu<ssescalarsize> for TI and OI modes.
12236 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12238 Update copyright years.
12240 2018-01-03 Martin Liska <mliska@suse.cz>
12243 * ipa-visibility.c (function_and_variable_visibility): Skip
12244 functions with noipa attribure.
12246 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12248 * gcc.c (process_command): Update copyright notice dates.
12249 * gcov-dump.c (print_version): Ditto.
12250 * gcov.c (print_version): Ditto.
12251 * gcov-tool.c (print_version): Ditto.
12252 * gengtype.c (create_file): Ditto.
12253 * doc/cpp.texi: Bump @copying's copyright year.
12254 * doc/cppinternals.texi: Ditto.
12255 * doc/gcc.texi: Ditto.
12256 * doc/gccint.texi: Ditto.
12257 * doc/gcov.texi: Ditto.
12258 * doc/install.texi: Ditto.
12259 * doc/invoke.texi: Ditto.
12261 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12263 * vector-builder.h (vector_builder::m_full_nelts): Change from
12264 unsigned int to poly_uint64.
12265 (vector_builder::full_nelts): Update prototype accordingly.
12266 (vector_builder::new_vector): Likewise.
12267 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12268 (vector_builder::operator ==): Likewise.
12269 (vector_builder::finalize): Likewise.
12270 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12271 Take the number of elements as a poly_uint64 rather than an
12273 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12274 from unsigned int to poly_uint64.
12275 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12276 (vec_perm_indices::new_vector): Likewise.
12277 (vec_perm_indices::length): Likewise.
12278 (vec_perm_indices::nelts_per_input): Likewise.
12279 (vec_perm_indices::input_nelts): Likewise.
12280 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12281 number of elements per input as a poly_uint64 rather than an
12282 unsigned int. Use the original encoding for variable-length
12283 vectors, rather than clamping each individual element.
12284 For the second and subsequent elements in each pattern,
12285 clamp the step and base before clamping their sum.
12286 (vec_perm_indices::series_p): Handle polynomial element counts.
12287 (vec_perm_indices::all_in_range_p): Likewise.
12288 (vec_perm_indices_to_tree): Likewise.
12289 (vec_perm_indices_to_rtx): Likewise.
12290 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12291 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12292 (tree_vector_builder::new_binary_operation): Handle polynomial
12293 element counts. Return false if we need to know the number
12294 of elements at compile time.
12295 * fold-const.c (fold_vec_perm): Punt if the number of elements
12296 isn't known at compile time.
12298 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12300 * vec-perm-indices.h (vec_perm_builder): Change element type
12301 from HOST_WIDE_INT to poly_int64.
12302 (vec_perm_indices::element_type): Update accordingly.
12303 (vec_perm_indices::clamp): Handle polynomial element_types.
12304 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12305 (vec_perm_indices::all_in_range_p): Likewise.
12306 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12308 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12309 polynomial vec_perm_indices element types.
12310 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12311 * fold-const.c (fold_vec_perm): Likewise.
12312 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12313 * tree-vect-generic.c (lower_vec_perm): Likewise.
12314 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12315 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12316 element type to HOST_WIDE_INT.
12318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12319 Alan Hayward <alan.hayward@arm.com>
12320 David Sherwood <david.sherwood@arm.com>
12322 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12323 rather than an int. Use plus_constant.
12324 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12325 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12327 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12328 Alan Hayward <alan.hayward@arm.com>
12329 David Sherwood <david.sherwood@arm.com>
12331 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12332 a HOST_WIDE_INT to a poly_int64.
12334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12335 Alan Hayward <alan.hayward@arm.com>
12336 David Sherwood <david.sherwood@arm.com>
12338 * calls.c (load_register_parameters): Cope with polynomial
12339 mode sizes. Require a constant size for BLKmode parameters
12340 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12341 forces a parameter to be padded at the lsb end in order to
12342 fill a complete number of words, require the parameter size
12343 to be ordered wrt UNITS_PER_WORD.
