1 2018-05-18 Richard Biener <rguenther@suse.de>
3 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
5 (evrp_dom_walker::before_dom_children): Call it.
7 2018-05-18 Richard Biener <rguenther@suse.de>
9 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
10 results when processing array refs with variable index.
12 2018-05-18 Toon Moene <toon@moene.org>
14 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
15 directly after that of -floop-interchange. Indicate that both
16 options are enabled by default when specifying -O3.
18 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
21 iterator. Delete separate integer-mode vec_set<mode> expander.
22 (aarch64_simd_vec_setv2di): Delete.
23 (vec_setv2di): Delete.
24 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
25 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
26 the "w, r" alternative.
28 2018-05-18 Martin Liska <mliska@suse.cz>
30 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
31 * tree-pass.h (make_pass_lower_switch_O0): New function.
32 * tree-switch-conversion.c (node_has_low_bound): Remove.
33 (node_has_high_bound): Likewise.
34 (node_is_bounded): Likewise.
35 (class pass_lower_switch): Make it a template type and create
37 (pass_lower_switch::execute): Add template argument.
38 (make_pass_lower_switch): New function.
39 (make_pass_lower_switch_O0): New function.
40 (do_jump_if_equal): Remove.
41 (emit_case_nodes): Simplify to just handle all 3 cases and leave
42 all the hard work to tree optimization passes.
44 2018-05-18 Martin Liska <mliska@suse.cz>
46 * dbgcnt.c (limit_low): Renamed from limit.
47 (limit_high): New variable.
48 (dbg_cnt_is_enabled): Check for upper limit.
49 (dbg_cnt): Adjust dumping.
50 (dbg_cnt_set_limit_by_index): Add new argument for high
52 (dbg_cnt_set_limit_by_name): Likewise.
53 (dbg_cnt_process_single_pair): Parse new format.
54 (dbg_cnt_process_opt): Use strtok.
55 (dbg_cnt_list_all_counters): Remove 'value' and add
57 * doc/invoke.texi: Document changes.
59 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
61 * doc/sourcebuild.texi (scalar_all_fma): Document.
62 * tree.def (FMA_EXPR): Delete.
63 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
64 * internal-fn.c (ternary_direct): New macro.
65 (expand_ternary_optab_fn): Likewise.
66 (direct_ternary_optab_supported_p): Likewise.
67 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
68 * builtins.c (fold_builtin_fma): Delete.
69 (fold_builtin_3): Don't call it.
70 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
71 * expr.c (expand_expr_real_2): Likewise.
72 * fold-const.c (operand_equal_p): Likewise.
73 (fold_ternary_loc): Likewise.
74 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
75 * gimple.c (DEFTREECODE): Likewise.
76 * gimplify.c (gimplify_expr): Likewise.
77 * optabs-tree.c (optab_for_tree_code): Likewise.
78 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
79 * tree-eh.c (operation_could_trap_p): Likewise.
80 (stmt_could_throw_1_p): Likewise.
81 * tree-inline.c (estimate_operator_cost): Likewise.
82 * tree-pretty-print.c (dump_generic_node): Likewise.
83 (op_code_prio): Likewise.
84 * tree-ssa-loop-im.c (stmt_cost): Likewise.
85 * tree-ssa-operands.c (get_expr_operands): Likewise.
86 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
87 * fold-const-call.h (fold_fma): Delete.
88 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
89 CFN_FNMA and CFN_FNMS.
91 * genmatch.c (combined_fn): New enum.
92 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
93 (commutative_op): New function.
94 (commutate): Use it. Handle more than 2 operands.
95 (dt_operand::gen_gimple_expr): Use commutative_op.
96 (parser::parse_expr): Allow :c to be used with non-binary
97 operators if the commutative operand is known.
98 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
99 CFN_FMS, CFN_FNMA and CFN_FNMS.
100 (backprop::process_assign_use): Remove FMA_EXPR handling.
101 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
102 (gen_hsa_fma): New function.
103 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
104 IFN_FNMA and IFN_FNMS.
105 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
106 * gimple-fold.h (follow_all_ssa_edges): Declare.
107 * gimple-fold.c (follow_all_ssa_edges): New function.
108 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
109 gimple_build interface and use follow_all_ssa_edges to fold the result.
110 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
111 instead of checking for optabs directly.
112 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
113 rather than FMA_EXPRs.
114 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
115 call to IFN_FMA instead of an FMA_EXPR.
117 2018-05-17 Jim Wilson <jimw@sifive.com>
119 * expr.c (do_tablejump): When converting index to Pmode, if we have a
120 sign extended promoted subreg, and the range does not have the sign bit
121 set, then do a sign extend.
123 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
124 test, check for sign extended subreg and/or constant operands, and
125 do a sign extend in that case.
127 2018-05-17 Steve Ellcey <sellcey@cavium.com>
129 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
130 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
132 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
133 Change logics_shift_reg to logics_shift_imm.
134 (thunderx2t99_fp_loadpair_basic): Delete.
135 (thunderx2t99_fp_storepair_basic): Delete.
136 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
137 (thunderx2t99_asimd_polynomial): Delete.
138 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
139 and neon_fp_mul_d_scalar_q.
140 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
141 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
142 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
143 (thunderx2t99_asimd_lut): Add missing tbl types.
144 (thunderx2t99_asimd_ext): Delete.
145 (thunderx2t99_asimd_load1_1_mult): Delete.
146 (thunderx2t99_asimd_load1_2_mult): Delete.
147 (thunderx2t99_asimd_load1_ldp): New.
148 (thunderx2t99_asimd_load1): New.
149 (thunderx2t99_asimd_load2): Add missing *load2* types.
150 (thunderx2t99_asimd_load3): New.
151 (thunderx2t99_asimd_load4): New.
152 (thunderx2t99_asimd_store1_1_mult): Delete.
153 (thunderx2t99_asimd_store1_2_mult): Delete.
154 (thunderx2t99_asimd_store2_mult): Delete.
155 (thunderx2t99_asimd_store2_onelane): Delete.
156 (thunderx2t99_asimd_store_stp): New.
157 (thunderx2t99_asimd_store1): New.
158 (thunderx2t99_asimd_store2): New.
159 (thunderx2t99_asimd_store3): New.
160 (thunderx2t99_asimd_store4): New.
162 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
164 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
165 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
167 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
168 Segher Boessenkool <segher@kernel.crashing.org>
171 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
174 2018-05-17 Richard Biener <rguenther@suse.de>
176 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
177 for pruning loop and prune defs feeding only already visited PHIs.
179 2018-05-17 Richard Biener <rguenther@suse.de>
181 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
183 2018-05-17 Bin Cheng <bin.cheng@arm.com>
184 Richard Biener <rguenther@suse.de>
186 PR tree-optimization/85793
187 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
188 for VMAT_ELEMENTWISE.
190 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
192 * internal-fn.h (lookup_internal_fn): Declare
193 * internal-fn.c (lookup_internal_fn): New function.
194 * gimple.c (gimple_build_call_from_tree): Handle calls to
196 * gimple-pretty-print.c (dump_gimple_call): Print "." before
197 internal function names.
198 * tree-pretty-print.c (dump_generic_node): Likewise.
199 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
201 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
203 * gimple-fold.h (gimple_build): Make the function forms take
204 combined_fn rather than built_in_function.
205 (gimple_simplify): Likewise.
206 * gimple-match-head.c (gimple_simplify): Likewise.
207 * gimple-fold.c (gimple_build): Likewise.
208 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
209 rather than gimple_build_call_internal.
210 (get_initial_defs_for_reduction): Likewise.
211 (vect_create_epilog_for_reduction): Likewise.
212 (vectorizable_live_operation): Likewise.
214 2018-05-17 Martin Liska <mliska@suse.cz>
216 * gimple-ssa-sprintf.c (format_directive): Do not use
217 space in between 'G_' and '('.
219 2018-05-17 Jakub Jelinek <jakub@redhat.com>
222 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
223 even if the mask is not all ones.
226 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
228 (ix86_gimple_fold_builtin): Likewise.
231 * config/i386/i386.c: Include tree-vector-builder.h.
232 (ix86_vector_shift_count): New function.
233 (ix86_fold_builtin): Fold shift builtins by scalar count.
234 (ix86_gimple_fold_builtin): Likewise.
236 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
237 _mm512_setzero): New intrinsics.
239 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
240 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
242 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
243 code generation for cases where splatting a value is not useful.
244 * simplify-rtx.c (simplify_ternary_operation): Simplify
245 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
246 mode to a vec_concat.
248 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
250 * config.gcc: Support "goldmont-plus".
251 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
253 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
254 PROCESSOR_GOLDMONT_PLUS.
255 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
256 (processor_target_table): Add "goldmont-plus".
257 (PTA_GOLDMONT_PLUS): Define.
258 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
259 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
260 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
261 (fold_builtin_cpu): Add "goldmont-plus".
262 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
263 (ix86_option_override_internal): Add "goldmont-plus".
264 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
265 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
266 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
267 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
269 2018-05-17 Richard Biener <rguenther@suse.de>
271 PR tree-optimization/85757
272 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
273 remove defs that only feed that PHI from further processing.
275 2018-05-16 Jim Wilson <jimw@sifive.com>
277 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
279 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
281 2018-05-16 Mark Wielaard <mark@klomp.org>
283 * dwarf2out.c (count_index_strings): New function.
284 (output_indirect_strings): Call count_index_strings and generate
285 header for dwarf_version >= 5.
287 2018-05-16 Mark Wielaard <mark@klomp.org>
289 * dwarf2out.c (dwarf_FORM): New function.
290 (set_indirect_string): Use dwarf_FORM.
291 (reset_indirect_string): Likewise.
292 (size_of_die): Likewise.
293 (value_format): Likewise.
294 (output_die): Likewise.
295 (add_skeleton_AT_string): Likewise.
296 (output_macinfo_op): Likewise.
297 (index_string): Likewise.
298 (output_index_string_offset): Likewise.
299 (output_index_string): Likewise.
300 (count_index_strings): Likewise.
302 2018-05-16 Carl Love <cel@us.ibm.com>
304 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
305 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
307 2018-05-16 Martin Jambor <mjambor@suse.cz>
309 * ipa-prop.c (ipa_free_all_edge_args): Remove.
310 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
312 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
314 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
315 (fnma<mode>4): Likewise.
316 (fms<mode>4): Likewise.
317 (fnms<mode>4): Likewise.
318 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
319 (aarch64_fnma<mode>4): Likewise.
320 (aarch64_fms<mode>4): Likewise.
321 (aarch64_fnms<mode>4): Likewise.
322 (aarch64_fnmadd<mode>4): Likewise.
324 2018-05-16 Jason Merrill <jason@redhat.com>
326 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
328 2018-05-16 Richard Biener <rguenther@suse.de>
330 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
331 (dump_stmt_cost): Declare.
332 (add_stmt_cost): Dump cost we add.
333 (add_stmt_costs): New function.
334 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
336 (vect_analyze_stmt): Adjust prototype.
337 (vectorizable_condition): Likewise.
338 (vectorizable_live_operation): Likewise.
339 (vectorizable_reduction): Likewise.
340 (vectorizable_induction): Likewise.
341 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
342 cost vector to pass to vectorizable_ and record afterwards.
343 (vect_model_reduction_cost): Take cost vector argument and adjust.
344 (vect_model_induction_cost): Likewise.
345 (vectorizable_reduction): Likewise.
346 (vectorizable_induction): Likewise.
347 (vectorizable_live_operation): Likewise.
348 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
349 SLP_TREE_NUMBER_OF_VEC_STMTS.
350 (vect_analyze_slp_cost_1): Remove.
351 (vect_analyze_slp_cost): Likewise.
352 (vect_slp_analyze_node_operations): Take visited args and
353 a target cost vector. Avoid processing already visited stmt sets.
354 (vect_slp_analyze_operations): Use a local cost vector to gather
355 costs and register those of non-discarded instances.
356 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
357 (vect_schedule_slp_instance): Remove copying of
358 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
360 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
361 adding cost. Record cost entry location.
362 (vect_prologue_cost_for_slp_op): Function to compute cost of
363 a constant or invariant generated for SLP vect in the prologue,
364 split out from vect_analyze_slp_cost_1.
365 (vect_model_simple_cost): Make static. Adjust for SLP costing.
366 (vect_model_promotion_demotion_cost): Likewise.
367 (vect_model_store_cost): Likewise, make static.
368 (vect_model_load_cost): Likewise.
369 (vectorizable_bswap): Add cost vector arg and adjust.
370 (vectorizable_call): Likewise.
371 (vectorizable_simd_clone_call): Likewise.
372 (vectorizable_conversion): Likewise.
373 (vectorizable_assignment): Likewise.
374 (vectorizable_shift): Likewise.
375 (vectorizable_operation): Likewise.
376 (vectorizable_store): Likewise.
377 (vectorizable_load): Likewise.
378 (vectorizable_condition): Likewise.
379 (vectorizable_comparison): Likewise.
380 (can_vectorize_live_stmts): Likewise.
381 (vect_analyze_stmt): Likewise.
382 (vect_transform_stmt): Adjust calls to vectorizable_*.
383 * tree-vectorizer.c: Include gimple-pretty-print.h.
384 (dump_stmt_cost): New function.
386 2018-05-16 Richard Biener <rguenther@suse.de>
388 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
389 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
390 * tree-ssa-dse.c: Include tree-ssa-loop.h.
391 (check_name): New callback.
392 (dse_classify_store): Track cycles via a visited bitmap of PHI
393 defs and simplify handling of in-loop and across loop dead stores
394 and properly fail for loop-variant refs. Handle byte-tracking with
395 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
398 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
400 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
401 (vect_get_mask_type_for_stmt): Likewise.
402 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
404 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
405 to determine the statement's vector type and the vector type that
406 should be used for calculating nunits. Deal with cases in which
407 the type has to be deferred.
408 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
409 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
410 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
411 (vect_determine_vf_for_stmt): New functions, split out from...
412 (vect_determine_vectorization_factor): ...here.
413 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
414 (vect_get_mask_type_for_stmt): New functions, split out from
415 vect_determine_vectorization_factor.
417 2018-05-16 Richard Biener <rguenther@suse.de>
419 * tree-cfg.c (verify_gimple_assign_ternary): Properly
420 verify the [VEC_]COND_EXPR embedded comparison.
422 2018-05-15 Martin Sebor <msebor@redhat.com>
424 PR tree-optimization/85753
425 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
426 RECORD_TYPE in addition to ARRAY_TYPE.
428 2018-05-15 Martin Sebor <msebor@redhat.com>
431 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
433 2018-05-15 Richard Biener <rguenther@suse.de>
435 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
436 add by_clobber_p one. Change algorithm to collect all defs
437 representing uses we need to walk and try reducing them to
438 a single one before failing.
439 (dse_dom_walker::dse_optimize_stmt): Adjust.
441 2018-05-13 Mark Wielaard <mark@klomp.org>
443 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
444 (size_of_loc_descr): Likewise.
445 (output_loc_operands): Likewise.
446 (output_loc_operands_raw): Likewise.
447 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
448 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
449 (hash_loc_operands): Likewise.
450 (compare_loc_operands): Likewise.
452 2018-05-14 Mark Wielaard <mark@klomp.org>
454 * dwarf2out.c (count_index_addrs): New function.
455 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
457 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
459 PR tree-optimization/83648
460 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
461 return value as malloc candidate.
463 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
466 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
467 as true in call to suggest_attribute.
469 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
471 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
474 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
476 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
478 (aarch64_crypto_aesd_fused): Likewise.
480 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
482 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
483 (movsi_aarch64): Likewise.
484 (load_pairsi): Likewise.
485 (load_pairdi): Likewise.
486 (store_pairsi): Likewise.
487 (store_pairdi): Likewise.
488 (load_pairsf): Likewise.
489 (load_pairdf): Likewise.
490 (store_pairsf): Likewise.
491 (store_pairdf): Likewise.
492 (zero_extend): Likewise.
493 (trunc): Swap alternatives.
494 (fcvt_target): Add '?' to prefer w over r.
496 2018-05-14 Jakub Jelinek <jakub@redhat.com>
499 * config/i386/i386.md: Disallow non-commutative arithmetics in
500 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
501 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
502 in the peephole2 before it.
504 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
506 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
507 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
508 (ix86_handle_option): Handle -mcldemote.
509 * config.gcc: New header.
510 * config/i386/cldemoteintrin.h: New file.
511 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
512 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
514 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
515 OPTION_MASK_ISA_CLDEMOTE.
516 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
517 (ix86_valid_target_attribute_inner_p): Ditto.
518 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
519 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
520 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
521 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
522 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
524 * config/i386/i386.opt: Add -mcldemote.
525 * config/i386/x86intrin.h: New header.
526 * doc/invoke.texi: Add -mcldemote.
528 2018-05-14 Richard Biener <rguenther@suse.de>
530 * doc/match-and-simplify.texi: Adjust :s documentation.
532 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
534 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
535 intended memcpy size.
536 (REORDER_45): Likewise.
538 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
541 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
542 * vec.c (qsort_chk): Use gcc_qsort.
543 * Makefile.in (OBJS-libcommon): Add sort.o.
544 (build/sort.o): New target. Use it...
545 (BUILD_RTL): ... here, and...
546 (build/gencfn-macros): ... here, and...
547 (build/genmatch): ... here.
549 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
550 Chung-Ju Wu <jasonwucj@gmail.com>
552 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
553 * config/nds32/nds32-graywolf.md: New file.
554 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
555 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
557 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
558 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
559 * config/nds32/nds32.md (pipeline_model): Add graywolf.
560 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
561 * config/nds32/pipelines.md: Include n15 settings.
563 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
564 Chung-Ju Wu <jasonwucj@gmail.com>
566 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
567 * config/nds32/nds32-n13.md: New file.
568 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
569 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
571 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
572 * config/nds32/nds32.md (pipeline_model): Add n13.
573 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
574 * config/nds32/pipelines.md: Include n13 settings.
576 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
577 Chung-Ju Wu <jasonwucj@gmail.com>
579 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
580 * config/nds32/nds32-n10.md: New file.
581 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
582 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
584 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
585 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
586 * config/nds32/nds32.md (pipeline_model): Add n10.
587 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
588 * config/nds32/pipelines.md: Include n10 settings.
590 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
591 Kito Cheng <kito.cheng@gmail.com>
592 Chung-Ju Wu <jasonwucj@gmail.com>
594 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
595 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
596 Add enum values for DSP extension instructions.
597 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
599 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
600 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
602 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
603 * config/nds32/nds32-dspext.md: New file for DSP implementation.
604 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
605 * config/nds32/nds32-intrinsic.md: Likewise.
606 * config/nds32/nds32_intrinsic.h: Likewise.
607 * config/nds32/nds32-md-auxiliary.c: Likewise.
608 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
609 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
610 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
611 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
612 * config/nds32/nds32-protos.h: New declarations for DSP extension.
613 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
614 TYPE_DMAC in switch statement.
615 * config/nds32/nds32.c: New checking and implementation for DSP
616 extension instructions.
617 * config/nds32/nds32.h: Likewise.
618 * config/nds32/nds32.md: Likewise.
619 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
620 * config/nds32/predicates.md: Implement new predicates for DSP
623 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
625 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
626 Reformat alternatives and attributes so it is easier to identify
627 which constraints/attributes go with which instruction.
628 (mov<mode>_hardfloat32, FMOVE64): Likewise.
629 (mov<mode>_softfloat32, FMOVE64): Likewise.
630 (mov<mode>_hardfloat64, FMOVE64): Likewise.
631 (mov<mode>_softfloat64, FMOVE64): Likewise.
633 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
635 * doc/extend.texi (PowerPC Built-in Functions): Rename this
637 (Basic PowerPC Built-in Functions): The new name of the
638 subsection previously known as "PowerPC Built-in Functions".
639 (Basic PowerPC Built-in Functions Available on all Configurations):
641 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
642 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
643 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
644 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
646 2018-05-11 Martin Jambor <mjambor@suse.cz>
649 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
652 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
655 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
657 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
659 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
660 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
661 (ix86_handle_option): Handle -mwaitpkg.
662 * config.gcc: New header.
663 * config/i386/cpuid.h (bit_WAITPKG): New bit.
664 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
665 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
667 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
668 OPTION_MASK_ISA_WAITPKG.
669 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
670 (ix86_option_override_internal): Add PTA_WAITPKG.
671 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
672 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
674 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
675 __builtin_ia32_umwait and __builtin_ia32_tpause.
676 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
677 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
678 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
679 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
680 UNSPECV_TPAUSE): New.
681 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
682 * config/i386/i386.opt: Add -mwaitpkg.
683 * config/i386/waitpkgintrin.h: New file.
684 * config/i386/x86intrin.h: New header.
685 * doc/invoke.texi: Add -mwaitpkg.
687 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
690 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
692 (cortex-m0): Use armv6s-m isa.
693 (cortex-m0plus): Likewise.
694 (cortex-m1): Likewise.
695 (cortex-m0.small-multiply): Likewise.
696 (cortex-m0plus.small-multiply): Likewise.
697 (cortex-m1.small-multiply): Likewise.
699 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
700 Jakub Jelinek <jakub@redhat.com>
702 PR tree-optimization/85692
703 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
704 source permute as well.
706 2018-05-11 Martin Liska <mliska@suse.cz>
709 * doc/extend.texi: Document LLVM style format for no_sanitize
712 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
714 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
715 mode_supports_vsx_dform_quad to mode_supports_dq_form.
716 (mode_supports_vsx_dform_quad): Likewise.
717 (mode_supports_vmx_dform): Move these functions to be next to the
718 other mode_supports functions.
719 (mode_supports_dq_form): Likewise.
720 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
721 mode_supports_dq_form.
722 (reg_offset_addressing_ok_p): Likewise.
723 (offsettable_ok_by_alignment): Likewise.
724 (rs6000_legitimate_offset_address_p): Likewise.
725 (legitimate_lo_sum_address_p): Likewise.
726 (rs6000_legitimize_address): Likewise.
727 (rs6000_legitimize_reload_address): Likewise.
728 (rs6000_secondary_reload_inner): Likewise.
729 (rs6000_preferred_reload_class): Likewise.
730 (rs6000_output_move_128bit): Likewise.
732 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
734 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
735 Generate SImode target register for null target.
736 <case IX86_BUILTIN_XGETBV>: Ditto.
737 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
738 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
740 2018-05-10 Carl Love <cel@us.ibm.com>
742 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
743 dcbtt and dcbtstt if operands[2] is 0.
745 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
748 * config/i386/sse.md (usadv64qi): New expander.
750 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
752 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
753 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
754 -maltivec=be support.
755 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
756 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
757 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
758 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
759 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
760 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
761 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
762 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
763 altivec_vsumsws): Adjust.
764 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
765 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
766 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
768 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
769 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
770 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
771 (altivec_lve<VI_char>x): Delete expand.
772 (*altivec_lve<VI_char>x_internal): Rename to...
773 (altivec_lve<VI_char>x): ... this.
774 (altivec_lvxl_<mode>): Delete expand.
775 (*altivec_lvxl_<mode>_internal): Rename to ...
776 (altivec_lvxl_<mode>): ... this.
777 (altivec_stvxl_<mode>): Delete expand.
778 (*altivec_stvxl_<mode>_internal): Rename to ...
779 (altivec_stvxl_<mode>): ... this.
780 (altivec_stve<VI_char>x): Delete expand.
781 (*altivec_stve<VI_char>x_internal): Rename to ...
782 (altivec_stve<VI_char>x): ... this.
783 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
784 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
785 reduc_plus_scal_<mode>): Adjust.
786 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
788 (rs6000_cpu_cpp_builtins): Adjust.
789 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
790 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
791 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
792 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
793 -maltivec=be support.
794 (rs6000_split_vec_extract_var): Adjust.
795 (rs6000_split_v4si_init): Adjust.
796 (swap_selector_for_mode): Delete.
797 (altivec_expand_lvx_be, altivec_expand_stvx_be,
798 altivec_expand_stvex_be): Delete.
799 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
800 -maltivec=be support.
801 (rs6000_gimple_fold_builtin): Ditto.
802 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
804 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
805 (TARGET_DIRECT_MOVE_64BIT): Adjust.
806 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
807 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
808 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
809 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
810 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
811 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
812 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
813 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
814 anonymous split): Adjust.
815 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
816 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
818 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
820 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
821 when --with-gxx-include-dir is also specified.
822 * configure: Regenerate.
824 2018-05-09 Jim Wilson <jimw@sifive.com>
827 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
828 * config/riscv/t-withmultilib: New.
829 * config/riscv/withmultilib.h: New.
830 * doc/install.texi: Document RISC-V --with-multilib-list support.
832 2018-05-09 Richard Biener <rguenther@suse.de>
834 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
836 (vect_bb_vectorization_profitable_p): Adjust. Compute
837 actual scalar cost using the cost vector and the add_stmt_cost
840 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
842 PR rtl-optimization/85645
843 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
844 in the REG_CFA_REGISTER note for LR, don't leave it empty.
846 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
848 PR rtl-optimization/85645
849 * shrink-wrap.c (spread_components): Return a boolean saying if
850 anything was changed.
851 (try_shrink_wrapping_separate): Iterate spread_components until
852 nothing changes anymore.
854 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
856 PR rtl-optimization/85645
857 * regrename.c (build_def_use): Also kill the chains that include the
858 destination of a REG_CFA_REGISTER note.
860 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
862 PR rtl-optimization/85645
863 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
864 insn that has a REG_CFA_REGISTER note.
866 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
868 * cfgexpand.c (expand_clobber): New function.
869 (expand_gimple_stmt_1): Use it.
870 * tree-vect-stmts.c (vect_clobber_variable): New function,
872 (vectorizable_simd_clone_call): ...here.
873 (vectorizable_store): Emit a clobber either side of an
874 IFN_STORE_LANES sequence.
875 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
877 2018-05-09 Tom de Vries <tom@codesourcery.com>
880 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
881 (define_insn "trap_if_false"): Add exit after trap.
883 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
885 PR rtl-optimization/85638
886 * bb-reorder.c: Include common/common-target.h.
887 (create_forwarder_block): New function extracted from...
888 (fix_up_crossing_landing_pad): ...here. Rename into...
889 (dw2_fix_up_crossing_landing_pad): ...this.
890 (sjlj_fix_up_crossing_landing_pad): New function.
891 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
892 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
893 from both partitions and exit the loop after one iteration.
895 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
898 * doc/extend.texi (PowerPC Built-in Functions): Rename this
900 (Basic PowerPC Built-in Functions): The new name of the
901 subsection previously known as "PowerPC Built-in Functions".
902 (Basic PowerPC Built-in Functions Available on all Configurations):
904 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
906 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
908 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
910 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
913 2018-05-08 Jim Wilson <jimw@sifive.com>
915 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
916 (LD_EMUL_SUFFIX): New.
919 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
921 * doc/extend.texi (PowerPC Built-in Functions): Rename this
923 (Basic PowerPC Built-in Functions): The new name of the
924 subsection previously known as "PowerPC Built-in Functions".
925 (Basic PowerPC Built-in Functions Available on all Configurations):
927 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
929 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
931 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
933 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
936 2018-05-08 Jakub Jelinek <jakub@redhat.com>
939 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
940 after cmpelim optimization.
942 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
944 * config.gcc: Support "goldmont".
945 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
946 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
948 * config/i386/i386.c (m_GOLDMONT): Define.
949 (processor_target_table): Add "goldmont".
950 (PTA_GOLDMONT): Define.
951 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
952 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
953 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
954 (fold_builtin_cpu): Add "goldmont".
955 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
956 (ix86_option_override_internal): Add "goldmont".
957 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
958 (processor_type): Add PROCESSOR_GOLDMONT.
959 * config/i386/i386.md: Add CPU "glm".
960 * config/i386/glm.md: New file.
961 * config/i386/x86-tune.def: Add m_GOLDMONT.
962 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
964 2018-05-08 Jakub Jelinek <jakub@redhat.com>
967 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
969 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
970 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
971 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
974 * config/i386/i386.c (ix86_fold_builtin): Handle
975 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
978 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
979 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
981 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
984 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
985 (check_arch): Likewise.
986 (check_fpu): Return the result rather than printing it.
987 (end arch): Fix operator precedence.
989 (END): Print the result from check_fpu.
991 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
992 Alan Hayward <alan.hayward@arm.com>
993 David Sherwood <david.sherwood@arm.com>
995 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
996 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
997 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
998 (*fcmuo<mode>_and): New patterns.
1000 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1002 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1003 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1004 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1005 (cmp_op, sve_imm_con): New code attributes.
1006 (SVE_COND_INT_CMP, imm_con): Delete.
1007 (cmp_op): Remove above unspecs from int attribute.
1008 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1010 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1011 comparison-specific unspecs.
1012 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1013 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1014 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1015 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1016 (*vec_fcm<cmp_op><mode>): Rename to...
1017 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1018 (*vec_fcmuo<mode>): Rename to...
1019 (*fcmuo<mode>): ...this and adjust likewise.
1020 (*pred_fcm<cmp_op><mode>): New pattern.
1021 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1022 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1024 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1026 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1027 (aarch64_emit_sve_predicated_cond): New function.
1028 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1029 (aarch64_emit_unspec_cond_or): Replace with...
1030 (aarch64_emit_sve_or_conds): ...this new function. Use
1031 aarch64_emit_sve_ptrue_op for the individual comparisons and
1032 aarch64_emit_binop to OR them together.
1033 (aarch64_emit_inverted_unspec_cond): Replace with...
1034 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1035 aarch64_emit_sve_ptrue_op for the comparison and
1036 aarch64_emit_unop to invert the result.
1037 (aarch64_expand_sve_vec_cmp_float): Update after the above
1038 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1040 2018-05-07 Nathan Sidwell <nathan@acm.org>
1042 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1043 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1044 (Backwards Compatibility): Likewise.
1046 2018-05-07 Luis Machado <luis.machado@linaro.org>
1050 2018-05-07 Luis Machado <luis.machado@linaro.org>
1052 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1053 <prefetch_dynamic_strides>: New const bool field.
1054 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1055 prefetch_dynamic_strides.
1056 (exynosm1_prefetch_tune): Likewise.
1057 (thunderxt88_prefetch_tune): Likewise.
1058 (thunderx_prefetch_tune): Likewise.
1059 (thunderx2t99_prefetch_tune): Likewise.
1060 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1061 (aarch64_override_options_internal): Update to set
1062 PARAM_PREFETCH_DYNAMIC_STRIDES.
1063 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1064 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1065 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1066 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1067 prefetch-dynamic-strides setting.
1069 2018-05-07 Luis Machado <luis.machado@linaro.org>
1071 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1072 <minimum_stride>: New const int field.
1073 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1074 minimum_stride field.
1075 (exynosm1_prefetch_tune): Likewise.
1076 (thunderxt88_prefetch_tune): Likewise.
1077 (thunderx_prefetch_tune): Likewise.
1078 (thunderx2t99_prefetch_tune): Likewise.
1079 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1080 (aarch64_override_options_internal): Update to set
1081 PARAM_PREFETCH_MINIMUM_STRIDE.
1082 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1083 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1084 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1085 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1086 stride is constant and is below the minimum stride threshold.
1088 2018-05-07 Luis Machado <luis.machado@linaro.org>
1090 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1093 2018-05-07 Luis Machado <luis.machado@linaro.org>
1095 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1096 <prefetch_dynamic_strides>: New const bool field.
1097 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1098 prefetch_dynamic_strides.
1099 (exynosm1_prefetch_tune): Likewise.
1100 (thunderxt88_prefetch_tune): Likewise.
1101 (thunderx_prefetch_tune): Likewise.
1102 (thunderx2t99_prefetch_tune): Likewise.
1103 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1104 (aarch64_override_options_internal): Update to set
1105 PARAM_PREFETCH_DYNAMIC_STRIDES.
1106 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1107 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1108 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1109 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1110 prefetch-dynamic-strides setting.
1112 2018-05-07 Luis Machado <luis.machado@linaro.org>
1114 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1115 <minimum_stride>: New const int field.
1116 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1117 minimum_stride field.
1118 (exynosm1_prefetch_tune): Likewise.
1119 (thunderxt88_prefetch_tune): Likewise.
1120 (thunderx_prefetch_tune): Likewise.
1121 (thunderx2t99_prefetch_tune): Likewise.
1122 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1123 (aarch64_override_options_internal): Update to set
1124 PARAM_PREFETCH_MINIMUM_STRIDE.
1125 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1126 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1127 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1128 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1129 stride is constant and is below the minimum stride threshold.
1131 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1134 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1135 the type is addressable. Don't force op into register if it has
1138 2018-05-05 Roland McGrath <mcgrathr@google.com>
1141 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1142 any section for which we don't know a specific type it should have,
1143 regardless of name. Previously this was done only for the exact
1144 names ".init_array", ".fini_array", and ".preinit_array".
1145 (default_elf_asm_named_section): Add comment about
1146 relationship with default_section_type_flags and SECTION_NOTYPE.
1147 (get_section): Don't consider it a type conflict if one side has
1148 SECTION_NOTYPE and the other doesn't, as long as neither has the
1149 SECTION_BSS et al used in the default_section_type_flags logic.
1151 2018-05-05 Tom de Vries <tom@codesourcery.com>
1154 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1155 (workaround_barsyncs): New function.
1156 (nvptx_reorg): Use workaround_barsyncs.
1157 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1158 (define_expand "nvptx_membar_cta"): New define_expand.
1159 (define_insn "*nvptx_membar_cta"): New insn.
1161 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1163 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1164 To improve optimization opportunities.
1165 * builtin-types.def: The new needed builtin types for the above.
1167 2018-05-04 Richard Biener <rguenther@suse.de>
1169 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1170 * gimple-ssa-store-merging.c
1171 (imm_store_chain_info::output_merged_store): Remove redundant create,
1172 release split_store vector contents on failure.
1173 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1174 scalar stmt vector on cache hit.
1176 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1178 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1180 * config.gcc (powerpc-xilinx-eabi*): Remove.
1181 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1183 (fusion_addis_mem_combo_load): Ditto.
1184 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1186 (rs6000_cpu_cpp_builtins): Ditto.
1187 * config/rs6000/rs6000-linux.c
1188 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1189 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1190 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1192 (rs6000_setup_reg_addr_masks): Ditto.
1193 (rs6000_init_hard_regno_mode_ok): Ditto.
1194 (rs6000_option_override_internal): Ditto.
1195 (legitimate_lo_sum_address_p): Ditto.
1196 (rs6000_legitimize_address): Ditto.
1197 (rs6000_legitimize_reload_address): Ditto.
1198 (rs6000_legitimate_address_p): Ditto.
1199 (abi_v4_pass_in_fpr): Ditto.
1200 (setup_incoming_varargs): Ditto.
1201 (rs6000_gimplify_va_arg): Ditto.
1202 (rs6000_split_multireg_move): Ditto.
1203 (rs6000_savres_strategy): Ditto.
1204 (rs6000_emit_prologue_components): Ditto.
1205 (rs6000_emit_epilogue_components): Ditto.
1206 (rs6000_emit_prologue): Ditto.
1207 (rs6000_emit_epilogue): Ditto.
1208 (rs6000_elf_file_end): Ditto.
1209 (rs6000_function_value): Ditto.
1210 (rs6000_libcall_value): Ditto.
1211 * config/rs6000/rs6000.h: Ditto.
1212 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1213 (TARGET_MINMAX): ... this. New.
1214 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1215 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1216 (*movsi_internal1_single): Delete.
1217 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1218 mfpu=, mxilinx-fpu): Delete.
1219 * config/rs6000/singlefp.h: Delete.
1220 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1221 * config/rs6000/t-rs6000: Ditto.
1222 * config/rs6000/t-xilinx: Delete.
1223 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1224 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1225 (VStype_simple): Delete.
1226 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1227 * config/rs6000/xfpu.h: Delete.
1228 * config/rs6000/xfpu.md: Delete.
1229 * config/rs6000/xilinx.h: Delete.
1230 * config/rs6000/xilinx.opt: Delete.
1231 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1232 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1234 2018-05-04 Tom de Vries <tom@codesourcery.com>
1237 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1240 2018-05-04 Richard Biener <rguenther@suse.de>
1243 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1244 (expand_complex_div_wide): Likewise.
1245 (expand_complex_operations_1): Likewise.
1246 (expand_complex_libcall): Preserve EH info of the original stmt.
1247 (tree_lower_complex): Handle removed blocks.
1248 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1249 on complex multiplication and division libcall builtins.
1251 2018-05-04 Richard Biener <rguenther@suse.de>
1254 * fold-const.c (negate_expr_p): Restrict negation of operand
1255 zero of a division to when we know that can happen without
1257 (fold_negate_expr_1): Likewise.
1259 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1262 * real.h (real_nextafter): Declare.
1263 * real.c (real_nextafter): New function.
1264 * fold-const-call.c (fold_const_nextafter): New function.
1265 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1266 CASE_CFN_NEXTTOWARD.
1267 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1268 even when arg1_mode is different from arg0_mode.
1270 2018-05-03 Nathan Sidwell <nathan@acm.org>
1272 * doc/extend.texi (Deprecated Features): Remove
1274 (Backwards Compatibility): Likewise.
1275 * doc/invoke.texi (C++ Language Options): Likewise.
1276 (C++ Dialect Options): Likewise.
1278 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1281 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1282 _mm512_mask_mullox_epi64): New intrinsics.
1284 2018-05-03 Tom de Vries <tom@codesourcery.com>
1287 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1288 dump files): Add offload-tree.
1290 2018-05-03 Richard Biener <rguenther@suse.de>
1292 PR tree-optimization/85615
1293 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1294 to loops not nested in BBs loop father to avoid creating multi-entry
1297 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1299 PR tree-optimization/70291
1300 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1301 arguments. Change return type to tree. Emit libcall as a new
1302 statement rather than replacing existing one when inplace_p is true.
1303 (expand_complex_multiplication_components): New function.
1304 (expand_complex_multiplication): Expand floating-point complex
1305 multiplication using the above.
1306 (expand_complex_division): Rename inner_type parameter to type.
1307 Update expand_complex_libcall call-site.
1308 (expand_complex_operations_1): Update expand_complex_multiplication
1309 and expand_complex_division call-sites.
1311 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1314 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1315 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1316 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1317 the highest significant bit of the shift count mask is clear. In
1318 check whether and[sq]i3 is needed verify that all significant bits
1319 of the shift count other than the highest are set.
1321 2018-05-02 Tom de Vries <tom@codesourcery.com>
1324 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1325 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1326 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1327 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1328 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1329 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1330 __builtin_goacc_parlevel_size.
1332 2018-05-02 Richard Biener <rguenther@suse.de>
1334 PR tree-optimization/85597
1335 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1336 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1339 2018-05-02 Tom de Vries <tom@codesourcery.com>
1342 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1343 dump files): Add ltrans-tree.
1345 2018-05-02 Tom de Vries <tom@codesourcery.com>
1348 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1349 dump files): Add wpa-ipa.
1351 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1353 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1354 powerpc*-*-linux*paired* target.
1355 * config/rs6000/750cl.h: Delete.
1356 * config/rs6000/paired.h: Delete.
1357 * config/rs6000/paired.md: Delete.
1358 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1360 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1361 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1362 comment. Remove paired float support.
1363 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1364 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1366 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1367 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1369 * config/rs6000/rs6000.c: Remove paired float support.
1370 (paired_expand_vector_init, paired_expand_vector_move,
1371 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1372 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1373 paired_expand_builtin, paired_expand_predicate_builtin,
1374 paired_init_builtins): Delete.
1375 * config/rs6000/rs6000.h: Remove paired float support.
1376 * config/rs6000/rs6000.md: Remove paired float support.
1377 (move_from_CR_ov_bit): Delete.
1378 * config/rs6000/rs6000.opt (mpaired): Delete.
1379 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1380 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1382 2018-05-02 Richard Biener <rguenther@suse.de>
1385 * gimplify.c (gimplify_save_expr): When in SSA form allow
1386 SAVE_EXPRs to compute to SSA vars.
1388 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1391 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1392 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1393 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1394 clobber operands[2], instead use a new pseudo. Formatting fixes.
1396 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1398 PR tree-optimization/85586
1399 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1400 exit early for statements in the same group if the accesses are
1403 2018-05-02 Tom de Vries <tom@codesourcery.com>
1406 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1409 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1411 PR tree-optimization/85143
1412 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1414 2018-05-01 Tom de Vries <tom@codesourcery.com>
1417 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1418 not found" error message.
1420 2018-05-01 Tom de Vries <tom@codesourcery.com>
1423 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1424 * vec.c (test_ordered_remove_if): New function.
1425 (vec_c_tests): Call test_ordered_remove_if.
1426 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1427 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1428 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1429 VEC_ORDERED_REMOVE_IF.
1431 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1433 PR tree-optimization/82665
1434 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1435 pointer subtraction where arguments come from a memchr call.
1437 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1439 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1440 --push-state --as-needed and --pop-state instead of --as-needed and
1441 --no-as-needed if ld supports it.
1442 * configure: Regenerated.
1445 * doc/install.texi2html: Replace _002d with - and _002a with * in
1446 generated html files using sed.
1448 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1451 * gcc-rich-location.c (blank_line_before_p): New function.
1452 (use_new_line): New function.
1453 (gcc_rich_location::add_fixit_insert_formatted): New function.
1454 * gcc-rich-location.h
1455 (gcc_rich_location::add_fixit_insert_formatted): New function.
1457 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1459 * selftest.c (assert_streq): Rename "expected" and "actual" to
1460 "val1" and "val2". Extend NULL-handling to cover both inputs
1461 symmetrically, while still requiring both to be non-NULL for a pass.
1462 * selftest.h (assert_streq): Rename "expected" and "actual" to
1464 (ASSERT_EQ): Likewise.
1465 (ASSERT_EQ_AT): Likewise.
1466 (ASSERT_KNOWN_EQ): Likewise.
1467 (ASSERT_KNOWN_EQ_AT): Likewise.
1468 (ASSERT_NE): Likewise.
1469 (ASSERT_MAYBE_NE): Likewise.
1470 (ASSERT_MAYBE_NE_AT): Likewise.
1471 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1472 the assertion to pass.
1473 (ASSERT_STREQ_AT): Likewise.
1475 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1477 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1478 interaction with -pie.
1480 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1482 * selftest.h: Fix alphabetization of per-source-file selftest
1485 2018-04-30 Jason Merrill <jason@redhat.com>
1487 PR c++/61982 - dead stores to destroyed objects.
1488 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1491 2018-04-30 Jason Merrill <jason@redhat.com>
1493 * tree.c (build_clobber): New.
1494 * tree.h: Declare it.
1495 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1497 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1499 * diagnostic-show-locus.c (layout::layout): Update for
1500 location_get_source_line returning a char_span.
1501 (struct char_span): Move to input.h.
1502 (struct correction): Update for fields in char_span becoming
1504 (struct source_line): Update for location_get_source_line
1505 returning a char_span.
1506 (layout::print_line): Likewise.
1507 * edit-context.c (edited_file::print_content): Likewise.
1508 (edited_file::print_diff_hunk): Likewise.
1509 (edited_file::print_run_of_changed_lines): Likewise.
1510 (edited_file::get_num_lines): Likewise.
1511 (edited_line::edited_line): Likewise.
1512 * final.c (asm_show_source): Likewise.
1513 * input.c (location_get_source_line): Convert return type
1514 from const char * to char_span, losing the final "line_len"
1516 (dump_location_info): Update for the above.
1517 (get_substring_ranges_for_loc): Likewise. Use a char_span
1518 when handling the literal within the line.
1519 (test_reading_source_line): Update for location_get_source_line
1520 returning a char_span.
1521 * input.h (class char_span): Move here from
1522 diagnostic-show-locus.c, converting from a struct to a class.
1523 Make data members private.
1524 (char_span::operator bool): New.
1525 (char_span::length): New.
1526 (char_span::get_buffer): New.
1527 (char_span::operator[]): New.
1528 (char_span::subspan): Make const.
1529 (char_span::xstrdup): New.
1530 (location_get_source_line): Convert return type from const char *
1531 to char_span, losing the final "line_size" param.
1533 2018-04-30 Jan Hubicka <jh@suse.cz>
1535 * lto-wrapper.c (ltrans_priorities): New static var.
1536 (cmp_priority): New.
1537 (run_gcc): Read priorities and if doing parallel build order
1538 the Makefile by them.
1540 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1542 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1544 2018-04-30 Richard Biener <rguenther@suse.de>
1546 * tree-cfg.c (verify_address): Remove base argument, add
1547 flag whether to check TREE_ADDRESSABLE and do that.
1548 (verify_expr): Remove.
1549 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1550 (verify_gimple_assign_single): Likewise.
1551 (verify_gimple_switch): Likewise.
1552 (verify_expr_location_1): Dereference tp once. Add (disabled)
1553 piece from verify_expr.
1554 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1556 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1558 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1560 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1562 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1563 (small_data_pattern): Likewise.
1564 (arc_rewrite_small_data): Likewise.
1565 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1566 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1567 (get_symbol_alignment): New function.
1568 (legitimate_small_data_address_p): Likewise.
1569 (legitimate_scaled_address): Update, call
1570 legitimate_small_data_address_p.
1571 (output_sdata): New static variable.
1572 (arc_print_operand): Update how we handle small data operands.
1573 (arc_print_operand_address): Likewise.
1574 (arc_legitimate_address_p): Update, use
1575 legitimate_small_data_address_p.
1576 (arc_rewrite_small_data_p): Remove.
1577 (arc_rewrite_small_data_1): Likewise.
1578 (arc_rewrite_small_data): Likewise.
1579 (small_data_pattern): Likewise.
1580 (compact_sda_memory_operand): Update to use
1581 legitimate_small_data_address_p and get_symbol_alignment.
1582 (prepare_move_operands): Don't rewite sdata pattern.
1583 (prepare_extend_operands): Remove.
1584 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1586 (zero_extendqisi2): Likewise.
1587 (zero_extendhisi2): Likewise.
1588 (extendqihi2): Likewise.
1589 (extendqisi2): Likewise.
1590 (extendhisi2): Likewise.
1594 * config/arc/constraints.md (Usd): Change it to memory constraint.
1596 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1598 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1599 as source of std instructions.
1600 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1601 allow 6-bit constants as source for store instructions.
1602 (movdi_insn): Update instruction pattern to allow 6-bit constants
1603 as source for store instructions.
1605 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1607 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1609 2018-04-30 Nathan Sidwell <nathan@acm.org>
1610 Sandra Loosemore <sandra@codesourcery.com>
1612 * dumpfile.c (dump_open): Allow '-' for stdout.
1613 * doc/invoke.texi (Developer Options): Document dump filename
1614 determination early. Document stdin/stdout selection.
1616 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1618 Microblaze Target: PIC data text relative
1620 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1621 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1623 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1624 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1625 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1626 New addressing mode for data-text relative position indepenedent code.
1627 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1628 'ADDRESS_SYMBOLIC_TXT_REL'.
1629 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1630 (microblaze_legitimate_pic_operand): Exclude function calls from
1631 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1632 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1634 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1635 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1636 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1637 for 'address + offset'.
1638 (microblaze_expand_prologue): Add new function prologue call for
1640 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1641 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1642 table in case of TARGET_PIC_DATA_TEXT_REL.
1643 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1644 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1645 Add new macros 'UNSPEC_TEXT',
1646 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1647 + exclude function calls from 'UNSPEC_PLT' in case of data text
1649 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1650 new target hook for generating address diff vector tables in case of
1652 * doc/tm.texi : Regenerate.
1653 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1654 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1655 of addr diff vector generation.
1656 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1657 target hook definition.
1658 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1659 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1660 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1661 option for data text relative.
1663 2018-04-30 Richard Biener <rguenther@suse.de>
1665 * tree-chrec.h (evolution_function_is_constant_p): Remove
1667 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1669 2018-04-30 Richard Biener <rguenther@suse.de>
1672 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1674 2018-04-30 Richard Biener <rguenther@suse.de>
1676 PR tree-optimization/28364
1677 PR tree-optimization/85275
1678 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1679 copying first exit test.
1681 2018-04-28 Mark Wielaard <mark@klomp.org>
1683 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1685 (dwarf_AT): Handle DW_AT_addr_base.
1686 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1688 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1691 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1692 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1693 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1694 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1696 2018-04-28 Richard Biener <rguenther@suse.de>
1698 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1699 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1700 to reflect use. Only add interesting stmts.
1702 2018-04-27 Martin Jambor <mjambor@suse.cz>
1705 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1706 the jump function allows for passing through aggregate values.
1708 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1710 * input.h (in_system_header_at): Convert from macro to inline
1712 (from_macro_expansion_at): Likewise.
1713 (from_macro_definition_at): Likewise.
1715 2018-04-27 Jeff Law <law@redhat.com>
1717 * config.gcc: Mark tile* targets as deprecated/obsolete.
1719 2018-04-27 Richard Biener <rguenther@suse.de>
1721 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1724 2018-04-27 Richard Biener <rguenther@suse.de>
1726 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1728 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1730 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1731 with Yd constraint. Set "preferred_for_speed" attribute from
1732 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1734 (*movdi_internal): Ditto.
1735 (movti_interunit splitters): Remove
1736 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1737 (movdi_interunit splitters): Ditto.
1738 * config/i386/constraints.md (Ye): Remove.
1739 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1741 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1744 * config/aarch64/constraints.md (Usg): Limit to 31.
1747 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1749 PR tree-optimization/85529
1750 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1751 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1752 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1753 zero extension or masking of the MSB bit.
1754 (optimize_range_tests): Add FIRST_BB argument, pass it through
1755 to optimize_range_tests_var_bound.
1756 (maybe_optimize_range_tests, reassociate_bb): Adjust
1757 optimize_range_tests callers.
1759 2018-04-26 Richard Biener <rguenther@suse.de>
1760 Jakub Jelinek <jakub@redhat.com>
1762 * cgraph.h (symbol_table): Just declare debug method here.
1763 * symtab.c (symbol_table::debug): Define.