12345 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12346 Alan Hayward <alan.hayward@arm.com>
12347 David Sherwood <david.sherwood@arm.com>
12349 * reload1.c (spill_stack_slot_width): Change element type
12350 from unsigned int to poly_uint64_pod.
12351 (alter_reg): Treat mode sizes as polynomial.
12353 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12354 Alan Hayward <alan.hayward@arm.com>
12355 David Sherwood <david.sherwood@arm.com>
12357 * reload.c (complex_word_subreg_p): New function.
12358 (reload_inner_reg_of_subreg, push_reload): Use it.
12360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12361 Alan Hayward <alan.hayward@arm.com>
12362 David Sherwood <david.sherwood@arm.com>
12364 * lra-constraints.c (process_alt_operands): Reject matched
12365 operands whose sizes aren't ordered.
12366 (match_reload): Refer to this check here.
12368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12369 Alan Hayward <alan.hayward@arm.com>
12370 David Sherwood <david.sherwood@arm.com>
12372 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12373 that the mode size is in the set {1, 2, 4, 8, 16}.
12375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12376 Alan Hayward <alan.hayward@arm.com>
12377 David Sherwood <david.sherwood@arm.com>
12379 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12380 Use plus_constant instead of gen_rtx_PLUS.
12382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12383 Alan Hayward <alan.hayward@arm.com>
12384 David Sherwood <david.sherwood@arm.com>
12386 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12387 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12388 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12389 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12390 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12391 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12392 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12393 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12394 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12395 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12397 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12398 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12399 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12400 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12401 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12402 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12403 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12404 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12405 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12406 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12408 * expr.c (emit_move_resolve_push): Treat the input and result
12409 of PUSH_ROUNDING as a poly_int64.
12410 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12411 (emit_push_insn): Likewise.
12412 * lra-eliminations.c (mark_not_eliminable): Likewise.
12413 * recog.c (push_operand): Likewise.
12414 * reload1.c (elimination_effects): Likewise.
12415 * rtlanal.c (nonzero_bits1): Likewise.
12416 * calls.c (store_one_arg): Likewise. Require the padding to be
12417 known at compile time.
12419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12420 Alan Hayward <alan.hayward@arm.com>
12421 David Sherwood <david.sherwood@arm.com>
12423 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12424 Use plus_constant instead of gen_rtx_PLUS.
12426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12427 Alan Hayward <alan.hayward@arm.com>
12428 David Sherwood <david.sherwood@arm.com>
12430 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12431 rather than an int.
12433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12434 Alan Hayward <alan.hayward@arm.com>
12435 David Sherwood <david.sherwood@arm.com>
12437 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12438 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12439 via stack temporaries. Treat the mode size as polynomial too.
12441 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12442 Alan Hayward <alan.hayward@arm.com>
12443 David Sherwood <david.sherwood@arm.com>
12445 * expr.c (expand_expr_real_2): When handling conversions involving
12446 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12447 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12448 as a poly_uint64 too.
12450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12451 Alan Hayward <alan.hayward@arm.com>
12452 David Sherwood <david.sherwood@arm.com>
12454 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12457 Alan Hayward <alan.hayward@arm.com>
12458 David Sherwood <david.sherwood@arm.com>
12460 * combine.c (can_change_dest_mode): Handle polynomial
12461 REGMODE_NATURAL_SIZE.
12462 * expmed.c (store_bit_field_1): Likewise.
12463 * expr.c (store_constructor): Likewise.
12464 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12465 and polynomial REGMODE_NATURAL_SIZE.
12466 (gen_lowpart_common): Likewise.
12467 * reginfo.c (record_subregs_of_mode): Likewise.
12468 * rtlanal.c (read_modify_subreg_p): Likewise.
12470 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12471 Alan Hayward <alan.hayward@arm.com>
12472 David Sherwood <david.sherwood@arm.com>
12474 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12475 numbers of elements.
12477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12478 Alan Hayward <alan.hayward@arm.com>
12479 David Sherwood <david.sherwood@arm.com>
12481 * match.pd: Cope with polynomial numbers of vector elements.
12483 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12484 Alan Hayward <alan.hayward@arm.com>
12485 David Sherwood <david.sherwood@arm.com>
12487 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12488 in a POINTER_PLUS_EXPR.