1765 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1767 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1769 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1771 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1772 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1773 (*movdi_internal): Substitute Yi and Yj constraint with x
1774 and Ym and Yn constraint with y constraint. Update "isa"
1775 attribute and set "preferred_for_speed" attribute from
1776 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1777 (*movsi_internal): Ditto.
1778 (*movdf_internal): Ditto.
1779 (*movsf_internal): Ditto.
1780 (*zero_extendsidi2): Ditto.
1781 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1782 (sse2_loadld): Ditto.
1783 (*vec_extract<ssevecmodelower>_0): Ditto.
1784 (*vec_extractv4si_0_zext_sse4): Ditto.
1785 (vec_concatv2di): Ditto.
1786 (*vec_dup<mode>): Ditto.
1787 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1788 * config/i386/constraints.md (Yi): Remove.
1793 2018-04-26 Nathan Sidwell <nathan@acm.org>
1795 * dumpfile.c (dump_open): New.
1796 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1797 (dump_finish): Detect stdio/stderr by value not name.
1799 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1801 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1803 2018-04-26 Tom de Vries <tom@codesourcery.com>
1806 * config/nvptx/nvptx.c (verify_neutering_jumps)
1807 (verify_neutering_labels): New function
1808 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1810 2018-04-26 Tom de Vries <tom@codesourcery.com>
1813 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1814 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1817 2018-04-26 Richard Biener <rguenther@suse.de>
1818 Tom de Vries <tom@codesourcery.com>
1821 * lto-streamer-out.c (output_function): Fixup loops if required to match
1822 discovery done in the reader.
1824 2018-04-26 Richard Biener <rguenther@suse.de>
1826 PR tree-optimization/85116
1827 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1828 have a loop exit from the single latch predecessor. Remove
1829 case of header with just condition.
1830 (ch_base::copy_headers): Exclude infinite loops from any
1832 (pass_ch::execute): Record exits.
1834 2018-04-26 Richard Biener <rguenther@suse.de>
1836 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1837 prologue cost vector and pass it to vect_get_load_cost.
1838 (vect_get_peeling_costs_all_drs): Likewise.
1839 (vect_peeling_hash_get_lowest_cost): Likewise.
1840 (vect_enhance_data_refs_alignment): Likewise.
1842 2018-04-26 Richard Biener <rguenther@suse.de>
1845 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1846 checking of integer<->pointer conversions.
1847 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1848 sign-/zero-extending pointer types.
1849 (expand_omp_for_static_chunk): Likewise.
1851 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1852 Jean Lee <xiaoyur347@gmail.com>
1854 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1855 (TARGET_ASAN_SHADOW_OFFSET): Define.
1856 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1857 true for -fsanitize=address.
1859 2018-04-25 Mark Wielaard <mark@klomp.org>
1861 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1864 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1866 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1867 than "alu", remove explicit "memory" and "imm_disp" attributes.
1868 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1871 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1872 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1873 gen_lowpart_no_emit.
1875 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1878 * config/i386/i386.c (ix86_expand_builtin): Change memory
1879 operand to XI, extend p0 to Pmode.
1880 * config/i386/i386.md: Change unspec volatile and operand
1881 1 mode to XI, change operand 0 mode to P.
1883 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1885 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1886 GET_MODE_MASK before any checking.
1887 (nds32_can_use_bset_p): Likewise.
1888 (nds32_can_use_btgl_p): Likewise.
1890 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1892 * config/nds32/nds32-doubleword.md: New define_split pattern for
1893 illegal register number.
1895 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1897 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1899 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1901 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1903 2018-04-25 Richard Biener <rguenther@suse.de>
1905 * lto-streamer.h (LTO_major_version): Bump to 8.
1907 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1909 * BASE-VER: Set to 9.0.0.
1911 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1913 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1914 in __abskf2 and __powikf2.
1916 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1919 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1920 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1921 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1922 Use the above on operand 2. Reindent.
1923 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1925 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1928 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1930 * config/i386/i386.opt (mcet): Removed.
1931 * doc/install.texi: Remove -mcet documentation.
1932 * doc/invoke.texi: Likewise.
1934 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1937 * doc/install.texi: Remove -mcet from bootstrap-cet.
1939 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1942 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1943 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1947 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1948 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1949 containing a CONST_VECTOR.
1951 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1953 * doc/install.texi: Update newlib dependency for nvptx.
1955 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1958 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1959 instead of INTVAL when shifting x left.
1961 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1963 PR tree-optimization/85478
1964 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1965 vect_grouped_store_supported for single element vectors.
1967 2018-04-24 Richard Biener <rguenther@suse.de>
1970 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1971 load cost increase to the case of non-constant step.
1973 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1976 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1977 destination if any_malformed_asm.
1979 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1982 * expr.c (store_field): In the bitfield case, if the value comes from
1983 a function call and is returned in registers by means of a PARALLEL,
1984 do not change the mode of the temporary unless BLKmode and VOIDmode.
1986 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1988 PR rtl-optimization/85423
1989 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1990 dependencies to debug insns when the previous insn is non-debug.
1992 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1994 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1995 enums into a single definition.
1996 (fls): Fix predicates and printing.
1999 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2001 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2002 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2003 and short u6 immediate.
2004 (check_if_valid_sleep_operand): Remove.
2005 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2007 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2009 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2010 flag_always_save_lp condition.
2011 * config/nds32/nds32.opt (malways-save-lp): New option.
2013 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2015 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2016 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2017 * config/nds32/nds32.h
2018 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2019 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2021 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2023 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2024 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2026 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2027 Chung-Ju Wu <jasonwucj@gmail.com>
2029 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2031 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2032 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2034 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2036 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2038 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2040 * config/nds32/nds32-protos.h (nds32_data_alignment,
2041 nds32_local_alignment): Declare.
2042 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2043 nds32_local_alignment): New functions.
2044 (TARGET_CONSTANT_ALIGNMENT): Define.
2045 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2047 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2049 * config/nds32/nds32.c
2050 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2051 (TARGET_MODES_TIEABLE_P): Likewise.
2053 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2055 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2058 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2059 Chung-Ju Wu <jasonwucj@gmail.com>
2061 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2062 for unaligned access.
2063 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2065 * config/nds32/nds32-intrinsic.md: Likewise.
2066 * config/nds32/nds32_intrinsic.h: Likewise.
2067 * config/nds32/nds32.h (nds32_builtins): Likewise.
2068 * config/nds32/nds32.opt (munaligned-access): New option.
2069 * config/nds32/nds32.c (nds32_asm_file_start): Display
2070 flag_unaligned_access status.
2072 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2074 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2075 -mno-relax is present.
2076 * config/riscv/linux.h (LINK_SPEC): Ditto.
2078 2018-04-20 Martin Sebor <msebor@redhat.com>
2081 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2083 (gimple_fold_builtin_stxcpy_chk): Same.
2084 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2086 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2089 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2090 __powikf2 when long double is IEEE 128-bit.
2092 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2094 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2095 step to make sure stack always aligned.
2097 2018-04-20 Carl Love <cel@us.ibm.com>
2100 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2101 size check for arg0.
2103 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2104 Tom de Vries <tom@codesourcery.com>
2107 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2108 Emit insns for calls too.
2109 (nvptx_find_par): Always look for worker-level predecessor insn.
2110 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2112 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2113 (nvptx_process_pars): Propagate frames for calls.
2115 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2118 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2120 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2121 (ix86_handle_option): Don't handle OPT_mibt.
2122 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2124 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2126 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2127 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2128 (ix86_target_macros): Define __CET__ with flag_cf_protection
2129 for -fcf-protection.
2130 * config/i386/i386.c (isa2_opts): Remove -mibt.
2131 * config/i386/i386.h (TARGET_IBT): Removed.
2132 (TARGET_IBT_P): Likewise.
2133 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2134 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2135 * config/i386/i386.opt (mcet): Update help message.
2138 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2139 -mcet as an alias for -mshstk.
2141 2018-04-20 Richard Biener <rguenther@suse.de>
2144 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2145 complexity by forcing a single use of the multiply operand.
2147 2018-04-20 Martin Jambor <mjambor@suse.cz>
2150 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2151 recursion dependency to only apply to non-clones.
2153 2018-04-20 Martin Jambor <mjambor@suse.cz>
2156 * ipa-cp.c (create_specialized_node): Check that clones of
2157 self-recursive edges exist during IPA-CP.
2159 2018-04-19 Toon Moene <toon@moene.org>
2161 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2164 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2166 PR tree-optimization/85467
2167 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2168 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2169 VECTOR_CST element to type.
2171 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2174 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2175 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2176 (builtin_longjmp): Likewise.
2177 (save_stack_nonlocal): New pattern.
2178 (restore_stack_nonlocal): Likewise.
2180 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2183 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2184 Replace ASM_OUTPUT_LABEL with fprintf.
2186 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2189 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2190 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2191 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2192 define __IBT__ and __SHSTK__ for -fcf-protection.
2193 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2195 (ix86_trampoline_init): Likewise.
2196 (x86_output_mi_thunk): Likewise.
2197 (ix86_notrack_prefixed_insn_p): Likewise.
2198 (ix86_option_override_internal): Don't disallow -fcf-protection.
2199 * config/i386/i386.md (rdssp<mode>): Also enable for
2201 (incssp<mode>): Likewise.
2202 (nop_endbr): Likewise.
2203 * config/i386/i386.opt (mcet): Change help message to built-in
2207 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2208 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2209 enable CET built-in functions.
2211 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2213 * common/config/i386/i386-common.c
2214 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2215 OPTION_MASK_ISA_MOVDIRI_UNSET,
2216 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2217 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2218 * config.gcc (movdirintrin.h): New header.
2219 * config/i386/cpuid.h (bit_MOVDIRI,
2220 bit_MOVDIR64B): New bits.
2221 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2223 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2224 (VOID, PVOID, PCVOID)): New function types.
2225 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2226 __builtin_ia32_directstoreu_u64,
2227 __builtin_ia32_movdir64b): New builtins.
2228 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2229 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2231 (ix86_valid_target_attribute_inner_p): Ditto.
2232 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2233 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2234 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2235 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2236 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2237 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2238 (movdiri<mode>, movdir64b_<mode>): New.
2239 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2240 * config/i386/immintrin.h: Include movdirintrin.h.
2241 * config/i386/movdirintrin.h: New file.
2242 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2244 2018-04-19 Richard Biener <rguenther@suse.de>
2247 * cfg.c (clear_bb_flags): When loop state says we have
2248 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2250 2018-04-19 Richard Biener <rguenther@suse.de>
2252 PR tree-optimization/84737
2253 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2254 copying restrict info.
2255 (vect_setup_realignment): Use it.
2256 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2257 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2258 the first DR to all generated stores.
2259 (vectorizable_load): Likewise for loads.
2261 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2263 PR tree-optimization/85446
2264 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2265 the integral and pointer types to have the same precision.
2267 * doc/install.texi: Document --disable-cet being the default and
2270 2018-04-18 Martin Liska <mliska@suse.cz>
2272 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2275 2018-04-18 Martin Liska <mliska@suse.cz>
2278 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2281 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2282 arguments if they are comparable.
2284 2018-04-18 Martin Liska <mliska@suse.cz>
2287 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2290 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2293 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2296 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2297 ENDBR after calling __morestack.
2299 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2302 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2303 by using gcc_base_ver to generate a gcc_driver_version, and use
2304 it when generating GCC_DRIVER_NAME.
2305 * configure: Regenerate.
2307 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2310 * config.gcc: Obsolete powerpc*-*-*spe*.
2312 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2315 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2316 (stabstr_D): Change type of unum from unsigned int to
2317 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2320 2018-04-17 Jim Wilson <jimw@sifive.com>
2323 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2324 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2325 Set arg_pointer_offset after using pretend_args_size.
2327 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2329 PR rtl-optimization/85431
2330 * dse.c (record_store): Ignore zero width stores.
2333 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2334 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2335 __builtin_stack_restore rather than after it.
2336 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2337 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2338 argument instead of virtual_dynamic_stack_rtx.
2340 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2342 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2344 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2345 Add note to error message to explain internal mapping of overloaded
2346 built-in function name to non-overloaded built-in function name.
2347 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2350 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2353 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2354 where the inputs overlap with the output.
2356 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2358 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2359 (=v, v) alternative and explicit "memory" attribute.
2360 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2361 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2363 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2364 "sselog1" type instead of "sselog".
2365 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2366 "sselog". Remove explicit "memory" attribute.
2367 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2368 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2370 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2371 "sselog1" type instead of "sselog". Remove explicit "memory"
2373 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2374 use "sselog1" type instead of "sselog". Remove explicit "memory"
2376 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2377 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2379 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2380 "sselog1" type instead of "sselog". Remove explicit "memory"
2382 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2383 use "sselog1" type instead of "sselog". Remove explicit "memory"
2387 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2390 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2393 2018-04-17 Martin Jambor <mjambor@suse.cz>
2396 * ipa-cp.c (create_specialized_node): Call
2397 expand_all_artificial_thunks if necessary.
2399 2018-04-17 Martin Liska <mliska@suse.cz>
2402 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2403 in message, remote space in between '_G' and '('.
2405 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2408 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2409 avx512f_vmcmp<mode>3<round_saeonly_name>,
2410 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2411 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2412 avx512f_rndscale<mode><round_saeonly_name>,
2413 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2414 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2415 Use %<iptr>2 instead of %2 for -masm=intel.
2416 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2417 avx512f_vcvttss2usi<round_saeonly_name>,
2418 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2420 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2421 avx512f_vcvttsd2usi<round_saeonly_name>,
2422 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2423 Use %q1 instead of %1 for -masm=intel.
2424 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2425 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2426 of %3 for -masm=intel.
2427 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2429 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2431 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2433 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2434 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2436 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2438 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2439 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2440 %g1 and one with %0 and %1.
2441 (avx512er_vmrcp28<mode><round_saeonly_name>,
2442 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2444 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2445 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2446 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2447 of %0 and %{%4%} for -masm=intel.
2448 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2449 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2450 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2451 order of %0 and %{%5%}%{z%} for -masm=intel.
2453 2018-04-17 Jan Hubicka <jh@suse.cz>
2456 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2458 2018-04-17 Martin Liska <mliska@suse.cz>
2461 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2462 for target_clone error message. Make default implementation
2463 clone to be a local declaration.
2464 (separate_attrs): Add new argument and check for an empty
2466 (expand_target_clones): Handle it.
2467 (ipa_target_clone): Make redirection just for target_clones
2470 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2471 Tom de Vries <tom@codesourcery.com>
2474 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2475 tiled basic blocks without omp continue statements.
2477 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2480 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2481 vec_extract expression as having side effects to make sure it gets
2484 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2487 * config/i386/i386.c (get_builtin_code_for_version): Check
2490 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2493 * gcc/config.gcc: Support "skylake".
2494 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2496 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2497 (processor_target_table): Add "skylake".
2498 (ix86_option_override_internal): Add "skylake".
2499 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2500 PROCESSOR_CANNONLAKE.
2501 (get_builtin_code_for_version): Fix priority for
2502 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2503 PROCESSOR_SKYLAKE-AVX512.
2504 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2505 (processor_type): Add PROCESSOR_SKYLAKE.
2507 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2508 Jason Merrill <jason@redhat.com>
2511 * convert.c (convert_to_integer_1): Use direct recursion for
2512 enumeral types and types with a precision less than the number
2513 of bits in their mode.
2515 2018-04-16 Julia Koval <julia.koval@intel.com>
2518 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2519 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2521 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2524 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2525 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2526 and -mno-direct-move.
2528 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2531 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2532 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2533 Ensure negative shifts result in {0}.
2535 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2537 PR rtl-optimization/79916
2538 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2539 regs (if any) to define how to gnerate SD moves when LRA is in
2542 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2544 PR rtl-optimization/85393
2545 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2546 * except.c (expand_dw2_landing_pad_for_region): Make static.
2547 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2548 a label and unconditional jump to old_bb, rather than
2549 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2552 PR rtl-optimization/85376
2553 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2554 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2555 instead of a specific value.
2557 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2558 Bin Cheng <bin.cheng@arm.com>
2560 PR tree-optimization/82965
2561 PR tree-optimization/83991
2562 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2563 by_profile_only parameter.
2564 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2565 information if the loop was predicted to iterate too many times.
2566 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2568 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2571 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2574 2018-04-13 Martin Liska <mliska@suse.cz>
2575 Jakub Jelinek <jakub@redhat.com>
2578 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2579 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2580 * builtins.c (expand_builtin_memory_copy_args): Use
2581 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2582 handle dest_addr == pc_rtx.
2584 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2587 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2588 asked to not generate direct moves.
2589 (fix_trunc<mode>si2_stfiwx): Similar.
2590 (fix_trunc<mode>si2_internal): Similar.
2592 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2595 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2596 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2597 lookup if dest in some wider mode is known to be const0_rtx and
2598 if so, record permanent equivalence for it to be ZERO_EXTEND of
2599 the narrower mode destination.
2601 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2603 * lto-streamer-out.c (output_function): Revert 259346.
2604 * omp-expand.c (expand_oacc_for): Likewise.
2606 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2608 PR rtl-optimization/85354
2609 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2610 * sel-sched.c (sel_global_init): ... here.
2612 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2615 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2616 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2617 mode for PE-COFF targets.
2618 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2619 (i386_pe_asm_lto_end): Likewise.
2620 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2621 (TARGET_ASM_LTO_END): Likewise.
2622 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2623 (i386_pe_asm_lto_start): New function.
2624 (i386_pe_asm_lto_end): Likewise.
2626 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2627 Richard Biener <rguenther@suse.de>
2630 * lto-streamer-out.c (output_function): Fix CFG loop state before
2632 * omp-expand.c (expand_oacc_for): Handle calls to internal
2633 functions like regular functions.
2635 2018-04-12 Richard Biener <rguenther@suse.de>
2638 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2639 for the early LTO debug to properly generate references to it
2640 during DIE emission. Do not re-use that for the skeleton for
2642 (dwarf2out_early_finish): Likewise.
2644 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2647 * config/i386/sse.md
2648 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2649 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2650 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2651 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2652 and output is a reg, avoid creating invalid lowpart subreg, but
2653 instead split into a 512-bit move. Don't split if not AVX512VL,
2654 input is xmm16+ reg and output is a mem.
2655 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2656 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2657 xmm16+ reg and output is a mem.
2659 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2661 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2662 also for flag_dwarf2_cfi_asm.
2664 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2666 PR rtl-optimization/85342
2667 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2668 a bool scalar var inside of the loop instead. Don't try to update
2669 recog_data.operand after failed apply_change_group.
2671 2018-04-12 Tom de Vries <tom@codesourcery.com>
2674 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2675 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2676 array with flexible array member as array without given dimension.
2677 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2678 argument for undefined param to true.
2680 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2683 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2684 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2685 from PowerPC section.
2686 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2687 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2688 help text that is too long.
2689 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2690 help text that is too long.
2691 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2692 help text that is too long.
2694 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2696 * config/alpha/alpha.md (stack_probe_internal): Rename
2697 from "probe_stack". Update all callers.
2699 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2701 PR rtl-optimization/84566
2702 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2703 sched_macro_fuse_insns.
2705 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2708 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2709 (compute_block_dependences): ... from here.
2711 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2713 PR tree-optimization/85331
2714 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2715 from int to HOST_WIDE_INT.
2717 2018-04-11 Martin Jambor <mjambor@suse.cz>
2720 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2721 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2722 not the same as the source val.
2723 (cgraph_edge_brings_value_p): New parameter.
2724 (gather_edges_for_value): Pass destination value to
2725 cgraph_edge_brings_value_p.
2726 (perhaps_add_new_callers): Likewise.
2727 (get_info_about_necessary_edges): Likewise and exclude values brought
2728 only by self-recursive edges.
2729 (create_specialized_node): Redirect only clones of self-calling edges.
2730 (+self_recursive_pass_through_p): New function.
2731 (find_more_scalar_values_for_callers_subset): Use it.
2732 (find_aggregate_values_for_callers_subset): Likewise.
2733 (known_aggs_to_agg_replacement_list): Removed.
2734 (decide_whether_version_node): Re-calculate known constants for all
2735 remaining context clones.
2737 2018-04-11 Richard Biener <rguenther@suse.de>
2740 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2741 from early DWARF output.
2742 (dwarf2out_early_finish): Output line info unconditionally into
2743 early DWARF and add reference to it.
2745 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2748 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2749 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2750 other than V2DFmode using iptr mode attribute.
2751 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2753 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2755 PR rtl-optimization/84659
2756 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2758 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2761 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2763 (output_loc_list): Pass address of a dummy size variable even in the
2764 locview handling loop.
2765 (index_location_lists): Add comment on why skip_loc_list_entry can't
2768 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2771 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2774 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2777 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2778 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2779 and -mstring-compare-inline-limit.
2781 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2784 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2785 for stack clash protection in a register whenever we need it to be in
2788 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2790 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2791 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2793 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2796 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2798 (mlong-double-): Ditto.
2799 * config/rs6000/sysv4.opt (msdata=): Ditto.
2800 (mtls-size=): Ditto.
2802 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2804 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2805 erroneous entries for
2806 "vector int vec_ldl (int, long int *)", and
2807 "vector unsigned int vec_ldl (int, unsigned long int *)".
2808 Add comments and entries for
2809 "vector bool char vec_ldl (int, bool char *)",
2810 "vector bool short vec_ldl (int, bool short *)",
2811 "vector bool int vec_ldl (int, bool int *)",
2812 "vector bool long long vec_ldl (int, bool long long *)",
2813 "vector pixel vec_ldl (int, pixel *)",
2814 "vector long long vec_ldl (int, long long *)",
2815 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2816 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2817 type tree bool_long_long_type_node and correct definition of
2818 bool_V2DI_type_node to make reference to this new type tree.
2819 (rs6000_mangle_type): Replace erroneous reference to
2820 bool_long_type_node with bool_long_long_type_node.
2821 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2822 comments to emphasize sign distinctions for char and int types and
2823 replace RS6000_BTI_bool_long constant with
2824 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2825 use of RS6000_BTI_pixel.
2826 (bool_long_type_node): Remove this macro definition.
2827 (bool_long_long_type_node): New macro definition
2829 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2831 PR rtl-optimization/85300
2832 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2833 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2834 simplify_unary_operation fails.
2836 2018-04-10 Martin Liska <mliska@suse.cz>
2838 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2839 cgraph_edge and ipa_ref.
2841 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2845 * config/i386/sse.md
2846 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2847 computation of the VEC_MERGE selector from mask.
2848 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2849 Fix decoding of the VEC_MERGE selector into mask.
2851 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2853 PR tree-optimization/85286
2854 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2856 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2858 * final.c (final_1): Set insn_last_address as well as
2859 insn_current_address.
2861 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2864 * explow.c (emit_stack_probe): Call validize_mem on memory location
2865 before passing it to gen_probe_stack. Create address operand and
2866 legitimize it for the probe_stack_address case.
2868 2018-04-09 Jan Hubicka <jh@suse.cz>
2871 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2872 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2873 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2875 (free_lang_data): Rebuild type inheritance graph.
2877 2018-04-09 Martin Sebor <msebor@redhat.com>
2879 * invoke.texi (-finline-small-functions): Mention other optimization
2881 (-findirect-inlining, -fpartial-inlining): Same.
2882 (-finline-functions-called-once): Same.
2883 (-freorder-blocks-and-partition): Same.
2885 2018-04-09 Jan Hubicka <jh@suse.cz>
2888 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2889 jumps; choose last target that matches the criteria (i.e.
2890 no partition changes for non-crossing jumps).
2891 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2892 support for redirecting crossing jumps to non-crossing.
2894 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2896 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2897 also for naked functions.
2899 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2901 * config/arc/arc.md (add_shift): New pattern.
2902 (add_shift2): Likewise.
2903 (sub_shift): Likewise.
2904 (sub_shift_cmp0_noout): Likewise.
2905 (compare_si_ashiftsi): Likewise.
2906 (xbfu_cmp0_noout): New combine pattern.
2907 (xbfu_cmp0"): Likewise.
2908 (movsi_set_cc_insn): Place the predicable variant first.
2909 (commutative_binary_cmp0_noout): Remove clobber.
2910 (commutative_binary_cmp0): New pattern.
2911 (noncommutative_binary_cmp0): Likewise.
2912 (noncommutative_binary_cmp0_noout): Likewise.
2913 (noncommutative_binary_comparison_result_used): Removed.
2914 (rsub_cmp0): New pattern.
2915 (rsub_cmp0_noout): Likewise.
2916 (extzvsi): Changed, keep only meaningful variants.
2917 (SQH, SEZ): New iterators.
2918 (SQH_postfix): New mode attribute.
2919 (SEZ_prefix): New code attribute.
2920 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2921 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2922 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2924 (noncommutative_operator): Check the availability of barrel
2927 2018-04-09 Richard Biener <rguenther@suse.de>
2929 PR tree-optimization/85284
2930 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2931 Only use the niter constraining form of simple_iv when the exit
2934 2018-04-09 Tom de Vries <tom@codesourcery.com>
2937 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2938 (define_expand "*memory_barrier"): New define_expand.
2939 (define_insn "memory_barrier"): New insn.
2941 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2943 PR rtl-optimization/80463
2944 PR rtl-optimization/83972
2945 PR rtl-optimization/83480
2947 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2948 correct producer for the insn.
2949 (tidy_control_flow): Fixup seqnos in case of debug insns.
2951 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2953 PR rtl-optimization/83913
2955 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2956 different sched-times when merging exprs.
2958 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2960 PR rtl-optimization/83962
2962 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2963 tidy_fallthru_edge and tidy_control_flow.
2965 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2967 PR rtl-optimization/83530
2969 * sel-sched.c (force_next_insn): New global variable.
2970 (remove_insn_for_debug): When force_next_insn is true, also leave only
2971 next insn in the ready list.
2972 (sel_sched_region): When the region wasn't scheduled, make another pass
2973 over it with force_next_insn set to 1.
2975 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2977 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2979 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2980 for interrupt control.
2981 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2982 functions for interrupt control.
2983 * config/nds32/nds32-intrinsic.md: Likewise.
2984 * config/nds32/nds32_intrinsic.h: Likewise.
2985 * config/nds32/nds32.h (nds32_builtins): Likewise.
2987 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2989 * config/nds32/nds32.c (nds32_init_machine_status,
2990 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2991 strict_aligned_p field.
2992 (nds32_expand_to_rtl_hook): New function.
2993 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2994 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2996 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2997 Chung-Ju Wu <jasonwucj@gmail.com>
2999 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3000 * config/nds32/nds32-n7.md: New file.
3001 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3002 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3004 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3005 * config/nds32/nds32.md (pipeline_model): Add n7.
3006 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3007 * config/nds32/pipelines.md: Include n7 settings.
3009 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3010 Chung-Ju Wu <jasonwucj@gmail.com>
3012 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3013 * config/nds32/nds32-e8.md: New file.
3014 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3015 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3017 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3018 * config/nds32/nds32.md (pipeline_model): Add e8.
3019 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3020 * config/nds32/pipelines.md: Include e8 settings.
3022 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3023 Chung-Ju Wu <jasonwucj@gmail.com>
3025 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3026 * config/nds32/nds32-n8.md: New file.
3027 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3028 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3030 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3031 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3032 * config/nds32/nds32.md (pipeline_model): Add n8.
3033 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3034 * config/nds32/pipelines.md: Include n8 settings.
3036 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3037 Chung-Ju Wu <jasonwucj@gmail.com>
3039 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3040 * config/nds32/nds32-n9-2r1w.md: New file.
3041 * config/nds32/nds32-n9-3r2w.md: New file.
3042 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3043 nds32_register_ports): New or modify for cpu n9.
3044 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3046 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3047 * config/nds32/nds32-utils.c: New file.
3048 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3049 TARGET_MUL_SLOW): Define.
3050 * config/nds32/nds32.md (pipeline_model): New attribute.
3051 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3052 New options that support cpu n9.
3053 * config/nds32/pipelines.md: Include n9 settings.
3054 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3056 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3058 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3059 information if necessary.
3060 (output_cond_branch_compare_zero): Likewise.
3061 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3062 (nds32_target_alignment): Refine for alignment.
3063 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3064 (FUNCTION_BOUNDARY): Modify.
3065 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3067 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3069 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3071 * config/nds32/constants.md (unspec_volatile_element): Add values for
3072 TLB operation and data prefetch.
3073 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3074 functions for TLB operation and data prefetch.
3075 * config/nds32/nds32-intrinsic.md: Likewise.
3076 * config/nds32/nds32_intrinsic.h: Likewise.
3077 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3078 (nds32_print_operand): Likewise.
3079 * config/nds32/nds32.h (nds32_builtins): Likewise.
3081 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3082 Andrew Pinski <pinsika@gcc.gnu.org>
3085 * match.pd: Use constant_boolean_node of correct type instead of
3086 boolean_true_node or boolean_false_node for simplifying
3087 pointer comparisons to zero.
3089 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3091 PR tree-optimization/80021
3092 * tree.c (verify_type_variant): Make error call in verify_variant_match
3093 translatable and remove final full stop.
3095 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3097 * config/nds32/constants.md (unspec_volatile_element): Add
3098 UNSPEC_VOLATILE_EH_RETURN.
3099 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3100 nds32_output_stack_pop): Support dwarf exception handling process.
3101 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3102 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3103 exception handling process.
3104 (nds32_compute_stack_frame): Likewise.
3105 (nds32_return_addr_rtx): Likewise.
3106 (nds32_initial_elimination_offset): Likewise.
3107 (nds32_expand_prologue): Likewise.
3108 (nds32_expand_epilogue): Likewise.
3109 (nds32_dynamic_chain_address): New function.
3110 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3112 (DYNAMIC_CHAIN_ADDRESS): Define.
3113 (EH_RETURN_DATA_REGNO): Define.
3114 (EH_RETURN_STACKADJ_RTX): Define.
3115 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3116 patterns for dwarf exception handling.
3118 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3120 * config/nds32/nds32.h: Clean up obsolete macros.
3122 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3124 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3125 Add enum values for particular instructions.
3126 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3127 particular intrinsic functions.
3128 * config/nds32/nds32-intrinsic.md: Likewise.
3129 * config/nds32/nds32_intrinsic.h: Likewise.
3130 * config/nds32/nds32.h (nds32_builtins): Likewise.
3131 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3132 (btst, ave): New patterns for particular instructions.
3134 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3136 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3137 Add enum values for atomic load/store and memory sync.
3138 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3140 * config/nds32/nds32-intrinsic.md: Likewise.
3141 * config/nds32/nds32_intrinsic.h: Likewise.
3142 * config/nds32/nds32.h (nds32_builtins): Likewise.
3144 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3146 PR tree-optimization/85257
3147 * fold-const.c (native_encode_vector): If not all elts could fit
3148 and off is -1, return 0 rather than offset.
3149 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3150 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3151 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3152 adjust buffer in native_interpret_expr call.
3154 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3156 * config/nds32/constants.md (unspec_volatile_element): Add cache
3157 control enum values.
3158 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3159 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3160 * config/nds32/nds32.c (nds32_cctl_names): New.
3161 (nds32_print_operand): Handle cache control register names.
3162 * config/nds32/nds32.h (nds32_builtins): New enum values.
3163 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3165 * config/nds32/nds32.md (type): Add mmu.
3166 * config/nds32/pipelines.md (simple_insn): Add mmu.
3168 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3170 * config/nds32/nds32.md (type): Remove call.
3171 * config/nds32/pipelines.md (simple_insn): Likewise.
3173 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3175 * config/nds32/constants.md (unspec_volatile_element): Add
3176 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3177 UNSPEC_VOLATILE_FMFCFG.
3178 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3179 description for fmfcfg and fmfcsr.
3180 (bdesc_1arg): Add fmtcsr.
3181 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3182 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3183 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3184 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3185 unspec_fmfcfg): New patterns.
3186 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3187 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3188 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3189 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3190 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3191 __nds32__fmfcfg): Define.
3193 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3195 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3196 intrinsic register names.
3197 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3198 intrinsic register enum values and macros.
3200 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3202 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3203 for load/store addressing form.
3204 (nds32_print_operand_address): Likewise.
3206 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3209 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3210 based on LABEL_REF. Remove useless assertion.
3211 (pic_address_needs_scratch): Fix formatting.
3212 (sparc_legitimize_pic_address): Minor tweaks.
3213 (sparc_delegitimize_address): Adjust assertion accordingly.
3214 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3215 into symbolic_operand.
3216 (movsi_high_pic_label_ref): Likewise.
3217 (movsi_lo_sum_pic_label_ref): Likewise.
3218 (movdi_pic_label_ref): Likewise.
3219 (movdi_high_pic_label_ref): Likewise.
3220 (movdi_lo_sum_pic_label_ref): Likewise.
3222 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3224 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3225 custom LIB_SPEC setup.
3227 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3228 Kito Cheng <kito.cheng@gmail.com>
3230 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3231 * config/riscv/freebsd.h: New.
3233 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3235 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3236 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3239 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3240 Kito Cheng <kito.cheng@gmail.com>
3242 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3243 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3244 * config/nds32/nds32-protos.h (nds32_output_call,
3245 nds32_output_return): Declare.
3246 * config/nds32/nds32.md: Refine all the call and return patterns.
3248 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3251 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3252 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3254 PR rtl-optimization/84872
3255 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3256 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3259 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3261 * expr.c (copy_blkmode_to_reg): Revert 254862.
3262 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3264 2018-04-06 Richard Biener <rguenther@suse.de>
3267 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3268 after seeing a component reference with an adjacent field. Treat
3269 refs to arrays at struct end of external decls similar to
3270 refs to unconstrained commons.
3272 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3275 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3276 look through SAVE_EXPRs with non-side-effects argument. Adjust
3278 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3281 2018-04-06 Richard Biener <rguenther@suse.de>
3284 * alias.c (find_base_term): New wrapper around find_base_term
3285 unwinding CSELIB_VAL_PTR changes.
3286 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3289 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3291 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3293 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3294 constant definitions.
3295 ("nop"): lr 0,0 -> nopr r0
3296 ("nop_lr0", "nop_lr1"): New insn definitions.
3298 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3300 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3301 NDS32_V3PUSH_AVAILABLE_P macro.
3303 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3304 Chung-Ju Wu <jasonwucj@gmail.com>
3306 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3307 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3308 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3309 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3310 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3311 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3312 * config/nds32/constraints.md: New constraints and checking for hard
3313 float configuration.
3314 * config/nds32/iterators.md: New mode iterator and attribute for hard
3315 float configuration.
3316 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3318 * config/nds32/nds32-fpu.md: New file.
3319 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3320 deal with hard float code generation.
3321 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3323 (abi_type, float_reg_number): New enum type.
3324 * config/nds32/nds32-predicates.c: New predicates for hard float.
3325 * config/nds32/nds32-protos.h: Declare functions for hard float.
3326 * config/nds32/nds32.c: Implementation for hard float configuration.
3327 * config/nds32/nds32.h: Definitions for hard float configuration.
3328 * config/nds32/nds32.md: Include hard float machine description and
3329 modify patterns for hard float configuration.
3330 * config/nds32/nds32.opt: New options for hard float configuration.
3331 * config/nds32/predicates.md: New predicates for hard float
3334 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3336 * common/config/nds32/nds32-common.c
3337 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3339 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3342 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3343 CONSTRUCTOR_ELT (ctor, ...)->value.
3345 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3348 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3350 2018-04-05 Tom de Vries <tom@codesourcery.com>
3353 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3356 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3357 Kito Cheng <kito.cheng@gmail.com>
3359 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3360 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3361 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3362 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3364 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3365 Kito Cheng <kito.cheng@gmail.com>
3367 * config/nds32/constraints.md (Ufe): New memory constraint.
3368 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3369 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3370 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3372 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3373 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3375 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3377 * config/nds32/nds32.md: Use optimize_size in the condition for
3378 alu-shift instructions.
3380 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3382 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3384 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3386 * config/nds32/nds32.md (negsi2): Refine pattern.
3388 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3389 Chung-Ju Wu <jasonwucj@gmail.com>
3391 * config/nds32/iterators.md (shift_rotate): New code iterator.
3392 (shift): New code attribute.
3393 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3394 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3395 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3396 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3397 bit-wise operations.
3398 (andsi3, *andsi3): Ditto.
3399 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3400 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3401 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3402 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3403 nds32_ior_operand, nds32_xor_operand): New predicates.
3405 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3407 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3408 (addsi3, subsi3): ... this.
3410 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3412 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3414 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3416 * config/nds32/nds32.md: Adjust indention.
3418 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3420 * config/nds32/nds32.md (feature): New attribute.
3422 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3424 * config/nds32/nds32.md (subtype): New attribute.
3426 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3429 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3430 expansion to perform a bitwise AND of the argument followed by a
3431 boolean negation of the result.
3433 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3435 PR rtl-optimization/84878
3436 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3437 the basic block. Assert the use reference is not artificial and that
3438 it has an associated insn.
3440 2018-04-04 Michael Matz <matz@suse.de>
3442 * builtins.c (compute_objsize): Pass correct operand
3443 to array_at_struct_end_p.
3445 2018-04-04 Richard Biener <rguenther@suse.de>
3448 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3449 from contexts for DINFO_LEVEL_TERSE and below.
3451 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3453 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3454 resiter_operand condition.
3455 * config/nds32/nds32.md (*move<mode>): Ditto.
3457 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3458 Monk Chiang <sh.chiang04@gmail.com>
3460 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3462 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3464 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3466 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3467 Kito Cheng <kito.cheng@gmail.com>
3469 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3470 nds32_cond_code_str, output_cond_branch,
3471 output_cond_branch_compare_zero, nds32_expand_cbranch,
3472 nds32_expand_cstore, nds32_expand_movcc,
3473 nds32_output_cbranchsi4_equality_zero,
3474 nds32_output_cbranchsi4_equality_reg,
3475 nds32_output_cbranchsi4_equality_reg_or_const_int,
3476 nds32_output_cbranchsi4_greater_less_zero: New functions.
3477 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3478 nds32_expand_cstore, nds32_expand_movcc,
3479 nds32_output_cbranchsi4_equality_zero,
3480 nds32_output_cbranchsi4_equality_reg,
3481 nds32_output_cbranchsi4_equality_reg_or_const_int,
3482 nds32_output_cbranchsi4_greater_less_zero): Declare.
3483 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3484 nds32_rimm11s_operand): New predicates.
3485 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3486 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3489 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3491 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3492 * config/nds32/nds32.md: Ditto.
3493 * config/nds32/pipelines.md: Ditto.
3495 2018-04-04 Richard Biener <rguenther@suse.de>
3497 PR tree-optimization/85168
3498 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3499 propagating abnormals.
3501 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3503 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3505 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3506 Kito Cheng <kito.cheng@gmail.com>
3508 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3509 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3510 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3511 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3512 * config/nds32/nds32.md (sibcall_internal): New.
3513 (sibcall_register): Remove.
3514 (sibcall_immediate): Remove.
3515 (sibcall_value_internal): New.
3516 (sibcall_value_register): Remove.
3517 (sibcall_value_immediate): Remove.
3518 * config/nds32/predicates.md (nds32_general_register_operand): New.
3519 (nds32_call_address_operand): New.
3521 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3523 PR rtl-optimization/85167
3524 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3525 bb_defs if *split_p, instead preinitialize it to NULL.
3527 PR tree-optimization/85156
3528 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3529 evaluating the argument multiple times.
3531 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3533 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3535 (_mm_cvtpd_ps): Likewise.
3536 (_mm_cvttpd_epi32): Likewise.
3537 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3538 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3539 vector, pixel, and bool following altivec.h include.
3541 2018-04-03 Martin Sebor <msebor@redhat.com>
3543 * doc/extend.texi (Common Function Attributes): Clarify.
3544 (const attribute): Likewise.
3545 (pure attribute): Likewise.
3547 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3550 * config/i386/i386.c (ix86_expand_vector_set): Use
3551 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3553 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3555 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3556 instructions when changing rounding bits to preserve precision bits
3557 in the x87 control word.
3559 2018-04-03 Martin Liska <mliska@suse.cz>
3561 PR tree-optimization/82491
3562 * rtl.h (strip_offset_and_add): Replace += suboffset with
3563 poly_uint64 () + suboffset.
3565 2018-03-29 Martin Liska <mliska@suse.cz>
3566 Martin Jambor <mjambor@suse.cz>
3569 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3570 param_type is not an integral or pointer type.
3572 2018-04-03 Richard Biener <rguenther@suse.de>
3574 * sese.h (recompute_all_dominators): Remove.
3576 2018-04-02 Martin Sebor <msebor@redhat.com>
3578 * doc/invoke.texi (-Wrestrict): Fix typos.
3580 2018-04-02 Jim Wilson <jimw@sifive.com>
3582 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3583 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3584 (<optab>di3, <optab>si3_extend): Likewise.
3585 (<optab>si3_mask, <optab>si3_mask_1): New.
3586 (<optab>di3_mask, <optab>di3_mask_1): New.
3587 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3588 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3589 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3591 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3593 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3596 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3598 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3599 (nds32_canonicalize_comparison): New function.
3601 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3602 Kito Cheng <kito.cheng@gmail.com>
3603 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3605 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3606 * config/nds32/constants.md (unspec_volatile_element): Add
3607 UNSPEC_VOLATILE_RELAX_GROUP.
3608 * config/nds32/nds32-relax-opt.c: New file.
3609 * config/nds32/nds32-predicates.c
3610 (nds32_symbol_load_store_p): New function.
3611 * config/nds32/nds32-protos.h
3612 (nds32_symbol_load_store_p): Declare function.
3613 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3614 * config/nds32/nds32.c
3615 (nds32_register_pass): New function to register pass.
3616 (nds32_register_passes): New function to register passes.
3617 * config/nds32/nds32.md (relax_group): New pattern.
3618 * config/nds32/nds32.opt (mrelax-hint): New option.
3619 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3621 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3623 * config/nds32/t-nds32: Modify files dependency.
3625 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3627 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3628 (PROFILE_HOOK): Define its implementation.
3630 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3632 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3633 type and 32-bit size.
3635 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3638 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3639 (V_128_256): New mode iterator.
3640 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3641 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3642 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3644 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3645 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3647 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3650 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3651 NaN inputs correctly.
3653 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3656 * config/rs6000/vsx.md (??r): New mode attribute.
3657 (*vsx_mov<mode>_64bit): Use it.
3658 (*vsx_mov<mode>_32bit): Likewise.
3660 2018-03-30 Martin Sebor <msebor@redhat.com>
3662 PR tree-optimization/84818
3663 * builtins.c (check_access): Use warning_n.
3665 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3668 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3670 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3673 2018-03-30 Julia Koval <julia.koval@intel.com>
3676 * x86-tune.def (movx, partial_reg_dependency): Enable for
3679 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3682 * lra-constraints.c (process_alt_operands): Move setting
3683 this_alternative_matches below.
3685 2018-03-29 Martin Liska <mliska@suse.cz>
3688 * doc/invoke.texi: Document how LTO works with debug info.
3689 Describe auto-load support of binutils. Mention 'x86-64'
3690 as valid option value of -march option.
3692 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3694 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3697 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3698 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3699 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3702 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3705 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3706 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3707 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3708 for RS6000_BTM_POWERPC64.
3709 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3710 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3711 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3716 2018-03-28 Carl Love <cel@us.ibm.com>
3719 2017-09-27 Carl Love <cel@us.ibm.com>
3721 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3722 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3723 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3726 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3728 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3729 instead of __vector bool.
3730 (_mm_max_pu8): Likewise.
3731 (_mm_min_pi16): Likewise.
3733 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3736 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3737 (DIVWEUO): Likewise.
3739 (DIVDEUO): Likewise.
3740 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3741 DIVWEUO and DIVDEUO.
3742 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3743 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3744 (div_extend): Likewise.
3745 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3747 (__builtin_divweuo): Likewise.
3748 (__builtin_divdeo): Likewise.
3749 (__builtin_divdeuo): Likewise.
3751 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3754 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3755 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3757 PR tree-optimization/82004
3758 * gimple-match-head.c (optimize_pow_to_exp): New function.
3759 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3760 Don't fold to exp if optimize_pow_to_exp is false.
3762 2018-03-28 Martin Liska <mliska@suse.cz>
3765 * calls.c (initialize_argument_information): Fix trailing space.
3766 * common.opt: Fix typo and provide better explanation for
3767 -fsanitize-coverage option.
3768 * config/i386/i386.opt: Fix typo.
3770 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3771 Martin Liska <mliska@suse.cz>
3774 * gimplify.c (asan_poison_variable): Don't do the check for
3775 gimplify_omp_ctxp here.
3776 (gimplify_decl_expr): Do it here.
3777 (gimplify_target_expr): Likewise.
3779 2018-03-28 Martin Liska <mliska@suse.cz>
3782 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3783 chkp_type_bounds_count if MPX is not enabled.
3785 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3787 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3789 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3792 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3793 function to create the function decl for complex long double
3794 multiply and divide for -mabi=ieeelongdouble.
3795 (init_float128_ieee): Call it.
3797 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3800 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3801 -fcf-protection=branch -mibt.
3802 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3804 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3807 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3809 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3812 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3813 extern array declarations.
3815 2018-03-27 Richard Biener <rguenther@suse.de>
3818 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3819 explicit single_use checks.
3821 2018-03-27 Richard Biener <rguenther@suse.de>
3823 PR tree-optimization/85082
3824 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3827 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3829 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3830 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3831 Turn on fasynchronous-unwind-tables and funwind-tables.
3833 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3836 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3837 (*bmi_blsr_<mode>_ccz): Ditto.
3839 2018-03-26 Tom de Vries <tom@codesourcery.com>
3841 PR tree-optimization/85063
3842 * omp-general.c (offloading_function_p): New function. Factor out
3844 * omp-offload.c (pass_omp_target_link::gate): ... here.
3845 * omp-general.h (offloading_function_p): Declare.
3846 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3847 with attribute omp declare target for offloading functions.
3849 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3851 PR tree-optimization/84005
3852 * tree-data-ref.h (get_base_for_alignment): Declare.
3853 * tree-data-ref.c (get_base_for_alignment_1): New function.
3854 (get_base_for_alignment): Likewise.
3855 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3856 get_base_for_alignment to find a suitable base object, instead
3857 of always using drb->base_address.
3859 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3862 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3863 known size by default.
3865 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3868 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3869 and non BLKmode operands.
3871 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3874 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3875 Clean up attributes.
3877 2018-03-23 Richard Biener <rguenther@suse.de>
3880 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3881 we are going to emit early debug for LTO.
3883 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3886 * function.c (match_asm_constraints_1): Don't optimize if input
3887 doesn't satisfy general_operand predicate for output's mode.
3890 * alias.c (write_dependence_p): Don't require for x_canonicalized
3891 non-VOIDmode if x has VOIDmode.
3894 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3895 just don't try to optimize it rather than assert it never happens.
3897 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3899 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3900 macro expansions for definition of ST_INTERNAL_<mode> and
3901 LD_INTERNAL_<mode> builtins.
3902 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3904 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3906 (altivec_expand_st_builtin): Likewise.
3907 (altivec_expand_builtin): Remove calls to deleted functions.
3908 (rs6000_address_for_altivec): Delete this function.
3909 * config/rs6000/vector.md: Remove expands for
3910 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3912 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3915 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3916 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3917 re-computing once computed.
3918 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3919 (arm_init_machine_status): Initialize
3920 machine->static_chain_stack_bytes.
3922 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3925 * doc/extend.texi: Add four new prototypes for vec_ld.
3926 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3927 definitions for more logical presentation.
3928 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3929 entries for V1TI variants of __builtin_altivec_ld builtin.
3930 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3931 handling of V1TI variant of LVX icode pattern.
3932 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3933 (rs6000_gimple_fold_builtin): Likewise.
3934 (altivec_init_builtins): Add code to define
3935 __builtin_altivec_lvx_v1ti function.
3937 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3940 * function.c (match_asm_constraints_1): Don't do the optimization
3941 if input isn't a REG, SUBREG, MEM or constant.
3943 2018-03-22 Tom de Vries <tom@codesourcery.com>
3945 PR tree-optimization/84956
3946 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3947 bb_has_abnormal_pred.
3949 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3952 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3953 DECL_INITIAL (decl) to decl at the end.
3954 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3957 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3959 * doc/extend.texi (__builtin_tgmath): Document when complex
3960 integer types are treated as _Complex _Float64.
3962 2018-03-21 Tom de Vries <tom@codesourcery.com>
3964 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3966 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3968 PR tree-optimization/84960
3969 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3970 if it is ENTRY block, move them into single succ of ENTRY in that case.
3972 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3974 PR tree-optimization/84811
3975 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3976 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3977 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3978 (known_subrange_p): Update accordingly. Cast each value involved
3979 in the size comparison, rather than casting the result of the
3982 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3984 PR tree-optimization/84982
3985 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3986 by flipping the least significant bit rather than all bits from
3987 bitpos to bitpos + bitsize - 1.
3989 2018-03-21 Nathan Sidwell <nathan@acm.org>
3991 * doc/extend.texi (Deprecated Features): Remove mention of
3992 long-deleted deprecations.
3994 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3997 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3998 * configure: Regenerate.
4000 2018-03-21 Tom de Vries <tom@codesourcery.com>
4002 PR tree-optimization/83126
4003 * tree-parloops.c (num_phis): New function.
4004 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4006 2018-03-21 Nathan Sidwell <nathan@acm.org>
4008 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4009 mention anon-struct/union members and trailing attributes.
4011 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4013 PR tree-optimization/84969
4014 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4015 builtin memset partitions if they set different rhs values.
4017 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4019 PR rtl-optimization/84989
4020 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4021 VEC_DUPLICATE with scalar result mode.
4023 2018-03-21 Martin Liska <mliska@suse.cz>
4026 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4027 not intended return statement.
4029 2018-03-21 Martin Liska <mliska@suse.cz>
4032 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4033 (chkp_find_bound_slots_1): Limit number of iterations.
4035 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4038 * Minor grammar fixes for x86 options.
4040 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4043 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4044 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4046 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4049 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4050 (altivec_lvx_<mode>_1op): Likewise.
4051 (altivec_stvx_<mode>_2op): Likewise.
4052 (altivec_stvx_<mode>_1op): Likewise.