12490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12491 Alan Hayward <alan.hayward@arm.com>
12492 David Sherwood <david.sherwood@arm.com>
12494 * omp-simd-clone.c (simd_clone_subparts): New function.
12495 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12496 (ipa_simd_modify_function_body): Likewise.
12498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12499 Alan Hayward <alan.hayward@arm.com>
12500 David Sherwood <david.sherwood@arm.com>
12502 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12503 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12504 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12505 (expand_vector_condition, vector_element): Likewise.
12506 (subparts_gt): New function.
12507 (get_compute_type): Use subparts_gt.
12508 (count_type_subparts): Delete.
12509 (expand_vector_operations_1): Use subparts_gt instead of
12510 count_type_subparts.
12512 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12513 Alan Hayward <alan.hayward@arm.com>
12514 David Sherwood <david.sherwood@arm.com>
12516 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12517 (vect_compile_time_alias): ...this new function. Do the calculation
12518 on poly_ints rather than trees.
12519 (vect_prune_runtime_alias_test_list): Update call accordingly.
12521 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12522 Alan Hayward <alan.hayward@arm.com>
12523 David Sherwood <david.sherwood@arm.com>
12525 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12527 (vect_schedule_slp_instance): Likewise.
12529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12530 Alan Hayward <alan.hayward@arm.com>
12531 David Sherwood <david.sherwood@arm.com>
12533 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12534 constant and extern definitions for variable-length vectors.
12535 (vect_get_constant_vectors): Note that the number of units
12536 is known to be constant.
12538 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12539 Alan Hayward <alan.hayward@arm.com>
12540 David Sherwood <david.sherwood@arm.com>
12542 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12543 of units as polynomial. Choose between WIDE and NARROW based
12546 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12547 Alan Hayward <alan.hayward@arm.com>
12548 David Sherwood <david.sherwood@arm.com>
12550 * tree-vect-stmts.c (simd_clone_subparts): New function.
12551 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12554 Alan Hayward <alan.hayward@arm.com>
12555 David Sherwood <david.sherwood@arm.com>
12557 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12558 vectors as polynomial. Use build_index_vector for
12559 IFN_GOMP_SIMD_LANE.
12561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12562 Alan Hayward <alan.hayward@arm.com>
12563 David Sherwood <david.sherwood@arm.com>
12565 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12566 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12567 for variable-length vectors.
12568 (vectorizable_mask_load_store): Treat the number of units as
12569 polynomial, asserting that it is constant if the condition has
12570 already been enforced.
12571 (vectorizable_store, vectorizable_load): Likewise.
12573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12574 Alan Hayward <alan.hayward@arm.com>
12575 David Sherwood <david.sherwood@arm.com>
12577 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12578 of units as polynomial. Punt if we can't tell at compile time
12579 which vector contains the final result.
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-loop.c (vectorizable_induction): Treat the number
12586 of units as polynomial. Punt on SLP inductions. Use an integer
12587 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12588 cast of such a series for variable-length floating-point
12591 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12592 Alan Hayward <alan.hayward@arm.com>
12593 David Sherwood <david.sherwood@arm.com>
12595 * tree.h (build_index_vector): Declare.
12596 * tree.c (build_index_vector): New function.
12597 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12598 of units as polynomial, forcibly converting it to a constant if
12599 vectorizable_reduction has already enforced the condition.
12600 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12601 to create a {1,2,3,...} vector.
12602 (vectorizable_reduction): Treat the number of units as polynomial.
12603 Choose vectype_in based on the largest scalar element size rather
12604 than the smallest number of units. Enforce the restrictions
12607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12608 Alan Hayward <alan.hayward@arm.com>
12609 David Sherwood <david.sherwood@arm.com>
12611 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12612 number of units as polynomial.
12614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12615 Alan Hayward <alan.hayward@arm.com>
12616 David Sherwood <david.sherwood@arm.com>
12618 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12619 * target.def (autovectorize_vector_sizes): Return the vector sizes
12620 by pointer, using vector_sizes rather than a bitmask.
12621 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12622 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12623 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12625 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12626 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12627 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12628 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12629 * omp-general.c (omp_max_vf): Likewise.
12630 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12631 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12632 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12633 * tree-vect-slp.c (vect_slp_bb): Likewise.
12634 * doc/tm.texi: Regenerate.
12635 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12637 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12638 the vector size as a poly_uint64 rather than an unsigned int.
12639 (current_vector_size): Change from an unsigned int to a poly_uint64.
12640 (get_vectype_for_scalar_type): Update accordingly.
12641 * tree.h (build_truth_vector_type): Take the size and number of
12642 units as a poly_uint64 rather than an unsigned int.
12643 (build_vector_type): Add a temporary overload that takes
12644 the number of units as a poly_uint64 rather than an unsigned int.
12645 * tree.c (make_vector_type): Likewise.
12646 (build_truth_vector_type): Take the number of units as a poly_uint64
12647 rather than an unsigned int.
12649 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12650 Alan Hayward <alan.hayward@arm.com>
12651 David Sherwood <david.sherwood@arm.com>
12653 * target.def (get_mask_mode): Take the number of units and length
12654 as poly_uint64s rather than unsigned ints.
12655 * targhooks.h (default_get_mask_mode): Update accordingly.
12656 * targhooks.c (default_get_mask_mode): Likewise.
12657 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12658 * doc/tm.texi: Regenerate.
12660 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12661 Alan Hayward <alan.hayward@arm.com>
12662 David Sherwood <david.sherwood@arm.com>
12664 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12665 * omp-general.c (omp_max_vf): Likewise.
12666 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12667 (expand_omp_simd): Handle polynomial safelen.
12668 * omp-low.c (omplow_simd_context): Add a default constructor.
12669 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12670 (lower_rec_simd_input_clauses): Update accordingly.
12671 (lower_rec_input_clauses): Likewise.
12673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12674 Alan Hayward <alan.hayward@arm.com>
12675 David Sherwood <david.sherwood@arm.com>
12677 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12678 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12679 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12680 (vect_analyze_slp_cost): Likewise.
12681 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12682 (vect_model_load_cost): Likewise.
12684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12685 Alan Hayward <alan.hayward@arm.com>
12686 David Sherwood <david.sherwood@arm.com>
12688 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12689 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12690 from an unsigned int * to a poly_uint64_pod *.
12691 (calculate_unrolling_factor): New function.
12692 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12694 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12695 Alan Hayward <alan.hayward@arm.com>
12696 David Sherwood <david.sherwood@arm.com>
12698 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12699 from an unsigned int to a poly_uint64.
12700 (_loop_vec_info::slp_unrolling_factor): Likewise.
12701 (_loop_vec_info::vectorization_factor): Change from an int
12703 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12704 (vect_get_num_vectors): New function.
12705 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12706 (vect_get_num_copies): Use vect_get_num_vectors.
12707 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12708 to an unsigned int *.
12709 (vect_analyze_data_refs): Change min_vf from an int * to a
12711 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12712 than an unsigned HOST_WIDE_INT.
12713 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12714 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12715 to an unsigned int *.
12716 (vect_analyze_data_ref_dependences): Likewise.
12717 (vect_compute_data_ref_alignment): Handle polynomial vf.
12718 (vect_enhance_data_refs_alignment): Likewise.
12719 (vect_prune_runtime_alias_test_list): Likewise.
12720 (vect_shift_permute_load_chain): Likewise.
12721 (vect_supportable_dr_alignment): Likewise.
12722 (dependence_distance_ge_vf): Take the vectorization factor as a
12723 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12724 (vect_analyze_data_refs): Change min_vf from an int * to a
12726 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12727 vfm1 as a poly_uint64 rather than an int. Make the same change
12728 for the returned bound_scalar.
12729 (vect_gen_vector_loop_niters): Handle polynomial vf.
12730 (vect_do_peeling): Likewise. Update call to
12731 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12732 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12734 * tree-vect-loop.c (vect_determine_vectorization_factor)
12735 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12736 (vect_get_known_peeling_cost): Likewise.