4053 (altivec_lvx_<VM2:mode>): New define_expand.
4054 (altivec_stvx_<VM2:mode>): Likewise.
4055 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4056 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4057 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4058 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4059 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4060 (rs6000_gen_lvx): Likewise.
4061 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4062 (altivec_expand_stv_builtin): Likewise.
4063 (altivec_expand_builtin): Likewise.
4064 * config/rs6000/vector.md: Likewise.
4066 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4069 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4072 2018-03-20 Richard Biener <rguenther@suse.de>
4075 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4076 sign-conversions as zero, fall back to standard scalar_stmt
4079 2018-03-20 Martin Liska <mliska@suse.cz>
4082 * predict.c (rebuild_frequencies): Handle case when we have
4083 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4085 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4088 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4089 flag_section_anchors.
4090 * varasm.c (use_blocks_for_decl_p): Remove hack for
4091 dw2_force_const_mem.
4094 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4096 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4097 be created, use lowpart_subreg of operands[0] rather than operands[0]
4099 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4100 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4101 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4102 and n constraint instead of aarch64_shift_imm_di and Usd.
4103 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4104 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4106 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4109 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4110 to favor GPR over NEON registers.
4111 (<shift>di3_neon): Likewise.
4113 2018-03-20 Tom de Vries <tom@codesourcery.com>
4116 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4117 (nvptx_process_pars): Emit bar.sync asap and alap.
4119 2018-03-20 Tom de Vries <tom@codesourcery.com>
4122 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4123 seen_label if seen_label is already set.
4125 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4128 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4129 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4130 Use 1U instead of 1. Formatting fixes.
4133 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4134 instead of TREE_TYPE (s1) for the return value.
4136 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4138 PR tree-optimization/84946
4139 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4140 bitsize + bitsize in poly_uint64 rather than poly_int64.
4143 * dwarf2asm.c: Include fold-const.c.
4144 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4145 of decl rather than decl itself.
4147 PR rtl-optimization/84643
4148 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4150 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4153 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4154 calling assemble_variable.
4156 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4159 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4160 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4162 2018-03-19 Jim Wilson <jimw@sifive.com>
4165 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4166 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4167 (riscv_first_stack_step): Likewise.
4168 (riscv_option_override): Use STACK_BOUNDARY instead of
4170 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4172 (BIGGEST_ALIGNMENT): Set to 128.
4173 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4174 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4177 2018-03-19 Richard Biener <rguenther@suse.de>
4179 PR tree-optimization/84933
4180 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4181 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4183 2018-03-19 Richard Biener <rguenther@suse.de>
4185 PR tree-optimization/84859
4186 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4187 (cond_if_else_store_replacement): Perform sinking operation on
4188 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4189 Generalize what a BB with a single eligible store is.
4191 2018-03-19 Richard Biener <rguenther@suse.de>
4193 PR tree-optimization/84929
4194 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4195 chrec_is_positive against non-chrec arg.
4197 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4200 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4202 2018-03-18 Martin Liska <mliska@suse.cz>
4204 PR rtl-optimization/84635
4205 * regrename.c (build_def_use): Use matches_mode only when
4208 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4210 PR tree-optimization/84913
4211 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4212 vectorize chains of COND_EXPRs.
4214 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4216 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4218 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4220 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4222 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4224 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4226 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4227 Kito Cheng <kito.cheng@gmail.com>
4229 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4230 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4231 (nds32_adjust_reg_alloc_order): New function.
4232 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4234 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4236 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4237 nds32_print_operand, nds32_print_operand_address): Use
4238 HOST_WIDE_INT_PRINT_DEC instead.
4240 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4242 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4244 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4247 * config/i386/i386.c (initial_ix86_tune_features,
4248 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4250 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4251 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4252 rather than 1u << ix86_tune. Formatting fix.
4253 (ix86_option_override_internal): Change ix86_arch_mask from
4254 unsigned int to unsigned HOST_WIDE_INT, initialize to
4255 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4256 (ix86_function_specific_restore): Likewise.
4258 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4261 * postreload.c (reload_combine_recognize_pattern): Perform
4262 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4263 truncate_int_for_mode the result for the destination's mode.
4266 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4267 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4269 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4272 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4275 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4276 regno_allocno_class_array and sorted_pseudos.
4277 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4278 insns where regno is used.
4280 2018-03-16 Martin Liska <mliska@suse.cz>
4283 * multiple_target.c (create_dispatcher_calls): Redirect
4284 reference in the symbol table.
4286 2018-03-16 Martin Liska <mliska@suse.cz>
4289 * multiple_target.c (create_dispatcher_calls): Redirect also
4292 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4296 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4297 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4298 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4300 2018-03-16 Julia Koval <julia.koval@intel.com>
4302 * doc/invoke.texi (Skylake Server): Add CLWB.
4303 Cannonlake): Remove CLWB.
4305 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4307 PR tree-optimization/84841
4308 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4310 (FLOAT_ONE_CONST_TYPE): Define.
4311 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4312 (sort_by_operand_rank): Put entries with higher constant_type last
4313 rather than first to match comments.
4315 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4317 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4320 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4323 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4324 check and use address of target always.
4326 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4329 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4330 (indirect_thunk_bnd_needed): Likewise.
4331 (indirect_thunks_used): Likewise.
4332 (indirect_thunks_bnd_used): Likewise.
4333 (indirect_return_needed): New.
4334 (indirect_return_bnd_needed): Likewise.
4335 (output_indirect_thunk_function): Add a bool argument for
4337 (output_indirect_thunk_function): Don't generate alias for
4338 function return thunk.
4339 (ix86_code_end): Call output_indirect_thunk_function to generate
4340 function return thunks.
4341 (ix86_output_function_return): Set indirect_return_bnd_needed
4342 and indirect_return_needed instead of indirect_thunk_bnd_needed
4343 and indirect_thunk_needed.
4345 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4347 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4348 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4349 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4351 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4352 Paul Hua <paul.hua.gm@gmail.com>
4355 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4357 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4359 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4360 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4363 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4366 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4367 instead of GET_MODE_SIZE when comparing Units.
4369 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4372 * varasm.c (hash_section): Return an unchangeble hash value
4373 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4374 Return !aarch64_can_use_per_function_literal_pools_p ().
4376 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4379 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4380 rather than address of cmode as last argument to prepare_cmp_insn.
4382 2018-03-15 Julia Koval <julia.koval@intel.com>
4384 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4385 F_AVX512VNNI, F_AVX512BITALG): New.
4387 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4390 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4391 insn for floating-point loads and stores.
4393 2018-03-14 Carl Love <cel@us.ibm.com>
4395 * config/rs6000/rs6000-c.c: Add macro definitions for
4396 ALTIVEC_BUILTIN_VEC_PERMXOR.
4397 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4398 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4399 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4400 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4402 * config/doc/extend.texi: Add prototypes for vec_permxor.
4404 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4407 * diagnostic-show-locus.c (class layout_point): Convert m_line
4408 from int to linenum_type.
4409 (line_span::comparator): Use linenum "compare" function when
4410 comparing line numbers.
4411 (test_line_span): New function.
4412 (layout_range::contains_point): Convert param "row" from int to
4414 (layout_range::intersects_line_p): Likewise.
4415 (layout::will_show_line_p): Likewise.
4416 (layout::print_source_line): Likewise.
4417 (layout::should_print_annotation_line_p): Likewise.
4418 (layout::print_annotation_line): Likewise.
4419 (layout::print_leading_fixits): Likewise.
4420 (layout::annotation_line_showed_range_p): Likewise.
4421 (struct line_corrections): Likewise for field m_row.
4422 (line_corrections::line_corrections): Likewise for param "row".
4423 (layout::print_trailing_fixits): Likewise.
4424 (layout::get_state_at_point): Likewise.
4425 (layout::get_x_bound_for_row): Likewise.
4426 (layout::print_line): Likewise.
4427 (diagnostic_show_locus): Likewise for locals "last_line" and
4429 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4430 * input.c (selftest::test_linenum_comparisons): New function.
4431 (selftest::input_c_tests): Call it.
4432 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4433 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4434 * selftest.h (ASSERT_GT): New macro.
4435 (ASSERT_GT_AT): New macro.
4436 (ASSERT_LT): New macro.
4437 (ASSERT_LT_AT): New macro.
4439 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4441 PR rtl-optimization/84780
4442 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4444 2018-03-14 Martin Liska <mliska@suse.cz>
4446 * tree.c (record_node_allocation_statistics): Use
4447 get_stats_node_kind.
4448 (get_stats_node_kind): New function extracted from
4449 record_node_allocation_statistics.
4450 (free_node): Use get_stats_node_kind.
4452 2018-03-14 Richard Biener <rguenther@suse.de>
4454 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4455 that the value-set of ANTIC_IN doesn't grow.
4458 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4460 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4461 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4463 2018-03-14 Julia Koval <julia.koval@intel.com>
4465 * config.gcc (icelake-client, icelake-server): New.
4467 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4468 (initial_ix86_arch_features): Ditto.
4469 (PTA_SKYLAKE): Add SGX.
4470 (PTA_ICELAKE): Remove.
4471 (PTA_ICELAKE_CLIENT): New.
4472 (PTA_ICELAKE_SERVER): New.
4473 (ix86_option_override_internal): Split up icelake on icelake client and
4475 (get_builtin_code_for_version): Ditto.
4476 (fold_builtin_cpu): Ditto.
4477 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4478 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4479 * config/i386/i386.h (processor_type): Ditto.
4480 * doc/invoke.texi: Ditto.
4482 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4485 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4486 INTEGER_CST offset, add it together with bitpos / 8 and
4487 sign extend based on POINTER_SIZE.
4491 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4494 * config/i386/constraints.md (Yc): New register constraint.
4495 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4496 Use Yc constraint for alternative 2 of operand 0. Remove
4497 preferred_for_speed attribute.
4499 2018-03-14 Richard Biener <rguenther@suse.de>
4501 PR tree-optimization/84830
4502 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4503 with the old one to avoid oscillations.
4505 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4508 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4510 (assign_by_spills): Return a flag of reload assignment failure.
4511 Do not process the reload assignment failures. Do not spill other
4512 reload pseudos if they has the same reg class. Update n if
4514 (lra_assign): Add a return arg. Set up from the result of
4515 assign_by_spills call.
4516 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4517 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4518 usage_insns if it is not NULL.
4519 (spill_hard_reg_in_range): New function.
4520 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4521 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4522 function prototypes.
4523 (lra_assign): Change prototype.
4524 * lra.c (lra): Add code to deal with fails by splitting hard reg
4527 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4529 * config/riscv/riscv.opt (mrelax): New option.
4530 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4531 "norelax" when riscv_mrelax is disabled.
4532 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4534 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4537 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4538 reassociation for int modes.
4540 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4542 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4543 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4545 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4546 * config/aarch64/aarch64-sve.md
4547 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4548 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4549 (*extend<mode><Vwide>2): Rename to...
4550 (aarch64_sve_extend<mode><Vwide>2): ...this.
4551 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4552 renaming the old pattern to...
4553 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4555 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4556 define_expand, renaming the old pattern to...
4557 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4558 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4559 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4560 account when deciding which SVE instruction the optab should use.
4561 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4563 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4565 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4566 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4567 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4568 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4569 (tlsdesc_small_<mode>): Turn a define_expand and use
4570 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4571 (tlsdesc_small_advsimd_<mode>): ...this.
4572 (tlsdesc_small_sve_<mode>): New pattern.
4574 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4576 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4577 (UNSPEC_UMUL_HIGHPART): New constants.
4578 (MUL_HIGHPART): New int iteraor.
4579 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4580 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4582 (*<su>mul<mode>3_highpart): New define_insn.
4584 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4587 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4590 2018-03-13 Martin Liska <mliska@suse.cz>
4593 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4595 (sem_item_optimizer::~sem_item_optimizer): Release it.
4596 (sem_item_optimizer::merge_classes): Register variable aliases.
4597 (sem_item_optimizer::fixup_pt_set): New function.
4598 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4599 * ipa-icf.h: Declare new variables and functions.
4601 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4604 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4605 integer_pow2p@2 and test integer_pow2p in condition.
4606 (A < 0 ? C : 0): Similarly for @1.
4609 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4610 characters starting at p contain '\0' character, don't look beyond
4614 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4615 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4618 * reg-stack.c (change_stack): Change update_end var from int to
4619 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4620 also call set_block_for_insn on the newly added insns and rescan.
4623 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4624 on the last operand.
4627 * tree.c (stabilize_reference_1): Return save_expr (e) for
4628 STATEMENT_LIST even if it doesn't have side-effects.
4630 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4632 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4634 2018-03-12 Renlin Li <renlin.li@arm.com>
4636 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4637 aarch64_output_scalar_simd_mov_immediate.
4639 2018-03-12 Martin Sebor <msebor@redhat.com>
4641 PR tree-optimization/83456
4642 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4643 for perfectly overlapping calls to memcpy.
4644 (gimple_fold_builtin_memory_chk): Same.
4645 (gimple_fold_builtin_strcpy): Handle no-warning.
4646 (gimple_fold_builtin_stxcpy_chk): Same.
4647 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4649 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4651 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4652 parameter. Use it for SFmode.
4653 (rs6000_function_arg_advance_1): Adjust.
4654 (rs6000_function_arg): Adjust.
4655 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4657 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4659 PR rtl-optimization/84169
4660 PR rtl-optimization/84780
4661 * combine.c (can_combine_p): Check for a 2-insn combination whether
4662 the destination register is used between the two insns, too.
4664 2018-03-12 Richard Biener <rguenther@suse.de>
4666 PR tree-optimization/84803
4667 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4668 for refs DR analysis didn't process.
4670 2018-03-12 Richard Biener <rguenther@suse.de>
4672 PR tree-optimization/84777
4673 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4674 force-vectorize loops ignore whether we are optimizing for size.
4676 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4678 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4679 (TARGET_MD_ASM_ADJUST): Define.
4681 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4682 Kito Cheng <kito.cheng@gmail.com>
4683 Chung-Ju Wu <jasonwucj@gmail.com>
4685 * config/nds32/nds32.c (nds32_compute_stack_frame,
4686 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4687 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4688 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4689 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4690 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4691 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4692 * config/nds32/nds32.md (prologue, epilogue): Use macro
4693 NDS32_V3PUSH_AVAILABLE_P to do checking.
4695 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4698 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4699 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4700 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4701 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4702 addition of most attributes on !orig_type_die or the attribute not
4703 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4705 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4706 Chung-Ju Wu <jasonwucj@gmail.com>
4708 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4710 * config/nds32/nds32.opt (mvh): New option.
4712 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4713 Chung-Ju Wu <jasonwucj@gmail.com>
4715 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4717 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4718 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4721 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4722 Chung-Ju Wu <jasonwucj@gmail.com>
4724 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4726 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4727 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4729 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4730 Kito Cheng <kito.cheng@gmail.com>
4731 Chung-Ju Wu <jasonwucj@gmail.com>
4733 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4734 UNSPEC_FFMISM and UNSPEC_FLMISM.
4735 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4736 for ffb, ffmism and flmism.
4737 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4738 (unspec_ffmism): Ditto.
4739 (unspec_flmism): Ditto.
4740 (nds32_expand_builtin_impl): Check if string extension is available.
4741 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4742 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4744 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4747 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4750 * lra-assigns.c (assign_by_spills): Return a flag of reload
4751 assignment failure. Do not process the reload assignment
4752 failures. Do not spill other reload pseudos if they has the same
4754 (lra_assign): Add a return arg. Set up from the result of
4755 assign_by_spills call.
4756 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4757 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4758 usage_insns if it is not NULL.
4759 (spill_hard_reg_in_range): New function.
4760 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4761 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4762 function prototypes.
4763 (lra_assign): Change prototype.
4764 * lra.c (lra): Add code to deal with fails by splitting hard reg
4767 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4770 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4772 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4775 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4776 (dw_val_node): Add val_symbolic_view.
4777 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4778 (symview_upper_bound): New.
4779 (new_line_info_table): Initialize symviews_since_reset.
4780 (dwarf2out_source_line): Count symviews_since_reset and set
4781 symview_upper_bound.
4782 (dw_val_equal_p): Handle symview.
4783 (add_AT_symview): New.
4784 (print_dw_val): Handle symview.
4785 (attr_checksum, attr_checksum_ordered): Likewise.
4786 (same_dw_val_p, size_of_die): Likewise.
4787 (value_format, output_die): Likewise.
4788 (add_high_low_attributes): Use add_AT_symview for entry_view.
4789 (dwarf2out_finish): Reset symview_upper_bound, clear
4792 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4795 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4796 Add strict argument and use it.
4797 (rs6000_split_multireg_move): Update for new strict argument.
4798 (mem_operand_gpr): Disallow all non-offsettable addresses.
4799 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4801 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4804 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4805 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4806 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4809 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4810 decl, use remap_type if we want to use the type.
4812 2018-03-09 Martin Sebor <msebor@redhat.com>
4814 PR tree-optimization/84526
4815 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4817 (builtin_access::generic_overlap): Be prepared to handle non-array
4820 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4822 PR rtl-optimization/84682
4823 * lra-constraints.c (process_address_1): Check is_address flag
4824 for address constraints.
4825 (process_alt_operands): Likewise.
4826 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4827 preprocess_constraints.
4828 * recog.h (preprocess_constraints): Add oploc parameter.
4830 * recog.c (preprocess_constraints): Test address_operand for
4831 CT_ADDRESS constraints.
4833 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4836 * lra-assigns.c (assign_by_spills): Return a flag of reload
4837 assignment failure. Do not process the reload assignment
4838 failures. Do not spill other reload pseudos if they has the same
4840 (lra_assign): Add a return arg. Set up from the result of
4841 assign_by_spills call.
4842 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4843 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4844 usage_insns if it is not NULL.
4845 (spill_hard_reg_in_range): New function.
4846 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4847 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4848 function prototypes.
4849 (lra_assign): Change prototype.
4850 * lra.c (lra): Add code to deal with fails by splitting hard reg
4853 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4856 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4857 Accept complain bool parameter. Only emit errors if it is true.
4858 (arm_parse_cpu_option_name): Likewise.
4859 (arm_target_thumb_only): Adjust callers of the above.
4860 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4861 prototype to take a default true bool parameter.
4862 (arm_parse_arch_option_name): Likewise.
4864 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4865 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4869 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4870 * configure: Regenerate.
4871 * configure.ac ("linker --version-script option"): New.
4872 ("linker soname option"): New.
4874 2018-03-09 Richard Biener <rguenther@suse.de>
4876 PR tree-optimization/84775
4877 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4878 immediate uses of predicate stmts and mark them modified.
4881 PR tree-optimization/84178
4882 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4884 (version_loop_for_if_conversion): Delay update_ssa call.
4885 (tree_if_conversion): Delay update_ssa until after predicate
4888 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4891 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4892 when the function accesses prior frames.
4894 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4897 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4898 gen_llsym, otherwise call maybe_gen_llsym.
4901 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4902 has ',' character inside of it.
4904 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4907 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4908 as clobbering CC_REGNUM.
4910 2018-03-08 Richard Biener <rguenther@suse.de>
4913 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4914 (follow_copies_to_constant): Do not follow SSA names registered
4917 2018-03-08 Richard Biener <rguenther@suse.de>
4919 PR tree-optimization/84178
4920 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4922 (version_loop_for_if_conversion): Delay update_ssa call.
4923 (tree_if_conversion): Delay update_ssa until after predicate
4926 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4928 PR tree-optimization/84178
4929 * tree-if-conv.c (release_bb_predicate): Remove the
4930 the assertion that the stmts have NULL use_ops.
4931 Discard the statements, asserting that they haven't
4932 yet been added to a BB.
4934 2018-03-08 Richard Biener <rguenther@suse.de>
4936 PR tree-optimization/84746
4937 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4938 (phi_translate): Pass in destination ANTIC_OUT set.
4939 (phi_translate_1): Likewise. For a simplified result lookup
4940 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4941 (phi_translate_set): Adjust.
4942 (do_pre_regular_insertion): Likewise.
4943 (do_pre_partial_partial_insertion): Likewise.
4945 2018-03-08 Martin Liska <mliska@suse.cz>
4947 PR gcov-profile/84735
4948 * doc/gcov.texi: Document usage of profile files.
4949 * gcov-io.h: Document changes in the format.
4951 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4955 * dwarf2out.c (struct dw_line_info_table): Update comments for
4957 (FORCE_RESET_NEXT_VIEW): New.
4958 (FORCE_RESETTING_VIEW_P): New.
4959 (RESETTING_VIEW_P): Check for -1 too.
4960 (ZERO_VIEW_P): Likewise.
4961 (new_line_info_table): Force-reset next view.
4962 (dwarf2out_begin_function): Likewise.
4963 (dwarf2out_source_line): Simplify zero_view_p initialization.
4964 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4965 view directly. Omit view when omitting .loc at line 0.
4967 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4969 PR tree-optimization/84740
4970 * tree-switch-conversion.c (process_switch): Call build_constructors
4971 only if info.phi_count is non-zero.
4973 PR tree-optimization/84739
4974 * tree-tailcall.c (find_tail_calls): Check call arguments against
4975 DECL_ARGUMENTS (current_function_decl) rather than
4976 DECL_ARGUMENTS (func) when checking for tail recursion.
4978 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4980 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4981 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4982 Volker Reichelt's entry and add entries for people that perform
4983 GCC fuzzy testing and report numerous bugs.
4985 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4988 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4989 readonly data in sdata, if that is disabled.
4990 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4991 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4992 -mreadonly-in-sdata option.
4994 2018-03-07 Martin Sebor <msebor@redhat.com>
4996 PR tree-optimization/84468
4997 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4998 basic block when looking for nul assignment.
5000 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5003 * except.h (output_function_exception_table): Adjust prototype.
5004 * except.c (output_function_exception_table): Remove FNNAME parameter
5005 and add SECTION parameter. Ouput one part of the table at a time.
5006 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5007 the first part of the exception table and emit unwind directives.
5008 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5009 (i386_pe_seh_cold_init): Likewise.
5010 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5011 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5012 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5013 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5014 * config/i386/winnt.c: Include except.h.
5015 (struct seh_frame_state): Add reg_offset, after_prologue and
5016 in_cold_section fields.
5017 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5018 (i386_pe_seh_cold_init): New function.
5019 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5020 to seh->in_cold_section.
5021 (seh_emit_push): Record the offset of the push.
5022 (seh_emit_save): Record the offet of the save.
5023 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5024 Test seh->after_prologue to disregard the epilogue.
5025 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5026 (i386_pe_end_cold_function): New function.
5028 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5031 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5032 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5035 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5037 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5038 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5041 * multiple_target.c: Include tree-inline.h and intl.h.
5042 (expand_target_clones): Diagnose and fail if node->definition and
5043 !tree_versionable_function_p (node->decl).
5045 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5047 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5049 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5050 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5051 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5053 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5056 * combine.c (try_combine): Use reg_or_subregno instead of handling
5057 just paradoxical SUBREGs and REGs.
5059 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5061 * config/arc/arc.c (arc_finalize_pic): Remove function.
5062 (arc_must_save_register): We use single base PIC register, remove
5063 checks to save/restore the PIC register.
5064 (arc_expand_prologue): Likewise.
5065 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5067 (arc_verify_short): Likewise.
5068 (arc_attr_type): Likewise.
5069 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5070 (walk_stores): Likewise.
5071 (arc_address_cost): Make it static.
5072 (arc_verify_short): Likewise.
5073 (branch_dest): Likewise.
5074 (arc_attr_type): Likewise.
5075 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5076 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5077 (arc_final_prescan_insn): Remove inserting the nops due to
5078 hardware hazards. It is done in reorg step.
5079 (insn_length_variant_t): Remove.
5080 (insn_length_parameters_t): Likewise.
5081 (arc_insn_length_parameters): Likewise.
5082 (arc_get_insn_variants): Likewise.
5083 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5085 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5088 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5091 PR tree-optimization/84687
5092 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5094 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5096 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5098 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5099 Rename to ppc_speculation_barrier.
5100 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5101 __builtin_ppc_speculation_barrier.
5103 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5106 * combine.c (combine_simplify_rtx): Don't try to simplify if
5107 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5110 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5112 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5113 to 32 bytes when compiling for POWER9.
5115 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5118 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5119 regparm >= 3 with no arg reg available also for calls with
5120 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5123 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5125 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5127 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5130 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5132 2018-03-05 Richard Biener <rguenther@suse.de>
5134 PR tree-optimization/84486
5135 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5136 When inserting a __builtin_assume_aligned call set the LHS
5137 SSA name alignment info accordingly.
5139 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5141 PR tree-optimization/84114
5142 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5143 Avoid reassociation of FLOAT_MODE addition.
5145 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5147 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5148 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5149 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5150 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5151 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5152 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5153 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5155 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5156 __builtin_ia32_wbinvd): New builtins.
5157 (SPECIAL_ARGS2): New.
5158 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5159 (SPECIAL_ARGS2): New.
5160 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5161 (ix86_valid_target_attribute_inner_p): Ditto.
5162 (ix86_init_mmx_sse_builtins): Add special_args2.
5163 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5164 TARGET_WBNOINVD_P): New.