12737 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12738 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12739 (vect_transform_loop): Likewise. Use the lowest possible VF when
12740 updating the upper bounds of the loop.
12741 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12742 rather than an int.
12743 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12744 polynomial unroll factors.
12745 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12746 (vect_make_slp_decision): Likewise.
12747 (vect_supported_load_permutation_p): Likewise, and polynomial
12749 (vect_analyze_slp_cost): Handle polynomial vf.
12750 (vect_slp_analyze_node_operations): Likewise.
12751 (vect_slp_analyze_bb_1): Likewise.
12752 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12753 than an unsigned HOST_WIDE_INT.
12754 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12755 (vectorizable_load): Handle polynomial vf.
12756 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12758 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12761 Alan Hayward <alan.hayward@arm.com>
12762 David Sherwood <david.sherwood@arm.com>
12764 * match.pd: Handle bit operations involving three constants
12765 and try to fold one pair.
12767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12769 * tree-vect-loop-manip.c: Include gimple-fold.h.
12770 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12771 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12772 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12773 Add a path that uses a step of VF instead of 1, but disable it
12775 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12776 and niters_no_overflow parameters. Update calls to
12777 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12778 Create a new SSA name if the latter choses to use a ste other
12779 than zero, and return it via niters_vector_mult_vf_var.
12780 * tree-vect-loop.c (vect_transform_loop): Update calls to
12781 vect_do_peeling, vect_gen_vector_loop_niters and
12782 slpeel_make_loop_iterate_ntimes.
12783 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12784 (vect_gen_vector_loop_niters): Update declarations after above changes.
12786 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12788 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12789 128-bit round to integer instructions.
12790 (ceil<mode>2): Likewise.
12791 (btrunc<mode>2): Likewise.
12792 (round<mode>2): Likewise.
12794 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12796 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12797 unaligned VSX load/store on P8/P9.
12798 (expand_block_clear): Allow the use of unaligned VSX
12799 load/store on P8/P9.
12801 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12803 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12805 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12806 swap associated with both a load and a store.
12808 2018-01-02 Andrew Waterman <andrew@sifive.com>
12810 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12811 * config/riscv/riscv.md (clear_cache): Use it.
12813 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12815 * web.c: Remove out-of-date comment.
12817 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12819 * expr.c (fixup_args_size_notes): Check that any existing
12820 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12821 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12822 (emit_single_push_insn): ...here.
12824 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12826 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12827 (const_vector_encoded_nelts): New function.
12828 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12829 (const_vector_int_elt, const_vector_elt): Declare.
12830 * emit-rtl.c (const_vector_int_elt_1): New function.
12831 (const_vector_elt): Likewise.
12832 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12833 of CONST_VECTOR_ELT.
12835 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12837 * expr.c: Include rtx-vector-builder.h.
12838 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12839 directly on the tree encoding.
12840 (const_vector_from_tree): Likewise.
12841 * optabs.c: Include rtx-vector-builder.h.
12842 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12843 sequence of "u" values.
12844 * vec-perm-indices.c: Include rtx-vector-builder.h.
12845 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12846 directly on the vec_perm_indices encoding.
12848 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12850 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12851 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12852 * rtx-vector-builder.h: New file.
12853 * rtx-vector-builder.c: Likewise.
12854 * rtl.h (rtx_def::u2): Add a const_vector field.
12855 (CONST_VECTOR_NPATTERNS): New macro.
12856 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12857 (CONST_VECTOR_DUPLICATE_P): Likewise.
12858 (CONST_VECTOR_STEPPED_P): Likewise.
12859 (CONST_VECTOR_ENCODED_ELT): Likewise.
12860 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12861 (unwrap_const_vec_duplicate): Likewise.
12862 (const_vec_series_p): Check for a non-duplicated vector encoding.
12863 Say that the function only returns true for integer vectors.
12864 * emit-rtl.c: Include rtx-vector-builder.h.
12865 (gen_const_vec_duplicate_1): Delete.
12866 (gen_const_vector): Call gen_const_vec_duplicate instead of
12867 gen_const_vec_duplicate_1.
12868 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12869 (gen_const_vec_duplicate): Use rtx_vector_builder.
12870 (gen_const_vec_series): Likewise.
12871 (gen_rtx_CONST_VECTOR): Likewise.
12872 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12873 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12874 Build a new vector rather than modifying a CONST_VECTOR in-place.
12875 (handle_special_swappables): Update call accordingly.
12876 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12877 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12878 Build a new vector rather than modifying a CONST_VECTOR in-place.
12879 (handle_special_swappables): Update call accordingly.
12881 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12883 * simplify-rtx.c (simplify_const_binary_operation): Use
12884 CONST_VECTOR_ELT instead of XVECEXP.
12886 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12888 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12889 the selector elements to be different from the data elements
12890 if the selector is a VECTOR_CST.
12891 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12892 ssizetype for the selector.
12894 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12896 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12897 before testing each element individually.
12898 * tree-vect-generic.c (lower_vec_perm): Likewise.
12900 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12902 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12903 * selftest-run-tests.c (selftest::run_tests): Call it.
12904 * vector-builder.h (vector_builder::operator ==): New function.
12905 (vector_builder::operator !=): Likewise.
12906 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12907 (vec_perm_indices::all_from_input_p): New function.
12908 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12909 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12910 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12911 instead of reading the VECTOR_CST directly. Detect whether both
12912 vector inputs are the same before constructing the vec_perm_indices,
12913 and update the number of inputs argument accordingly. Use the
12914 utility functions added above. Only construct sel2 if we need to.
12916 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12918 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12919 the broadcast of the low byte.
12920 (expand_mult_highpart): Use an explicit encoding for the permutes.
12921 * optabs-query.c (can_mult_highpart_p): Likewise.
12922 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12923 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12924 (vectorizable_bswap): Likewise.
12925 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12926 explicit encoding for the power-of-2 permutes.
12927 (vect_permute_store_chain): Likewise.
12928 (vect_grouped_load_supported): Likewise.
12929 (vect_permute_load_chain): Likewise.
12931 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12933 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12934 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12935 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12936 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12937 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12938 (vect_gen_perm_mask_any): Likewise.
12940 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12942 * int-vector-builder.h: New file.
12943 * vec-perm-indices.h: Include int-vector-builder.h.
12944 (vec_perm_indices): Redefine as an int_vector_builder.
12945 (auto_vec_perm_indices): Delete.
12946 (vec_perm_builder): Redefine as a stand-alone class.
12947 (vec_perm_indices::vec_perm_indices): New function.
12948 (vec_perm_indices::clamp): Likewise.
12949 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12950 (vec_perm_indices::new_vector): New function.
12951 (vec_perm_indices::new_expanded_vector): Update for new
12952 vec_perm_indices class.
12953 (vec_perm_indices::rotate_inputs): New function.
12954 (vec_perm_indices::all_in_range_p): Operate directly on the
12955 encoded form, without computing elided elements.
12956 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12957 encoding. Update for new vec_perm_indices class.
12958 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12959 the given vec_perm_builder.
12960 (expand_vec_perm_var): Update vec_perm_builder constructor.
12961 (expand_mult_highpart): Use vec_perm_builder instead of
12962 auto_vec_perm_indices.
12963 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12964 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12965 or double series encoding as appropriate.
12966 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12967 vec_perm_indices instead of auto_vec_perm_indices.
12968 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12969 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12970 (vect_permute_store_chain): Likewise.
12971 (vect_grouped_load_supported): Likewise.
12972 (vect_permute_load_chain): Likewise.
12973 (vect_shift_permute_load_chain): Likewise.
12974 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12975 (vect_transform_slp_perm_load): Likewise.
12976 (vect_schedule_slp_instance): Likewise.
12977 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12978 (vectorizable_mask_load_store): Likewise.
12979 (vectorizable_bswap): Likewise.
12980 (vectorizable_store): Likewise.
12981 (vectorizable_load): Likewise.
12982 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12983 vec_perm_indices instead of auto_vec_perm_indices. Use
12984 tree_to_vec_perm_builder to read the vector from a tree.