5165 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5166 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5167 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5168 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5169 * config/i386/pconfigintrin.h: New file.
5170 * config/i386/wbnoinvdintrin.h: Ditto.
5171 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5172 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5174 2018-03-05 Richard Biener <rguenther@suse.de>
5176 PR tree-optimization/84670
5177 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5179 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5180 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5181 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5182 doesn't grow if all successors (recursively) were visited at least
5185 2018-03-05 Richard Biener <rguenther@suse.de>
5187 PR tree-optimization/84650
5188 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5189 if executed in the loop pipeline.
5191 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5193 * doc/configfiles.texi (Configuration Files): Move info about
5194 conditionalizing $target-protos.h to...
5195 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5196 differs from $target-protos.h.
5198 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5199 Chung-Ju Wu <jasonwucj@gmail.com>
5201 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5202 * config/nds32/nds32-multiple.md (setmemsi): Define.
5203 * config/nds32/nds32-memory-manipulation.c
5204 (nds32_gen_dup_4_byte_to_word_value): New.
5205 (emit_setmem_word_loop): New.
5206 (emit_setmem_byte_loop): New.
5207 (nds32_expand_setmem_loop): New.
5208 (nds32_expand_setmem_loop_v3m): New.
5209 (nds32_expand_setmem_unroll): New.
5210 (nds32_expand_setmem): New.
5212 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5213 Chung-Ju Wu <jasonwucj@gmail.com>
5215 * config/nds32/nds32-memory-manipulation.c
5216 (nds32_emit_load_store): New.
5217 (nds32_emit_post_inc_load_store): New.
5218 (nds32_emit_mem_move): New.
5219 (nds32_emit_mem_move_block): New.
5220 (nds32_expand_movmemsi_loop_unknown_size): New.
5221 (nds32_expand_movmemsi_loop_known_size): New.
5222 (nds32_expand_movmemsi_loop): New.
5223 (nds32_expand_movmemsi_unroll): New.
5224 (nds32_expand_movmemqi): Rename ...
5225 (nds32_expand_movmemsi): ... to this.
5226 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5227 (movmemsi): ... to this.
5228 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5229 (nds32_expand_movmemsi): ... to this.
5231 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5232 Monk Chiang <sh.chiang04@gmail.com>
5233 Chung-Ju Wu <jasonwucj@gmail.com>
5235 * config/nds32/nds32-protos.h
5236 (nds32_expand_load_multiple): New arguments.
5237 (nds32_expand_store_multiple): Ditto.
5238 (nds32_valid_multiple_load_store): Rename ...
5239 (nds32_valid_multiple_load_store_p): ... to this.
5240 * config/nds32/nds32-memory-manipulation.c
5241 (nds32_expand_load_multiple): Refine implementation.
5242 (nds32_expand_store_multiple): Ditto.
5243 * config/nds32/nds32-multiple.md
5244 (load_multiple): Update nds32_expand_load_multiple interface.
5245 (store_multiple): Update nds32_expand_store_multiple interface.
5246 * config/nds32/nds32-predicates.c
5247 (nds32_valid_multiple_load_store): Rename ...
5248 (nds32_valid_multiple_load_store_p): ... to this and refine
5250 * config/nds32/predicates.md
5251 (nds32_load_multiple_and_update_address_operation): New predicate.
5252 (nds32_store_multiple_and_update_address_operation): New predicate.
5254 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5255 Chung-Ju Wu <jasonwucj@gmail.com>
5257 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5258 (combo): New attribute.
5259 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5261 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5263 * config/nds32/nds32.opt: Change -mcmodel= default value.
5265 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5266 Monk Chiang <sh.chiang04@gmail.com>
5267 Chung-Ju Wu <jasonwucj@gmail.com>
5269 * config/nds32/constants.md (unspec_element): New enum.
5270 * config/nds32/constraints.md (Umw): New constraint.
5271 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5272 * config/nds32/nds32-intrinsic.md: Likewise.
5273 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5274 (nds32_valid_smw_lwm_base_p): New.
5275 (nds32_output_smw_single_word): New.
5276 (nds32_output_lmw_single_word): New.
5277 (nds32_expand_unaligned_load): New.
5278 (nds32_expand_unaligned_store): New.
5279 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5280 (nds32_output_smw_single_word): Declare.
5281 (nds32_output_lmw_single_word): Declare.
5282 (nds32_expand_unaligned_load): Declare.
5283 (nds32_expand_unaligned_store): Declare.
5284 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5285 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5286 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5287 NDS32_BUILTIN_UASTORE_DW.
5288 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5291 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5292 Kito Cheng <kito.cheng@gmail.com>
5293 Chung-Ju Wu <jasonwucj@gmail.com>
5295 * config/nds32/nds32-intrinsic.c
5296 (nds32_expand_builtin_null_ftype_reg): Delete.
5297 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5298 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5299 (nds32_read_argument): New.
5300 (nds32_legitimize_target): Ditto.
5301 (nds32_legitimize_argument): Ditto.
5302 (nds32_check_constant_argument): Ditto.
5303 (nds32_expand_unop_builtin): Ditto.
5304 (nds32_expand_unopimm_builtin): Ditto.
5305 (nds32_expand_binop_builtin): Ditto.
5306 (nds32_builtin_decl_impl): Ditto.
5307 (builtin_description): Ditto.
5308 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5309 (nds32_init_builtins_impl): Ditto.
5310 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5311 (nds32_builtin_decl): New.
5312 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5313 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5315 2018-03-02 Jeff Law <law@redhat.com>
5317 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5318 (redundant_insn, fill_simple_delay_slots): Likewise.
5319 (fill_slots_from_thread): Likewise.
5320 * resource.c (mark_referenced_resources): Likewise.
5321 (mark_set_resources, find_dead_or_set_registers): Likewise.
5323 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5325 * substring-locations.h (format_warning_va): Formatting fix for
5327 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5329 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5330 * substring-locations.c: Include intl.h.
5331 (format_warning_va): Turned into small wrapper around
5332 format_warning_n_va, renamed to ...
5333 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5334 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5336 (format_warning_at_substring_n): New function.
5337 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5338 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5339 format_warning_at_substring with just a shorter name instead of
5340 const function pointer.
5341 (fmtwarn_n): New function.
5342 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5343 appropriate, get rid of all the fmtstr temporaries, move conditionals
5344 with G_() wrapped string literals directly into fmtwarn arguments,
5345 cast dir.len to (int), formatting fixes.
5347 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5349 * doc/invoke.texi: Remove "Cilk Plus" references.
5351 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5352 Richard Biener <rguenther@suse.de>
5355 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5356 for error or warning attributes if CALL_FROM_THUNK_P is set.
5359 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5362 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5363 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5366 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5367 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5369 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5370 instead of -1U in last predictors element's probability member.
5372 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5375 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5376 arguments if they are comparable.
5378 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5380 PR tree-optimization/84634
5381 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5382 masks and masked_loop_p with a single loop_masks, making sure it's
5383 null for bb vectorization.
5385 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5387 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5388 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5389 loop->force_vectorize to check whether there is no alias.
5391 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5394 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5396 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5398 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5399 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5400 instead of a loop around prev_real_insn.
5401 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5405 * config/i386/i386.c (ix86_print_operand): Use conditional
5406 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5409 2018-03-02 Richard Biener <rguenther@suse.de>
5411 PR tree-optimization/84427
5412 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5413 (bitmap_set_subtract_values): Rewrite to handle multiple
5416 (prune_clobbered_mems): Likewise.
5417 (phi_translate): Take edge instead of pred/phiblock.
5418 (phi_translate_1): Likewise.
5419 (phi_translate_set): Likewise. Insert all translated
5420 exprs for a value into the set, keeping possibly multiple
5421 expressions per value.
5422 (compute_antic_aux): Adjust for phi_translate changes.
5423 When intersecting union the expressions and prune those
5424 not in the final value set, keeping possibly multiple
5425 expressions per value. Do not use value-insertion
5426 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5427 all expressions. Add verification that the value-sets
5428 only shrink during iteration.
5429 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5430 (do_pre_regular_insertion): Likewise.
5431 (do_pre_partial_partial_insertion): Likewise.
5433 2018-03-02 Richard Biener <rguenther@suse.de>
5436 * config/darwin.c (saved_debug_info_level): New static global.
5437 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5438 (darwin_asm_lto_end): Restore debug info generation settings.
5440 2018-03-01 Martin Liska <mliska@suse.cz>
5443 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5446 2018-03-01 Richard Biener <rguenther@suse.de>
5449 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5450 type annotation with LTO when debug was disabled at compile-time.
5452 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5454 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5456 (mips_final_postscan_insn): Likewise.
5458 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5460 PR rtl-optimization/84528
5461 * alias.c (init_alias_target): Add commentary.
5462 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5463 a unique base value if the frame pointer is not eliminated
5464 to the stack pointer.
5466 2018-03-01 Tom de Vries <tom@codesourcery.com>
5468 PR rtl-optimization/83327
5469 * lra-int.h (hard_regs_spilled_into): Declare.
5470 * lra.c (hard_regs_spilled_into): Define.
5471 (init_reg_info): Init hard_regs_spilled_into.
5472 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5473 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5474 (process_bb_lives): Handle hard_regs_spilled_into.
5475 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5476 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5478 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5480 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5481 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5482 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5483 * config/rs6000/aix72.h: New file.
5485 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5487 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5488 instead of warning_at with conditional singular and plural messages
5492 * stor-layout.c (update_alignment_for_field): For
5493 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5494 && !DECL_PACKED (field), do the alignment update, just use
5495 only desired_align instead of MAX (type_align, desired_align)
5497 (place_field): Don't do known_align < desired_align handling
5498 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5499 is non-NULL, instead do it after rli->prev_field handling and
5500 only if not within a bitfield word. For DECL_PACKED (field)
5501 use type_align of BITS_PER_UNIT.
5503 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5505 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5506 superfluous parentheses and trailing spaces.
5508 2018-02-28 Richard Biener <rguenther@suse.de>
5510 PR tree-optimization/84584
5511 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5512 SCoPs with fake exit edge.
5514 2018-02-28 Martin Liska <mliska@suse.cz>
5517 * timevar.c (timer::print): Fix format to properly print 100%
5520 2018-02-28 Richard Biener <rguenther@suse.de>
5523 * genmatch.c (capture_info::walk_match): Do not mark
5524 captured expressions without operands as expr_p given
5525 they act more like predicates and should be subject to
5526 "lost tail" side-effect preserving.
5528 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5530 PR rtl-optimization/81611
5531 * auto-inc-dec.c (attempt_change): Move dead note from
5532 mem_insn if it's the next use of regno
5533 (find_address): Take address use of reg holding
5534 non-incremented value. Add parm to limit search to the named
5536 (merge_in_block): Attempt to use a mem insn that is the next
5537 use of the original regno.
5539 2018-02-27 Martin Sebor <msebor@redhat.com>
5542 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5543 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5545 2018-02-27 Martin Sebor <msebor@redhat.com>
5547 PR translation/84207
5548 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5549 n argument to unsigned HOST_WIDE_INT.
5550 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5551 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5552 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5553 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5555 2018-02-27 Richard Biener <rguenther@suse.de>
5557 PR tree-optimization/84512
5558 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5559 Do not use the estimate returned from record_stmt_cost for
5560 the scalar iteration cost but sum properly using add_stmt_cost.
5562 2018-02-27 Richard Biener <rguenther@suse.de>
5564 PR tree-optimization/84466
5565 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5566 Adjust last change to less strictly validate use operands.
5568 2018-02-27 Martin Liska <mliska@suse.cz>
5570 PR gcov-profile/84548
5571 * gcov.c (process_file): Allow partial overlap and consider it
5572 also as group functions.
5573 (output_lines): Properly calculate range of lines for a group.
5575 2018-02-27 Martin Liska <mliska@suse.cz>
5577 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5578 'ggc' suffixes. Change first column width.
5579 (timer::print): Fix formatting of the column.
5581 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5583 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5584 preserve inline entry blocks for the sake of debug inline
5585 entry point markers alone.
5586 (remove_unused_locals): Suggest in comments a better place to
5587 force the preservation of inline entry blocks that are
5588 otherwise unused, but do not preserve them.
5590 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5592 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5594 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5597 * config/i386/constraints.md (Bs): Replace
5598 ix86_indirect_branch_register with
5599 TARGET_INDIRECT_BRANCH_REGISTER.
5601 * config/i386/i386.md (indirect_jump): Likewise.
5602 (tablejump): Likewise.
5603 (*sibcall_memory): Likewise.
5604 (*sibcall_value_memory): Likewise.
5605 Peepholes of indirect call and jump via memory: Likewise.
5606 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5607 (*sibcall_value_GOT_32): Likewise.
5608 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5609 (GOT_memory_operand): Likewise.
5610 (call_insn_operand): Likewise.
5611 (sibcall_insn_operand): Likewise.
5612 (GOT32_symbol_operand): Likewise.
5613 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5615 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5617 PR rtl-optimization/83496
5618 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5619 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5620 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5621 redundant insn, if any.
5622 (relax_delay_slots): Likewise.
5623 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5625 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5627 PR tree-optimization/83965
5628 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5629 that grouped statements are part of a reduction chain. Return
5630 true if the statement is not marked as a reduction itself but
5632 (vect_recog_dot_prod_pattern): Don't check whether the statement
5633 is part of a group here.
5634 (vect_recog_sad_pattern): Likewise.
5635 (vect_recog_widen_sum_pattern): Likewise.
5637 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5640 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5642 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5645 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5647 (ix86_output_indirect_function_return): New prototype.
5648 (ix86_split_simple_return_pop_internal): Likewise.
5649 * config/i386/i386.c (indirect_return_via_cx): New.
5650 (indirect_return_via_cx_bnd): Likewise.
5651 (indirect_thunk_name): Handle return va CX_REG.
5652 (output_indirect_thunk_function): Create alias for
5653 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5654 (ix86_output_indirect_jmp): Remove the bool argument.
5655 (ix86_output_indirect_function_return): New function.
5656 (ix86_split_simple_return_pop_internal): Likewise.
5657 * config/i386/i386.md (*indirect_jump): Don't pass false
5658 to ix86_output_indirect_jmp.
5659 (*tablejump_1): Likewise.
5660 (simple_return_pop_internal): Change it to define_insn_and_split.
5661 Call ix86_split_simple_return_pop_internal to split it for
5663 (simple_return_indirect_internal): Call
5664 ix86_output_indirect_function_return instead of
5665 ix86_output_indirect_jmp.
5667 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5670 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5671 memset and value initialization afterwards.
5673 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5675 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5677 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5680 * common/config/aarch64/aarch64-common.c
5681 (aarch_option_optimization_table[]): Switch
5682 off fomit-frame-pointer
5684 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5685 Chung-Ju Wu <jasonwucj@gmail.com>
5687 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5689 (store_multiple): Ditto.
5691 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5693 * config.gcc: Add --with-cpu support for nds32 target.
5694 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5695 * config/nds32/nds32.opt: Add -mcpu= option.
5697 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5699 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5700 isel=yes): Warn for these deprecated options.
5702 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5704 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5705 ISA_2_5_MASKS_EMBEDDED.
5707 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5709 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5710 p->max as pointers rather than using iterative_hash_expr.
5712 2018-02-23 Carl Love <cel@us.ibm.com>
5714 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5715 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5717 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5718 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5719 P8V_BUILTIN_VEC_VUNSIGNED2.
5721 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5724 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5725 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5727 * lra-constraints.c (curr_insn_transform): Set up
5728 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5730 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5731 * lra-eliminations.c (spill_pseudos): Ditto.
5732 (process_insn_for_elimination): Ditto.
5733 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5734 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5735 LRA_NON_CLOBBERED_ALT.
5737 2018-02-22 Martin Sebor <msebor@redhat.com>
5739 PR tree-optimization/84480
5740 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5741 to maybe_diag_stxncpy_trunc. Call it.
5742 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5743 from gimple_fold_builtin_strcpy. Print inlining stack.
5744 (handle_builtin_stxncpy): Print inlining stack.
5745 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5747 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5750 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5751 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5752 and -fcheck-pointer-bounds are used together.
5753 (indirect_thunk_prefix): New enum.
5754 (indirect_thunk_need_prefix): New function.
5755 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5756 "_nt" instead of "_bnd" for NOTRACK prefix.
5757 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5758 (output_indirect_thunk_function): Likewise.
5760 (ix86_code_end): Update output_indirect_thunk_function calls.
5761 (ix86_output_indirect_branch_via_reg): Replace
5762 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5763 (ix86_output_indirect_branch_via_push): Likewise.
5764 (ix86_output_function_return): Likewise.
5765 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5766 incompatible with -fcf-protection=branch and
5767 -fcheck-pointer-bounds.
5769 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5772 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5773 Change gcc_assert call to output_operand_lossage.
5775 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5777 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5779 2018-02-22 DJ Delorie <dj@redhat.com>
5780 Sebastian Perta <sebastian.perta@renesas.com>
5781 Oleg Endo <olegendo@gcc.gnu.org>
5783 * config/rx/rx.c (rx_rtx_costs): New function.
5784 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5786 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5788 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5790 2018-02-22 Martin Liska <mliska@suse.cz>
5793 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5794 Add "native" as a possible value.
5796 2018-02-22 Martin Liska <mliska@suse.cz>
5799 * config/i386/i386.c (ix86_option_override_internal):
5800 Add "native" as a possible value for -march and -mtune.
5802 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5805 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5806 to all type variants.
5808 PR tree-optimization/84503
5809 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5810 width as info->bitpos + info->bitsize - start.
5811 (merged_store_group::merge_overlapping): Simplify width computation.
5812 (check_no_overlap): New function.
5813 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5814 start + width and last_order of the group, fail if check_no_overlap
5816 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5817 to group if check_no_overlap fails.
5819 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5821 * config/rs6000/altivec.md: Delete contraint arguments to
5822 define_expand, define_split, and define_peephole2, and in
5823 define_insn_and_split if always unused.
5824 * config/rs6000/darwin.md: Ditto.
5825 * config/rs6000/dfp.md: Ditto.
5826 * config/rs6000/rs6000.md: Ditto.
5827 * config/rs6000/sync.md: Ditto.
5828 * config/rs6000/vector.md: Ditto.
5829 * config/rs6000/vsx.md: Ditto.
5831 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5833 * config/rs6000/altivec.md: Write output control strings as braced
5834 blocks instead of double-quoted strings.
5835 * config/rs6000/darwin.md: Ditto.
5836 * config/rs6000/rs6000.md: Ditto.
5837 * config/rs6000/vector.md: Ditto.
5838 * config/rs6000/vsx.md: Ditto.
5840 2018-02-21 Jason Merrill <jason@redhat.com>
5842 PR c++/84314 - ICE with templates and fastcall attribute.
5843 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5845 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5847 * ipa-cp.c (determine_versionability): Fix comment typos.
5849 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5852 * ipa-cp.c (determine_versionability): Do not version functions caling
5855 2018-02-21 Martin Liska <mliska@suse.cz>
5858 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5859 Add "native" as a possible value.
5860 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5861 the macro when native cpu detection is available.
5863 2018-02-21 Martin Liska <mliska@suse.cz>
5866 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5867 Add "native" as a possible value.
5868 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5869 when native cpu detection is available.
5871 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5872 Martin Sebor <msebor@redhat.com>
5874 PR tree-optimization/84478
5875 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5877 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5878 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5879 support which is conservatively correct, for 2 only stay conservative
5880 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5881 argument to the 2 argument get_range_strlen, adjust 6 arg
5882 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5884 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5885 (gimple_fold_builtin_strlen): Pass true as last argument to
5888 2018-02-20 Martin Sebor <msebor@redhat.com>
5891 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5892 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5893 (builtin_memref::builtin_memref): Factor out parts into
5894 set_base_and_offset and call it.
5896 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5899 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5900 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5901 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5902 search at the associated MODE_INT.
5904 2018-02-20 Jeff Law <law@redhat.com>
5907 PR tree-optimization/81592
5909 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5910 for range data rather than using global data.
5911 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5912 range data rather than using global data.
5913 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5914 pass it to children as needed.
5915 (struct directive::fmtresult): Similarly.
5916 (struct directive::set_width): Similarly.
5917 (struct directive::set_precision): Similarly.
5918 (format_integer, format_directive, parse_directive): Similarly.
5919 (format_none): Accept unnamed vr_values parameter.
5920 (format_percent, format_floating, format_character): Similarly.
5921 (format_string, format_plain): Similarly.
5922 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5923 the EVRP range analyzer for range data rather than using global data.
5924 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5925 gimple-ssa-evrp-analyze.h
5926 (class sprintf_dom_walker): Add after_dom_children member function.
5927 Add evrp_range_analyzer member.
5928 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5929 range analyzer as needed.
5930 (sprintf_dom_walker::after_dom_children): New member function.
5931 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5933 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5934 (evrp_range_analyzer::pop_to_marker): Likewise.
5936 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5938 PR tree-optimization/84419
5939 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5940 with the required type if its current type is compatible but
5943 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5946 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5947 after vectorization.
5949 2018-02-20 Martin Liska <mliska@suse.cz>
5952 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5953 possible values if we don't have a hint.
5955 2018-02-20 Martin Liska <mliska@suse.cz>
5959 * final.c (shorten_branches): Build align_tab array with one
5961 * opts.c (finish_options): Add alignment option limit check.
5962 (MAX_CODE_ALIGN): Likewise.
5963 (MAX_CODE_ALIGN_VALUE): Likewise.
5964 * doc/invoke.texi: Document maximum allowed option value for
5965 all -falign-* options.
5967 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5970 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5971 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5972 * var-tracking.c (emit_note_insn_var_location): Remove all references
5973 to NOTE_INSN_CALL_ARG_LOCATION.
5974 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5975 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5976 Use copy_rtx_if_shared.
5977 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5978 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5979 (dwarf2out_var_location): Remove handling of
5980 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5982 * final.c (final_scan_insn): Remove all references to
5983 NOTE_INSN_CALL_ARG_LOCATION.
5984 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5985 before dumping final insns.
5986 * except.c (emit_note_eh_region_end): Remove all references to
5987 NOTE_INSN_CALL_ARG_LOCATION.
5988 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5989 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5990 * config/arc/arc.c (hwloop_optimize): Likewise.
5991 * config/arm/arm.c (create_fix_barrier): Likewise.
5992 * config/s390/s390.c (s390_chunkify_start): Likewise.
5993 * config/sh/sh.c (find_barrier): Likewise.
5994 * config/i386/i386.c (rest_of_insert_endbranch,
5995 ix86_seh_fixup_eh_fallthru): Likewise.
5996 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5997 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5998 * config/frv/frv.c (frv_function_prologue): Likewise.
5999 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6001 (note_outside_basic_block_p): Remove all references to
6002 NOTE_INSN_CALL_ARG_LOCATION.
6003 * gengtype.c (adjust_field_rtx_def): Likewise.
6004 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6006 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6007 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6010 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6013 PR tree-optimization/84452
6014 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6015 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6018 2018-02-19 Martin Liska <mliska@suse.cz>
6021 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6023 2018-02-19 Martin Liska <mliska@suse.cz>
6024 Richard Sandiford <richard.sandiford@linaro.org>
6026 PR tree-optimization/82491
6027 * gimple-fold.c (get_base_constructor): Make earlier bail out
6030 2018-02-19 Carl Love <cel@us.ibm.com>
6032 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6033 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6035 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6036 P8V_BUILTIN_VEC_NEG.
6038 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6040 * config/rl78/rl78.md (movdf): New define expand.
6042 2018-02-19 Martin Liska <mliska@suse.cz>
6045 * doc/invoke.texi: Fix typo.
6046 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6048 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6050 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6051 handle rs6000_single_float and rs6000_double_float specially for
6054 2018-02-16 Jeff Law <law@redhat.com>
6056 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6059 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6062 * ipa-inline.c (inline_small_functions): Fix a typo.
6064 2018-02-16 Nathan Sidwell <nathan@acm.org>
6066 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6068 2018-02-16 Carl Love <cel@us.ibm.com>
6070 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6071 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6072 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6073 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6074 expansion to P8V_BUILTIN_VEC_FLOAT2.
6076 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6078 PR rtl-optimization/70023
6079 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6080 src_regno into account.
6082 2018-02-16 Carl Love <cel@us.ibm.com>
6084 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6085 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6086 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6087 * config/rs6000/rs6000.c: Remove case statements for
6088 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6089 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6090 and P9V_BUILTIN_VEC_VINSERT4B.
6091 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6092 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6093 * config/rs6000/vsx.md:
6094 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6097 2018-02-16 Carl Love <cel@us.ibm.com>
6099 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6101 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6103 * config/rs6000/rs6000-c.c: Add the definitions for
6104 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6105 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6106 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6107 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6108 definition for insert4b and define insn *insert3b_internal.
6109 * doc/extend.texi: Add documentation for vec_extract4b.
6111 2018-02-16 Nathan Sidwell <nathan@acm.org>
6113 * doc/extend.texi (Backwards Compatibility): Mention friend
6114 injection. Note for-scope is deprecated.
6115 * doc/invoke.texi (-ffriend-injection): Deprecate.
6117 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6119 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6120 that moved to I2, also allow destinations that are a paradoxical
6121 subreg (instead of a normal reg).