12985 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12986 vec_perm_builder instead of a vec_perm_indices.
12987 (have_whole_vector_shift): Use vec_perm_builder and
12988 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12989 truncation to calc_vec_perm_mask_for_shift.
12990 (vect_create_epilog_for_reduction): Likewise.
12991 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12992 from auto_vec_perm_indices to vec_perm_indices.
12993 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12994 instead of changing individual elements.
12995 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12996 the vector in d.perm.
12997 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12998 from auto_vec_perm_indices to vec_perm_indices.
12999 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13000 instead of changing individual elements.
13001 (arm_vectorize_vec_perm_const): Use new_vector to install
13002 the vector in d.perm.
13003 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13004 Update vec_perm_builder constructor.
13005 (rs6000_expand_interleave): Likewise.
13006 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13007 (rs6000_expand_interleave): Likewise.
13009 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13011 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13012 to qimode could truncate the indices.
13013 * optabs.c (expand_vec_perm_var): Likewise.
13015 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13017 * Makefile.in (OBJS): Add vec-perm-indices.o.
13018 * vec-perm-indices.h: New file.
13019 * vec-perm-indices.c: Likewise.
13020 * target.h (vec_perm_indices): Replace with a forward class
13022 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13023 * optabs.h: Include vec-perm-indices.h.
13024 (expand_vec_perm): Delete.
13025 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13026 (expand_vec_perm_const): Declare.
13027 * target.def (vec_perm_const_ok): Replace with...
13028 (vec_perm_const): ...this new hook.
13029 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13030 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13031 * doc/tm.texi: Regenerate.
13032 * optabs.def (vec_perm_const): Delete.
13033 * doc/md.texi (vec_perm_const): Likewise.
13034 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13035 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13036 expand_vec_perm for constant permutation vectors. Assert that
13037 the mode of variable permutation vectors is the integer equivalent
13038 of the mode that is being permuted.
13039 * optabs-query.h (selector_fits_mode_p): Declare.
13040 * optabs-query.c: Include vec-perm-indices.h.
13041 (selector_fits_mode_p): New function.
13042 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13043 is defined, instead of checking whether the vec_perm_const_optab
13044 exists. Use targetm.vectorize.vec_perm_const instead of
13045 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13046 fit in the vector mode before using a variable permute.
13047 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13048 vec_perm_indices instead of an rtx.
13049 (expand_vec_perm): Replace with...
13050 (expand_vec_perm_const): ...this new function. Take the selector
13051 as a vec_perm_indices rather than an rtx. Also take the mode of
13052 the selector. Update call to shift_amt_for_vec_perm_mask.
13053 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13054 Use vec_perm_indices::new_expanded_vector to expand the original
13055 selector into bytes. Check whether the indices fit in the vector
13056 mode before using a variable permute.
13057 (expand_vec_perm_var): Make global.
13058 (expand_mult_highpart): Use expand_vec_perm_const.
13059 * fold-const.c: Includes vec-perm-indices.h.
13060 * tree-ssa-forwprop.c: Likewise.
13061 * tree-vect-data-refs.c: Likewise.
13062 * tree-vect-generic.c: Likewise.
13063 * tree-vect-loop.c: Likewise.
13064 * tree-vect-slp.c: Likewise.
13065 * tree-vect-stmts.c: Likewise.
13066 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13068 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13069 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13070 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13071 (aarch64_vectorize_vec_perm_const): ...this new function.
13072 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13073 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13074 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13075 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13076 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13077 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13078 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13080 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13081 check for NEON modes.
13082 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13083 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13084 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13085 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13087 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13088 the old VEC_PERM_CONST conditions.
13089 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13090 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13091 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13092 (ia64_vectorize_vec_perm_const_ok): Merge into...
13093 (ia64_vectorize_vec_perm_const): ...this new function.
13094 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13095 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13096 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13097 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13098 * config/mips/mips.c (mips_expand_vec_perm_const)
13099 (mips_vectorize_vec_perm_const_ok): Merge into...