6123 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6126 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6129 2018-02-16 Richard Biener <rguenther@suse.de>
6131 PR tree-optimization/84037
6132 PR tree-optimization/84016
6134 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6135 Adjust vec_construct for the fact we need additional higher latency
6136 128bit inserts for AVX256 and AVX512 vector builds.
6137 (ix86_add_stmt_cost): Scale vector construction cost for
6140 2018-02-16 Richard Biener <rguenther@suse.de>
6142 PR tree-optimization/84417
6143 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6144 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6145 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6147 2018-02-16 Martin Liska <mliska@suse.cz>
6150 * internal-fn.def (ASAN_CHECK): Set proper flags.
6151 (ASAN_MARK): Likewise.
6153 2018-02-16 Julia Koval <julia.koval@intel.com>
6155 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6156 from PTA_CANNONLAKE.
6158 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6161 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6162 Use ++iter rather than iter++ for std::list iterators.
6163 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6164 defer deleting them until all nodes in the forest are processed. Do
6165 free even leaf nodes. Change to_process into auto_vec.
6168 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6169 * vec.h (vec_default_construct): Use memset instead of placement new
6170 if BROKEN_VALUE_INITIALIZATION is defined.
6171 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6172 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6175 PR rtl-optimization/83723
6176 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6177 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6178 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6180 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6182 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6184 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6186 PR rtl-optimization/81443
6187 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6188 from inner REGs to paradoxical SUBREGs.
6190 2018-02-16 Richard Biener <rguenther@suse.de>
6192 PR tree-optimization/84399
6193 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6194 For operands we can analyze at their definition make sure we can
6195 analyze them at each use as well.
6197 2018-02-16 Richard Biener <rguenther@suse.de>
6199 PR tree-optimization/84190
6200 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6201 volatile accesses if the decl isn't volatile.
6203 2018-02-15 Jason Merrill <jason@redhat.com>
6205 PR c++/84314 - ICE with templates and fastcall attribute.
6206 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6207 TYPE_CANONICAL on an existing type.
6209 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6211 PR tree-optimization/84383
6212 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6213 dstoff nor call operand_equal_p if dstbase is NULL.
6215 PR tree-optimization/84334
6216 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6217 also a CONSTANT_CLASS_P, punt.
6219 2018-02-14 Jim Wilson <jimw@sifive.com>
6221 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6222 first SMALL_OPERAND check. New local min_second_step. Move assert
6223 to where locals are set. Add TARGET_RVC support.
6224 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6226 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6228 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6230 2018-02-14 Martin Sebor <msebor@redhat.com>
6232 PR tree-optimization/83698
6233 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6234 arrays constrain the offset range to their bounds.
6235 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6236 (builtin_access::overlap): Avoid setting the size of overlap if it's
6238 (maybe_diag_overlap): Also consider arrays when deciding what values
6239 of offsets to include in diagnostics.
6241 2018-02-14 Martin Sebor <msebor@redhat.com>
6244 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6245 that correspond to the kind of a declaration.
6247 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6250 * config/pa/pa.md: Load address of PIC label using the linkage table
6251 if the label is nonlocal.
6253 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6255 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6256 warning message if user requests -maltivec=be.
6257 * doc/invoke.texi: Document deprecation of -maltivec=be.
6259 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6262 * config/rs6000/rs6000-c.c: Update definitions for
6263 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6264 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6266 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6269 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6270 add _get_ssp intrinsics. Remove argument from
6271 __builtin_ia32_rdssp[d|q].
6272 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6273 * config/i386/i386-builtin.def: Remove argument from
6274 __builtin_ia32_rdssp[d|q].
6275 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6276 ix86_expand_special_args_builtin for _rdssp[d|q].
6277 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6278 Clear register before usage.
6279 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6280 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6282 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6284 PR tree-optimization/84357
6285 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6286 operand 1 of an ARRAY_REF too.
6288 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6291 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6292 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6294 (set_of_reg): New struct.
6295 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6296 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6297 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6299 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6300 Split into bitclr, bitset, bitinvert patterns if appropriate.
6301 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6302 use rx_fuse_in_memory_bitop.
6303 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6304 to named insn, correct maximum insn length.
6306 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6309 * machmode.def: Define a complex mode for PARTIAL_INT.
6310 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6312 * doc/rtl.texi: Document CSPImode.
6313 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6315 (msp430_hard_regno_nregs_with_padding): Likewise.
6317 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6320 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6322 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6324 PR rtl-optimization/84169
6325 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6326 we generated a parallel as new i3 and we split that to new i2 and i3
6327 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6328 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6329 those to i2, not i1. Partially rewrite this scan code.
6331 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6334 * stor-layout.c (place_field): For variable length fields, adjust
6335 offset_align afterwards not just based on the field's alignment,
6336 but also on the size.
6339 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6340 of exps and logs in the use_exp2 case.
6342 2018-02-13 Jeff Law <law@redhat.com>
6344 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6347 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6350 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6354 * common.opt (gas-loc-support, gas-locview-support): New.
6355 (ginline-points, ginternal-reset-location-views): New.
6356 * doc/invoke.texi: Document them. Use @itemx where intended.
6357 (gvariable-location-views): Adjust.
6358 * target.def (reset_location_view): New.
6359 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6360 (TARGET_RESET_LOCATION_VIEW): New.
6361 * doc/tm.texi: Rebuilt.
6362 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6363 (dwarf2out_default_as_locview_support): New.
6364 (output_asm_line_debug_info): Use option variables.
6365 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6366 (output_loc_list): Likewise.
6367 (add_high_low_attributes): Check option variables.
6368 Don't output entry view attribute in strict mode.
6369 (gen_inlined_subroutine_die): Check option variables.
6370 (dwarf2out_inline_entry): Likewise.
6371 (init_sections_and_labels): Likewise.
6372 (dwarf2out_early_finish): Likewise.
6373 (maybe_reset_location_view): New, from...
6374 (dwarf2out_var_location): ... here. Call it.
6375 * debug.h (dwarf2out_default_as_loc_support): Declare.
6376 (dwarf2out_default_as_locview_support): Declare.
6377 * hooks.c (hook_int_rtx_insn_0): New.
6378 * hooks.h (hook_int_rtx_insn_0): Declare.
6379 * toplev.c (process_options): Take -gas-loc-support and
6380 -gas-locview-support from dwarf2out. Enable
6381 -gvariable-location-views by default only with locview
6382 assembler support. Enable -ginternal-reset-location-views by
6383 default only if the target defines the corresponding hook.
6384 Enable -ginline-points by default if location views are
6385 enabled; force it disabled if statement frontiers are
6387 * tree-inline.c (expand_call_inline): Check option variables.
6388 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6390 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6392 PR tree-optimization/84321
6393 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6394 handling. Also check whether the anti-range contains any values
6395 that satisfy the mask; switch to a VR_RANGE if not.
6397 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6400 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6402 2018-02-13 Martin Jambor <mjambor@suse.cz>
6405 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6406 of call statements, also set location of a load to a temporary.
6408 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6410 * config/rl78/rl78.c (add_vector_labels): New function.
6411 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6412 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6413 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6414 which checks that no arguments are passed.
6415 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6416 * doc/extend.texi: Documentation for the new attribute.
6418 2018-02-13 Andreas Schwab <schwab@suse.de>
6420 * config/riscv/linux.h (CPP_SPEC): Define.
6422 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6425 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6426 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6427 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6428 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6429 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6430 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6431 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6432 temporarily for AES and PCLMUL builtins.
6434 PR tree-optimization/84339
6435 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6436 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6440 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6441 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6442 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6444 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6446 * omp-simd-clone.h: New file.
6447 * omp-simd-clone.c: Include omp-simd-clone.h.
6448 (expand_simd_clones): No longer static.
6449 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6450 cgraph.h and omp-simd-clone.h.
6451 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6452 (vect_recog_widen_shift_pattern): Formatting fix.
6453 (vect_pattern_recog_1): Don't check optab for calls.
6456 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6457 operands[2] into a REG before using gen_lowpart on it.
6459 2018-02-12 Jeff Law <law@redhat.com>
6462 * config/sh/sh.c (find_barrier): Consider a sibling call
6465 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6466 successfully back substituting a reg.
6468 2018-02-12 Richard Biener <rguenther@suse.de>
6470 PR tree-optimization/84037
6471 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6472 parameter, move visited init to caller.
6473 (vect_slp_analyze_operations): Separate cost from validity
6474 check, initialize visited once for all instances.
6475 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6477 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6479 (vect_model_promotion_demotion_cost): Likewise.
6480 (vectorizable_bswap): Guard cost modeling with !slp_node
6481 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6483 (vectorizable_call): Likewise.
6484 (vectorizable_conversion): Likewise.
6485 (vectorizable_assignment): Likewise.
6486 (vectorizable_shift): Likewise.
6487 (vectorizable_operation): Likewise.
6488 (vectorizable_store): Likewise.
6489 (vectorizable_load): Likewise.
6490 (vectorizable_condition): Likewise.
6491 (vectorizable_comparison): Likewise.
6493 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6496 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6497 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6498 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6500 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6503 * params.def (inline-min-speedup): Increase from 8 to 15.
6504 (max-inline-insns-auto): Decrease from 40 to 30.
6505 * ipa-split.c (consider_split): Add some buffer for function to
6506 be considered inlining candidate.
6507 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6510 2018-02-12 Richard Biener <rguenther@suse.de>
6512 PR tree-optimization/84037
6513 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6514 matched stmts if we cannot swap the non-matched ones.
6516 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6518 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6519 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6520 _mm_maskz_scalef_round_ss): New intrinsics.
6521 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6522 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6523 __builtin_ia32_scalefss_round): Remove.
6524 (__builtin_ia32_scalefsd_mask_round,
6525 __builtin_ia32_scalefss_mask_round): New intrinsics.
6526 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6527 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6528 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6529 "<round_constraint>")): Changed to ...
6530 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6531 "<round_scalar_constraint>")): ... this.
6532 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6533 %0, %1, %2<round_op3>}"): Changed to ...
6534 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6535 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6536 %2<round_scalar_mask_op3>}"): ... this.
6537 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6539 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6541 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6542 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6543 (_mm_maskz_sqrt_round_ss): New intrinsics.
6544 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6545 (__builtin_ia32_sqrtsd_mask_round)
6546 (__builtin_ia32_sqrtss_mask_round): New builtins.
6547 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6548 (__builtin_ia32_sqrtss_round): Remove.
6549 (__builtin_ia32_sqrtsd_mask_round)
6550 (__builtin_ia32_sqrtss_mask_round): New builtins.
6551 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6552 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6553 ((match_operand:VF_128 1 "vector_operand"
6554 "xBm,<round_constraint>")): Changed to ...
6555 ((match_operand:VF_128 1 "vector_operand"
6556 "xBm,<round_scalar_constraint>")): ... this.
6557 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6558 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6559 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6560 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6561 %<iptr>1<round_scalar_mask_op3>}): ... this.
6562 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6563 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6565 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6568 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6569 Cast vec_cmpeq result to correct type.
6570 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6571 Cast vec_cmpgt result to correct type.
6573 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6575 * final.c (final_scan_insn_1): Renamed from...
6576 (final_scan_insn): ... this. New wrapper, to recover
6577 seen from the outermost call in recursive ones.
6578 * config/sparc/sparc.c (output_return): Drop seen from call.
6579 (output_sibcall): Likewise.
6580 * config/visium/visium.c (output_branch): Likewise.
6582 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6584 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6587 2018-02-10 Alan Modra <amodra@gmail.com>
6590 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6591 Specify LR as an input.
6593 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6596 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6597 remove_member_access_dummy_vars): New functions.
6598 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6599 lower_omp_1, execute_lower_omp): Use them.
6601 PR rtl-optimization/84308
6602 * shrink-wrap.c (spread_components): Release todo vector.
6604 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6606 PR rtl-optimization/57193
6607 * ira-color.c (struct allocno_color_data): Add member
6608 conflict_allocno_hard_prefs.
6609 (update_conflict_allocno_hard_prefs): New.
6610 (bucket_allocno_compare_func): Add a preference based on
6611 conflict_allocno_hard_prefs.
6612 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6613 (color_allocnos): Remove a dead code. Initiate
6614 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6616 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6619 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6620 constraint from =wa to wa. Avoid a subreg on the output operand,
6621 instead use a pseudo and subreg it in a move.
6622 (p9_xxbrd_<mode>): Changed to ...
6623 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6624 (p9_xxbrd_v2df): New expander.
6625 (p9_xxbrw_<mode>): Changed to ...
6626 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6627 (p9_xxbrw_v4sf): New expander.
6629 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6631 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6633 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6636 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6637 multiply in 32-bit mode.
6638 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6639 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6642 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6644 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6645 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6646 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6647 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6649 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6652 * dwarf2out.c (is_trivial_indirect_ref): New function.
6653 (dwarf2out_late_global_decl): Do not generate a location
6654 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6655 and that are not defined in the current unit.
6657 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6659 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6660 instead of a libcall for UNORDERED.
6662 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6665 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6666 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6668 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6671 * config/s390/s390.c (s390_set_current_function): Invoke
6672 s390_indirect_branch_settings also if fndecl didn't change.
6674 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6676 * config/rs6000/rs6000.md (blockage): Set length to zero.
6678 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6680 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6682 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6685 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6686 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6690 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6691 PARALLEL incoming that failed vt_get_decl_and_offset check.
6694 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6695 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6697 (get_variable_section): For decls in named .bss* sections pass true as
6698 second argument to bss_initializer_p.
6700 2018-02-09 Marek Polacek <polacek@redhat.com>
6701 Jakub Jelinek <jakub@redhat.com>
6704 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6705 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6706 Sync some changes from cxx_fold_indirect_ref.
6708 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6710 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6712 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6713 (BLOCK_INLINE_ENTRY_LABEL): New.
6714 (dwarf2out_var_location): Disregard inline entry markers.
6715 (inline_entry_data): New struct.
6716 (inline_entry_data_hasher): New hashtable type.
6717 (inline_entry_data_hasher::hash): New.
6718 (inline_entry_data_hasher::equal): New.
6719 (inline_entry_data_table): New variable.
6720 (add_high_low_attributes): Add DW_AT_entry_pc and
6721 DW_AT_GNU_entry_view attributes if a pending entry is found
6722 in inline_entry_data_table. Add old entry_pc attribute only
6723 if debug nonbinding markers are disabled.
6724 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6725 markers are enabled.
6726 (block_within_block_p, dwarf2out_inline_entry): New.
6727 (dwarf2out_finish): Check that no entries remained in
6728 inline_entry_data_table.
6729 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6730 (final_scan_insn, notice_source_line): Likewise.
6731 (rest_of_clean_state): Skip inline entry markers.
6732 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6734 * gimple.c (gimple_build_debug_inline_entry): New.
6735 * gimple.h (enum gimple_debug_subcode): Add
6736 GIMPLE_DEBUG_INLINE_ENTRY.
6737 (gimple_build_debug_inline_entry): Declare.
6738 (gimple_debug_inline_entry_p): New.
6739 (gimple_debug_nonbind_marker_p): Adjust.
6740 * insn-notes.def (INLINE_ENTRY): New.
6741 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6742 inline entry marker notes.
6743 (print_insn): Likewise.
6744 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6745 (INSN_DEBUG_MARKER_KIND): Likewise.
6746 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6747 * tree-inline.c (expand_call_inline): Build and insert
6748 debug_inline_entry stmt.
6749 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6750 inline entry blocks early, if nonbind markers are enabled.
6751 (dump_scope_block): Dump fragment info.
6752 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6753 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6754 (gimple_build_debug_inline_entry): New.
6755 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6756 Enable/disable inline entry points too.
6757 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6758 (DEBUG_INSN): Describe inline entry markers.
6760 * common.opt (gvariable-location-views): New.
6761 (gvariable-location-views=incompat5): New.
6762 * config.in: Rebuilt.
6763 * configure: Rebuilt.
6764 * configure.ac: Test assembler for view support.
6765 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6766 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6767 * dwarf2out.c (var_loc_view): New typedef.
6768 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6769 (dwarf2out_locviews_in_attribute): New.
6770 (dwarf2out_locviews_in_loclist): New.
6771 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6772 (enum dw_line_info_opcode): Add LI_adv_address.
6773 (struct dw_line_info_table): Add view.
6774 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6775 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6776 (zero_view_p): New variable.
6777 (ZERO_VIEW_P): New macro.
6778 (output_asm_line_debug_info): New.
6779 (struct var_loc_node): Add view.
6780 (add_AT_view_list, AT_loc_list): New.
6781 (add_var_loc_to_decl): Add view param. Test it against last.
6782 (new_loc_list): Add view params. Record them.
6783 (AT_loc_list_ptr): Handle loc and view lists.
6784 (view_list_to_loc_list_val_node): New.
6785 (print_dw_val): Handle dw_val_class_view_list.
6786 (size_of_die): Likewise.
6787 (value_format): Likewise.
6788 (loc_list_has_views): New.
6789 (gen_llsym): Set vl_symbol too.
6790 (maybe_gen_llsym, skip_loc_list_entry): New.
6791 (dwarf2out_maybe_output_loclist_view_pair): New.
6792 (output_loc_list): Output view list or entries too.
6793 (output_view_list_offset): New.
6794 (output_die): Handle dw_val_class_view_list.
6795 (output_dwarf_version): New.
6796 (output_compilation_unit_header): Use it.
6797 (output_skeleton_debug_sections): Likewise.
6798 (output_rnglists, output_line_info): Likewise.
6799 (output_pubnames, output_aranges): Update version comments.
6800 (output_one_line_info_table): Output view numbers in asm comments.
6801 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6802 Call maybe_gen_llsym.
6803 (loc_list_from_tree_1): Adjust.
6804 (add_AT_location_description): Create view list attribute if
6805 needed, check it's absent otherwise.
6806 (convert_cfa_to_fb_loc_list): Adjust.
6807 (maybe_emit_file): Call output_asm_line_debug_info for test.
6808 (dwarf2out_var_location): Reset views as needed. Precompute
6809 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6810 attribute. Set view.
6811 (new_line_info_table): Reset next view.
6812 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6813 (dwarf2out_source_line): Likewise. Output view resets and labels to
6814 the assembler, or select appropriate line info opcodes.
6815 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6816 (optimize_string_length): Catch it. Adjust.
6817 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6818 dw_val_class_view_list, and remove it if no longer needed.
6819 (hash_loc_list): Hash view numbers.
6820 (loc_list_hasher::equal): Compare them.
6821 (optimize_location_lists): Check whether a view list symbol is
6822 needed, and whether the locview attribute is present, and
6823 whether they match. Remove the locview attribute if no longer
6825 (index_location_lists): Call skip_loc_list_entry for test.
6826 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6827 Use output_dwarf_version.
6828 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6829 (struct dw_val_node): Add val_view_list.
6830 * final.c (SEEN_NEXT_VIEW): New.
6831 (set_next_view_needed): New.
6832 (clear_next_view_needed): New.
6833 (maybe_output_next_view): New.
6834 (final_start_function): Rename to...
6835 (final_start_function_1): ... this. Take pointer to FIRST,
6836 add SEEN parameter. Emit param bindings in the initial view.
6837 (final_start_function): Reintroduce SEEN-less interface.
6838 (final): Rename to...
6839 (final_1): ... this. Take SEEN parameter. Output final pending
6840 next view at the end.
6841 (final): Reintroduce seen-less interface.
6842 (final_scan_insn): Output pending next view before switching
6843 sections or ending a block. Mark the next view as needed when
6844 outputting variable locations. Notify debug backend of section
6845 changes, and of location view changes.
6846 (rest_of_handle_final): Adjust.
6847 * toplev.c (process_options): Autodetect value for debug variable
6848 location views option. Warn on incompat5 without -gdwarf-5.
6849 * doc/invoke.texi (gvariable-location-views): New.
6850 (gvariable-location-views=incompat5): New.
6851 (gno-variable-location-views): New.
6853 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6855 PR tree-optimization/84136
6856 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6857 that the result of find_edge is non-NULL.
6859 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6862 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6863 storing integer register in SImode. Fix cost of 256 and 512
6864 byte aligned SSE register store.
6866 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6868 * config/i386/i386.c (ix86_multiplication_cost): Fix
6869 multiplication cost for TARGET_AVX512DQ.
6871 2018-02-08 Marek Polacek <polacek@redhat.com>
6873 PR tree-optimization/84238
6874 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6877 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6879 PR tree-optimization/84265
6880 * tree-vect-stmts.c (vectorizable_store): Don't treat
6881 VMAT_CONTIGUOUS accesses as grouped.
6882 (vectorizable_load): Likewise.
6884 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6886 PR tree-optimization/81635
6887 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6888 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6889 (test_round_for_mask): New functions.
6890 (wide_int_cc_tests): Call test_round_for_mask.
6891 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6892 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6893 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6894 range returned by get_range_info.
6896 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6899 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6900 * symtab.c: Include builtins.h
6901 (symtab_node::output_to_lto_symbol_table_p): Move here
6902 from lto-streamer-out.c:output_symbol_p.
6903 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6904 (output_symbol_p): Move all logic to symtab.c
6905 (produce_symtab): Update.
6907 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6909 * config/s390/s390-opts.h (enum indirect_branch): Define.
6910 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6911 (s390_indirect_branch_via_thunk)
6912 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6913 (enum s390_indirect_branch_type): Define.
6914 * config/s390/s390.c (struct s390_frame_layout, struct
6915 machine_function): Remove.
6916 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6917 (indirect_branch_table_label_no, indirect_branch_table_name):
6919 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6920 (enum s390_indirect_branch_option): Define.
6921 (s390_return_addr_from_memory): New function.
6922 (s390_handle_string_attribute): New function.
6923 (s390_attribute_table): Add new attribute handler.
6924 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6925 (s390_indirect_branch_via_thunk): New function.
6926 (s390_indirect_branch_via_inline_thunk): New function.
6927 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6928 sibling call optimization for non z10 compiles.
6929 (s390_emit_call): Force indirect branch target to be a single
6930 register. Add r1 clobber for non-z10 compiles.
6931 (s390_emit_epilogue): Emit return jump via return_use expander.
6932 (s390_reorg): Handle JUMP_INSNs as execute targets.
6933 (s390_option_override_internal): Perform validity checks for the
6934 new command line options.
6935 (s390_indirect_branch_attrvalue): New function.
6936 (s390_indirect_branch_settings): New function.
6937 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6938 (s390_output_indirect_thunk_function): New function.
6939 (s390_code_end): Implement target hook.
6940 (s390_case_values_threshold): Implement target hook.
6941 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6943 * config/s390/s390.h (struct s390_frame_layout)
6944 (struct machine_function): Move here from s390.c.
6945 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6946 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6947 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6948 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6949 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6950 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6951 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6952 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6953 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6954 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6955 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6956 (mnemonic attribute): Add values which aren't recognized
6958 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6959 pattern for branch conversion. Fix mnemonic attribute.
6960 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6961 indirect branch via thunk if requested.
6962 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6963 ("*indirect_jump"): Disable for branch conversion using out of
6965 ("indirect_jump_via_thunk<mode>_z10")
6966 ("indirect_jump_via_thunk<mode>")
6967 ("indirect_jump_via_inlinethunk<mode>_z10")
6968 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6969 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6970 ("casesi_jump_via_inlinethunk<mode>_z10")
6971 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6972 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6973 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6974 ("*indirect2_jump"): Disable for branch conversion.
6975 ("casesi_jump"): Turn into expander and expand patterns for branch
6977 ("return_use"): New expander.
6978 ("*return"): Emit return via thunk and rename it to ...
6979 ("*return<mode>"): ... this one.
6980 * config/s390/s390.opt: Add new options and and enum for the
6983 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6985 * lra-constraints.c (match_reload): Unconditionally use
6986 gen_lowpart_SUBREG, rather than selecting between that
6987 and equivalent gen_rtx_SUBREG code.
6989 2018-02-08 Richard Biener <rguenther@suse.de>
6991 PR tree-optimization/84233
6992 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6993 changed flag instead of boguously re-using phi_inserted.
6995 2018-02-08 Martin Jambor <mjambor@suse.cz>
6997 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6998 static local variables.
7000 2018-02-08 Richard Biener <rguenther@suse.de>
7002 PR tree-optimization/84278
7003 * tree-vect-stmts.c (vectorizable_store): When looking for
7004 smaller vector types to perform grouped strided loads/stores
7005 make sure the mode is supported by the target.
7006 (vectorizable_load): Likewise.
7008 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7010 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7011 Increase LDP/STP opportunities by adding adjacent callee-saves.
7013 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7015 PR rtl-optimization/84068
7016 PR rtl-optimization/83459
7017 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7019 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7021 PR tree-optimization/84224
7022 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7023 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7026 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7029 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7030 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7031 count, remove one USE.
7033 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7035 * doc/install.texi (Configuration): Document the
7036 --with-long-double-format={ibm,ieee} PowerPC configuration
7040 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7041 Convert from define_expand to be define_insn_and_split. Rework
7042 float/double/_Float128 conversions to QI/HI/SImode to work with
7043 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7044 conversions to QI/HImode types did a store and then a load to
7045 truncate the value. For conversions to VSX registers, don't split
7046 the insn, instead emit the code directly. Use the code iterator
7047 any_fix to combine signed and unsigned conversions.