13100 (mips_vectorize_vec_perm_const): ...this new function.
13101 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13102 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13103 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13104 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13105 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13106 (rs6000_expand_vec_perm_const): Delete.
13107 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13109 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13110 (altivec_expand_vec_perm_const_le): Take each operand individually.
13111 Operate on constant selectors rather than rtxes.
13112 (altivec_expand_vec_perm_const): Likewise. Update call to
13113 altivec_expand_vec_perm_const_le.
13114 (rs6000_expand_vec_perm_const): Delete.
13115 (rs6000_vectorize_vec_perm_const_ok): Delete.
13116 (rs6000_vectorize_vec_perm_const): New function.
13117 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13118 an element count and rtx array.
13119 (rs6000_expand_extract_even): Update call accordingly.
13120 (rs6000_expand_interleave): Likewise.
13121 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13122 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13123 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13124 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13125 (rs6000_expand_vec_perm_const): Delete.
13126 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13127 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13128 (altivec_expand_vec_perm_const_le): Take each operand individually.
13129 Operate on constant selectors rather than rtxes.
13130 (altivec_expand_vec_perm_const): Likewise. Update call to
13131 altivec_expand_vec_perm_const_le.
13132 (rs6000_expand_vec_perm_const): Delete.
13133 (rs6000_vectorize_vec_perm_const_ok): Delete.
13134 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13135 reference to the SPE evmerge intructions.
13136 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13137 an element count and rtx array.
13138 (rs6000_expand_extract_even): Update call accordingly.
13139 (rs6000_expand_interleave): Likewise.
13140 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13141 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13143 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13145 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13147 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13148 vector mode and that that mode matches the mode of the data
13150 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13151 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13152 directly using expand_vec_perm_1 when forcing selectors into
13154 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13156 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13158 * optabs-query.h (can_vec_perm_p): Delete.
13159 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13160 * optabs-query.c (can_vec_perm_p): Split into...
13161 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13162 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13163 particular selector is valid.
13164 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13165 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13166 (vect_grouped_load_supported): Likewise.
13167 (vect_shift_permute_load_chain): Likewise.
13168 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13169 (vect_transform_slp_perm_load): Likewise.
13170 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13171 (vectorizable_bswap): Likewise.
13172 (vect_gen_perm_mask_checked): Likewise.
13173 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13174 implementations of variable permutation vectors into account
13175 when deciding which selector to use.
13176 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13177 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13178 with a false third argument.
13179 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13180 to test whether the constant selector is valid and can_vec_perm_var_p
13181 to test whether a variable selector is valid.
13183 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13185 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13186 * optabs-query.c (can_vec_perm_p): Likewise.
13187 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13188 instead of vec_perm_indices.
13189 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13190 (vect_gen_perm_mask_checked): Likewise,
13191 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13192 (vect_gen_perm_mask_checked): Likewise,
13194 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13196 * optabs-query.h (qimode_for_vec_perm): Declare.
13197 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13198 (qimode_for_vec_perm): ...this new function.
13199 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13201 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13203 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13204 does not have a conditional at the top.
13206 2018-01-02 Richard Biener <rguenther@suse.de>
13208 * ipa-inline.c (big_speedup_p): Fix expression.
13210 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13213 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13216 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13220 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13221 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13222 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13223 cond_taken_branch_cost 3->4.
13225 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13227 PR tree-optimization/83581
13228 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13229 TODO_cleanup_cfg if any changes have been made.
13231 PR middle-end/83608
13232 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13233 convert_modes if target mode has the right side, but different mode
13236 PR middle-end/83609
13237 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13238 last argument when extracting from CONCAT. If either from_real or
13239 from_imag is NULL, use expansion through memory. If result is not
13240 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13241 the parts directly to inner mode, if even that fails, use expansion
13244 PR middle-end/83623
13245 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13246 check for bswap in mode rather than HImode and use that in expand_unop
13249 Copyright (C) 2018 Free Software Foundation, Inc.
13251 Copying and distribution of this file, with or without modification,
13252 are permitted in any medium without royalty provided the copyright
13253 notice and this notice are preserved.