7048 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7049 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7050 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7051 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7052 (fix_<mode>di2_hw): Likewise.
7053 (fixuns_<mode>di2_hw): Likewise.
7054 (fix_<mode>si2_hw): Likewise.
7055 (fixuns_<mode>si2_hw): Likewise.
7056 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7057 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7058 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7059 fix<uns>_trunc<SFDF:mode>si2_p8.
7060 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7062 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7063 (fix<uns>_<mode>_mem): Likewise.
7064 (fctiw<u>z_<mode>_mem): Likewise.
7065 (fix<uns>_<mode>_mem): Likewise.
7066 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7067 the register allocator from doing a direct move to the GPRs to do
7068 a store, and instead use the ISA 3.0 store byte/half-word from
7069 vector register instruction. For IEEE 128-bit floating point,
7070 also optimize stores of 32-bit ints.
7071 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7073 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7075 * genextract.c (push_pathstr_operand): New function to support
7077 (walk_rtx): Call push_pathstr_operand.
7078 (print_path): Support [a-zA-Z].
7080 2018-02-07 Richard Biener <rguenther@suse.de>
7082 PR tree-optimization/84037
7083 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7084 (cse_and_gimplify_to_preheader): Declare.
7085 (vect_get_place_in_interleaving_chain): Likewise.
7086 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7088 (_loop_vec_info::~_loop_vec_info): Delete it.
7089 (cse_and_gimplify_to_preheader): New function.
7090 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7091 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7092 (vectorizable_load): Likewise. For grouped stores always base
7093 the IV on the first element.
7094 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7095 condition before gimplifying.
7097 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7099 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7100 *DIV_EXPR and *MOD_EXPR.
7102 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7105 * config/i386/i386.c (ix86_option_override_internal): Mask out
7106 the CF_SET bit when checking -fcf-protection.
7108 2018-02-07 Tom de Vries <tom@codesourcery.com>
7111 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7114 2018-02-07 Richard Biener <rguenther@suse.de>
7116 PR tree-optimization/84204
7117 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7120 PR tree-optimization/84205
7121 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7122 special-case isl_ast_op_zdiv_r.
7124 PR tree-optimization/84223
7125 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7126 Only add conditions from within the region.
7127 (gather_bbs::after_dom_children): Adjust.
7129 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7132 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7133 * config/avr/avr.md: Only post-reload split REG-REG moves if
7134 either register is GENERAL_REG_P.
7136 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7138 PR tree-optimization/84235
7139 * tree-ssa-scopedtables.c
7140 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7141 if the subtraction is performed in floating point type where NaNs are
7142 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7143 build 1. Formatting fix.
7145 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7148 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7149 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7150 and skip it regardless of bb boundaries. Use CALL_P macro,
7151 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7152 unnecessarily, formatting fix.
7154 2018-02-06 Michael Collison <michael.collison@arm.com>
7156 * config/arm/thumb2.md:
7157 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7158 (*thumb_mov_notscc): Ditto.
7160 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7163 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7164 unsigned_fix, not "s".
7166 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7168 * configure.ac (gcc_fn_eh_frame_ro): New function.
7169 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7170 correct .eh_frame permissions.
7171 * configure: Regenerate.
7173 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7175 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7178 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7180 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7181 Display warning message for -mno-speculate-indirect-jumps.
7183 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7185 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7187 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7189 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7191 PR tree-optimization/84225
7192 * tree-eh.c (find_trapping_overflow): Only call
7193 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7195 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7198 * config/i386/i386.c: Reimplement the check of possible options
7199 -mibt/-mshstk conbination. Change error messages.
7200 * doc/invoke.texi: Fix a typo: remove extra '='.
7202 2018-02-06 Marek Polacek <polacek@redhat.com>
7204 PR tree-optimization/84228
7205 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7207 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7210 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7211 emitted arch directives.
7212 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7213 __ARM_FEATURE_COPROC before changing architectures.
7215 2018-02-06 Richard Biener <rguenther@suse.de>
7217 * config/i386/i386.c (print_reg): Fix typo.
7218 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7220 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7222 * configure: Regenerate.
7224 2018-02-05 Martin Sebor <msebor@redhat.com>
7226 PR tree-optimization/83369
7227 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7230 2018-02-05 Martin Liska <mliska@suse.cz>
7232 * doc/invoke.texi: Cherry-pick upstream r323995.
7234 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7236 * ira.c (ira_init_register_move_cost): Adjust comment.
7238 2018-02-05 Martin Liska <mliska@suse.cz>
7240 PR gcov-profile/84137
7241 * doc/gcov.texi: Fix typo in documentation.
7243 2018-02-05 Martin Liska <mliska@suse.cz>
7245 PR gcov-profile/83879
7246 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7247 using dlopen functionality.
7249 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7251 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7252 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7253 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7254 _mm_maskz_range_round_ss): New intrinsics.
7255 (__builtin_ia32_rangesd128_round)
7256 (__builtin_ia32_rangess128_round): Remove.
7257 (__builtin_ia32_rangesd128_mask_round,
7258 __builtin_ia32_rangess128_mask_round): New builtins.
7259 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7260 __builtin_ia32_rangess128_round): Remove.
7261 (__builtin_ia32_rangesd128_mask_round,
7262 __builtin_ia32_rangess128_mask_round): New builtins.
7263 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7264 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7265 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7266 "<round_saeonly_constraint>")): Changed to ...
7267 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7268 "<round_saeonly_scalar_constraint>")): ... this.
7269 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7270 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7271 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7272 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7273 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7275 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7277 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7279 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7280 Remove all values except native, 8540 and 8548.
7282 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7284 * config/i386/i386.c (ix86_output_function_return): Pass
7285 INVALID_REGNUM, instead of -1, as invalid register number to
7286 indirect_thunk_name and output_indirect_thunk.
7288 2018-02-02 Julia Koval <julia.koval@intel.com>
7290 * config.gcc: Add -march=icelake.
7291 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7292 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7293 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7294 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7295 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7296 (processor_target_table): Add icelake.
7297 (ix86_option_override_internal): Handle new PTAs.
7298 (get_builtin_code_for_version): Handle icelake.
7299 (M_INTEL_COREI7_ICELAKE): New.
7300 (fold_builtin_cpu): Handle icelake.
7301 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7302 * doc/invoke.texi: Add -march=icelake.
7304 2018-02-02 Julia Koval <julia.koval@intel.com>
7306 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7307 to wide_int_bitmask.
7308 * wide-int-bitmask.h: New.
7310 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7313 * config/i386/i386.md: Replace Pmode with word_mode in
7314 builtin_setjmp_setup and builtin_longjmp to support x32.
7316 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7320 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7322 (rs6000_supported_cpu_names): New static variable.
7323 (linux_cpu_translation_table): Likewise.
7324 (elf_platform) <cpu>: Define new static variable and use it.
7325 Translate kernel AT_PLATFORM name to canonical name if needed.
7326 Error if platform name is unknown.
7328 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7331 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7333 2018-02-01 Jeff Law <law@redhat.com>
7336 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7337 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7338 the scratch if RELEASE_VIA_POP is false.
7339 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7340 If we have to save a temporary register, decrement SIZE appropriately.
7341 Pass new arguments to release_scratch_register_on_entry.
7342 (ix86_adjust_stack_and_probe): Likewise.
7343 (ix86_emit_probe_stack_range): Pass new arguments to
7344 release_scratch_register_on_entry.
7346 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7348 PR rtl-optimization/84157
7349 * combine.c (change_zero_ext): Use REG_P predicate in
7350 front of HARD_REGISTER_P predicate.
7352 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7354 * config/avr/avr.c (avr_option_override): Move disabling of
7355 -fdelete-null-pointer-checks to...
7356 * common/config/avr/avr-common.c (avr_option_optimization_table):
7359 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7361 PR tree-optimization/81635
7362 * tree-data-ref.c (split_constant_offset_1): For types that
7363 wrap on overflow, try to use range info to prove that wrapping
7366 2018-02-01 Renlin Li <renlin.li@arm.com>
7369 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7371 (aarch64_register_move_cost): Likewise.
7372 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7374 (REG_CLASS_NAMES): Likewise.
7375 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7376 TAILCALL_ADDR_REGS. Remove IP registers.
7377 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7379 2018-02-01 Richard Biener <rguenther@suse.de>
7381 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7382 for specifying RPO order and allow NULL for that.
7383 * domwalk.c (dom_walker::dom_walker): Likewise.
7384 (dom_walker::walk): Handle NULL RPO order.
7385 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7387 (rewrite_update_dom_walker): Likewise.
7388 (mark_def_dom_walker): Likewise.
7390 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7392 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7393 (aarch64_maybe_expand_sve_subreg_move): Declare.
7394 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7395 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7397 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7398 that are semantically a reverse operation.
7399 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7400 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7401 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7403 (aarch64_can_change_mode_class): For big-endian, forbid changes
7404 between two SVE modes if they have different element sizes.
7406 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7408 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7409 the TImode handling for big-endian targets.
7411 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7413 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7414 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7416 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7417 Remove BSWAP handing for big-endian targets and use the form of
7418 LD1RQ appropariate for the mode.
7420 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7422 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7423 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7426 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7429 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7430 check for operands that need to go through aarch64_sve_reload_be.
7432 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7434 PR tree-optimization/81661
7435 PR tree-optimization/84117
7436 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7437 * tree-eh.c: Include gimplify.h.
7438 (find_trapping_overflow, replace_trapping_overflow,
7439 rewrite_to_non_trapping_overflow): New functions.
7440 * tree-vect-loop.c: Include tree-eh.h.
7441 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7442 * tree-data-ref.c: Include tree-eh.h.
7443 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7445 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7447 PR rtl-optimization/84123
7448 * combine.c (change_zero_ext): Check if hard register satisfies
7449 can_change_dest_mode before calling gen_lowpart_SUBREG.
7451 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7454 * ira.c (ira_init_register_move_cost): Remove assert.
7456 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7458 PR rtl-optimization/84071
7459 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7460 * doc/tm.texi: Regenerate.
7462 2018-01-31 Richard Biener <rguenther@suse.de>
7464 PR tree-optimization/84132
7465 * tree-data-ref.c (analyze_miv_subscript): Properly
7466 check whether evolution_function_is_affine_multivariate_p
7467 before calling gcd_of_steps_may_divide_p.
7469 2018-01-31 Julia Koval <julia.koval@intel.com>
7472 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7473 * config/i386/i386.md (rdpid_rex64) New.
7474 (rdpid): Make 32bit only.
7476 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7479 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7480 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7482 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7485 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7487 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7489 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7491 PR rtl-optimization/84071
7492 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7493 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7495 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7497 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7498 (arc_attribute_table): Add 'aux' attribute.
7499 (arc_in_small_data_p): Consider aux like variables.
7500 (arc_is_aux_reg_p): New function.
7501 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7502 (arc_get_aux_arg): New function.
7503 (prepare_move_operands): Handle aux-register access.
7504 (arc_handle_aux_attribute): New function.
7505 * doc/extend.texi (ARC Variable attributes): Add subsection.
7507 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7509 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7510 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7511 (arc_attribute_table): Add 'uncached' attribute.
7512 (arc_print_operand): Print '.di' flag for uncached memory
7514 (arc_in_small_data_p): Do not consider for small data the uncached
7516 (arc_is_uncached_mem_p): New function.
7517 * config/arc/predicates.md (compact_store_memory_operand): Check
7518 for uncached memory accesses.
7519 (nonvol_nonimm_operand): Likewise.
7520 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7522 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7525 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7526 falign-loops=): Add Optimization flag.
7528 2018-01-30 Jeff Law <law@redhat.com>
7531 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7532 INT_REGISTERS_SAVED. Check it prior to calling
7533 get_scratch_register_on_entry.
7534 (ix86_adjust_stack_and_probe): Similarly.
7535 (ix86_emit_probe_stack_range): Similarly.
7536 (ix86_expand_prologue): Corresponding changes.
7538 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7541 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7542 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7544 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7547 * lra-constraints.c (curr_insn_transform): Process AND in the
7550 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7552 PR rtl-optimization/83986
7553 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7554 dependence against last_pending_memory_flush in addition to
7557 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7559 PR tree-optimization/81611
7560 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7563 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7566 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7569 2018-01-30 Richard Biener <rguenther@suse.de>
7570 Jakub Jelinek <jakub@redhat.com>
7572 PR tree-optimization/84111
7573 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7574 inner loops added during recursion, as they don't have up-to-date
7577 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7580 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7581 (can_inline_edge_by_limits_p): ... here.
7582 (can_early_inline_edge_p, check_callers,
7583 update_caller_keys, update_callee_keys, recursive_inlining,
7584 add_new_edges_to_heap, speculation_useful_p,
7585 inline_small_functions,
7586 inline_small_functions, flatten_function,
7587 inline_to_all_callers_1): Update.
7589 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7591 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7594 2018-01-30 Richard Biener <rguenther@suse.de>
7596 PR tree-optimization/83008
7597 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7598 invariant and constant vector uses in stmts when they need
7601 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7604 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7605 * configure: Regenerate.
7607 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7609 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7611 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7612 Use gen_rtx_REG rather than gen_lowpart.
7614 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7616 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7617 rather than 0 when creating partial subregs.
7619 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7621 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7624 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7627 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7628 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7629 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7630 flags. This restores the settings used before the 2017-07-24.
7631 Turning off pre increment/decrement/modify allows IVOPTS to
7632 optimize DF/SF loops where the index is an int.
7634 2018-01-29 Richard Biener <rguenther@suse.de>
7635 Kelvin Nilsen <kelvin@gcc.gnu.org>
7638 * tree-vect-stmts.c (vectorizable_call): Don't call
7639 targetm.vectorize_builtin_md_vectorized_function if callee is
7642 2018-01-22 Carl Love <cel@us.ibm.com>
7644 * doc/extend.tex: Fix typo in second arg in
7645 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7647 2018-01-29 Richard Biener <rguenther@suse.de>
7649 PR tree-optimization/84086
7650 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7651 (flush_ssaname_freelist): When SSA names were released reset
7652 the SCEV hash table.
7654 2018-01-29 Richard Biener <rguenther@suse.de>
7656 PR tree-optimization/84057
7657 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7658 removed paths when removing edges.
7660 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7662 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7663 -mfunction-return=@var{choice}.
7665 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7668 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7670 (layout::print_source_line): Likewise.
7671 (test_get_line_width_without_trailing_whitespace): Add test cases.
7673 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7676 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7679 2018-01-26 Jim Wilson <jimw@sifive.com>
7681 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7683 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7686 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7688 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7689 and CMP + SUB-immediate -> SUBS.
7691 2018-01-26 Martin Sebor <msebor@redhat.com>
7693 PR tree-optimization/83896
7694 * tree-ssa-strlen.c (get_string_len): Rename...
7695 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7696 Avoid assuming length is constant.
7697 (handle_char_store): Use HOST_WIDE_INT for string length.
7699 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7702 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7703 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7705 2018-01-26 Richard Biener <rguenther@suse.de>
7707 PR rtl-optimization/84003
7708 * dse.c (record_store): Only record redundant stores when
7709 the earlier store aliases at least all accesses the later one does.
7711 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7713 PR rtl-optimization/83985
7714 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7715 REG_CFA_RESTORE insns.
7716 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7717 REG_CFA_RESTORE insns here.
7720 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7721 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7723 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7725 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7726 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7727 (arc_init): Likewise.
7728 (arc_override_options): Likewise.
7729 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7731 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7733 * config/arc/arc.h (TARGET_DBNZ): Define.
7734 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7735 properly set the tune attribute.
7736 (dbnz): Use TARGET_DBNZ guard.
7737 * config/arc/arc.opt (mtune): Add core3 option.
7739 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7741 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7742 recognize new pic like addresses.
7743 (arc_delegitimize_address): Clean up.
7745 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7747 * config/arc/arc-arches.def: Option mrf16 valid for all
7749 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7750 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7751 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7752 * config/arc/arc-tables.opt: Regenerate.
7753 * config/arc/arc.c (arc_conditional_register_usage): Handle
7754 reduced register file case.
7755 (arc_file_start): Set must have build attributes.
7756 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7758 * config/arc/arc.opt (mrf16): Add new option.
7759 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7760 * config/arc/genmultilib.awk: Handle new mrf16 option.
7761 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7762 * config/arc/t-multilib: Regenerate.
7763 * doc/invoke.texi (ARC Options): Document mrf16 option.
7765 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7767 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7768 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7769 (arc_attribute_table): Add 'secure_call' attribute.
7770 (arc_print_operand): Print secure call operand.
7771 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7773 (arc_is_secure_call_p): New function. * config/arc/arc.md
7774 (call_i): Add support for sjli instruction.
7775 (call_value_i): Likewise.
7776 * config/arc/constraints.md (Csc): New constraint.
7778 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7779 John Eric Martin <John.Martin@emmicro-us.com>
7781 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7782 * config/arc/arc.c (_arc_jli_section): New struct.
7783 (arc_jli_section): New type.
7784 (rc_jli_sections): New static variable.
7785 (arc_handle_jli_attribute): New function.
7786 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7787 (arc_file_end): New function.
7788 (TARGET_ASM_FILE_END): Define.
7789 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7790 (arc_add_jli_section): New function.
7791 (jli_call_scan): Likewise.
7792 (arc_reorg): Call jli_call_scan.
7793 (arc_output_addsi): Remove 'S' from printing asm operand.
7794 (arc_is_jli_call_p): New function.
7795 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7797 (movhi_insn): Likewise.
7798 (movsi_insn): Likewise.
7799 (movsi_set_cc_insn): Likewise.
7800 (loadqi_update): Likewise.
7801 (load_zeroextendqisi_update): Likewise.
7802 (load_signextendqisi_update): Likewise.
7803 (loadhi_update): Likewise.
7804 (load_zeroextendhisi_update): Likewise.
7805 (load_signextendhisi_update): Likewise.
7806 (loadsi_update): Likewise.
7807 (loadsf_update): Likewise.
7808 (movsicc_insn): Likewise.
7809 (bset_insn): Likewise.
7810 (bxor_insn): Likewise.
7811 (bclr_insn): Likewise.
7812 (bmsk_insn): Likewise.
7813 (bicsi3_insn): Likewise.
7814 (cmpsi_cc_c_insn): Likewise.
7815 (movsi_ne): Likewise.
7816 (movsi_cond_exec): Likewise.
7817 (clrsbsi2): Likewise.
7828 (call_i): Remove 'S' asm letter, add jli instruction.
7829 (call_value_i): Likewise.
7830 * config/arc/arc.op (mjli-always): New option.
7831 * config/arc/constraints.md (Cji): New constraint.
7832 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7834 (subsf3_fpx): Likewise.
7835 (mulsf3_fpx): Likewise.
7836 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7838 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7840 * doc/invoke.texi (ARC): Document mjli-always option.
7842 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7844 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7845 avoid addition with 0 and use incw and decw where possible.
7847 2018-01-26 Richard Biener <rguenther@suse.de>
7849 PR tree-optimization/81082
7850 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7851 association if it requires casting to unsigned.
7852 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7853 from fold_plusminus_mult_expr to catch important cases late when
7854 range info is available.
7856 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7858 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7859 * configure.ac (hidden_linkonce): New test.
7860 * configure: Regenerate.
7861 * config.in: Regenerate.
7863 2018-01-26 Julia Koval <julia.koval@intel.com>
7865 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7866 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7867 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7868 _mm_mask_bitshuffle_epi64_mask): Fix type.
7869 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7870 USI_FTYPE_V4DI_V4DI_USI): Remove.
7871 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7872 __builtin_ia32_vpshufbitqmb256_mask,
7873 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7874 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7875 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7877 2018-01-26 Alan Modra <amodra@gmail.com>
7880 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7881 UNSPEC_VBPERMQ. Sort other unspecs.
7883 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7885 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7887 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7890 * predict.c (drop_profile): Do not push/pop cfun; update also
7892 (handle_missing_profiles): Fix logic looking for zero profiles.
7894 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7897 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7898 on functions with #pragma omp declare simd or functions with simd
7900 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7901 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7902 Remove trailing \n from warning_at calls.
7904 2018-01-25 Tom de Vries <tom@codesourcery.com>
7907 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7908 for neutered workers.
7910 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7913 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7914 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7916 2018-01-24 Jeff Law <law@redhat.com>
7919 * i386.c (get_probe_interval): Move to earlier point.
7920 (ix86_compute_frame_layout): If -fstack-clash-protection and
7921 the frame is larger than the probe interval, then use pushes
7922 to save registers rather than reg->mem moves.
7923 (ix86_expand_prologue): Remove conditional for int_registers_saved
7926 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7929 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7930 min/max for never referenced object.
7932 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7935 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7937 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7938 attributes from DECL_ATTRIBUTES (decl) without affecting
7939 DECL_ATTRIBUTES (current_function_decl).
7940 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7941 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7943 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7945 PR tree-optimization/83979
7946 * fold-const.c (fold_comparison): Use constant_boolean_node
7947 instead of boolean_{true,false}_node.
7949 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7951 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7954 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7956 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7957 Simplify the clause that sets the length attribute.
7958 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7959 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7960 clause that sets the length attribute.
7961 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7963 2018-01-24 Tom de Vries <tom@codesourcery.com>
7966 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7967 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7968 Add strict parameter.
7969 (prevent_branch_around_nothing): Insert dummy insn between branch to
7970 label and label with no ptx insn inbetween.
7971 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7973 2018-01-24 Tom de Vries <tom@codesourcery.com>
7976 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7977 for neutered threads in warp.
7978 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7980 2018-01-24 Richard Biener <rguenther@suse.de>
7982 PR tree-optimization/83176
7983 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7986 2018-01-24 Richard Biener <rguenther@suse.de>
7988 PR tree-optimization/82819
7989 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7990 code generating pluses that are no-ops in the target precision.
7992 2018-01-24 Richard Biener <rguenther@suse.de>
7995 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7997 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7999 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8000 to merge probabilities.
8001 * predict.c (probably_never_executed): Also mark as cold functions
8002 with global 0 profile and guessed local profile.
8003 * profile-count.c (profile_probability::combine_with_count): New
8005 * profile-count.h (profile_probability::operator*,
8006 profile_probability::operator*=, profile_probability::operator/,
8007 profile_probability::operator/=): Reduce precision to adjusted
8008 and set value to guessed on contradictory divisions.
8009 (profile_probability::combine_with_freq): Remove.
8010 (profile_probability::combine_wiht_count): Declare.
8011 (profile_count::force_nonzero):: Set to adjusted.
8012 (profile_count::probability_in):: Set quality to adjusted.
8013 * tree-ssa-tail-merge.c (replace_block_by): Use
8016 2018-01-23 Andrew Waterman <andrew@sifive.com>
8017 Jim Wilson <jimw@sifive.com>
8019 * config/riscv/riscv.c (riscv_stack_boundary): New.
8020 (riscv_option_override): Set riscv_stack_boundary. Handle
8021 riscv_preferred_stack_boundary_arg.
8022 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8023 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8024 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8025 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8026 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8027 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8029 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8032 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8033 of struct ix86_frame.
8034 (ix86_expand_epilogue): Likewise. Add a local variable for
8035 the reg_save_offset field in struct ix86_frame.
8037 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8039 PR tree-optimization/82604
8040 * tree-loop-distribution.c (enum partition_kind): New enum item
8041 PKIND_PARTIAL_MEMSET.
8042 (partition_builtin_p): Support above new enum item.
8043 (generate_code_for_partition): Ditto.
8044 (compute_access_range): Differentiate cases that equality can be
8045 proven at all loops, the innermost loops or no loops.
8046 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8047 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8048 (finalize_partitions, distribute_loop): Don't fuse partition of
8049 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8050 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8053 2018-01-23 Martin Liska <mliska@suse.cz>
8055 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8056 order to ignore the predictor.
8057 (PRED_POLYMORPHIC_CALL): Likewise.
8058 (PRED_RECURSIVE_CALL): Likewise.
8060 2018-01-23 Martin Liska <mliska@suse.cz>
8062 * tree-profile.c (tree_profiling): Print function header to
8063 aware reader which function we are working on.
8064 * value-prof.c (gimple_find_values_to_profile): Do not print
8065 not interesting value histograms.
8067 2018-01-23 Martin Liska <mliska@suse.cz>
8069 * profile-count.h (enum profile_quality): Add
8070 profile_uninitialized as the first value. Do not number values
8071 as they are zero based.
8072 (profile_count::verify): Update sanity check.
8073 (profile_probability::verify): Likewise.
8075 2018-01-23 Nathan Sidwell <nathan@acm.org>
8077 * doc/invoke.texi (ffor-scope): Deprecate.
8079 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8081 PR tree-optimization/83510
8082 * domwalk.c (set_all_edges_as_executable): New function.
8083 (dom_walker::dom_walker): Convert bool param
8084 "skip_unreachable_blocks" to enum reachability. Move setup of
8085 edge flags to set_all_edges_as_executable and only do it when
8086 reachability is REACHABLE_BLOCKS.
8087 * domwalk.h (enum dom_walker::reachability): New enum.
8088 (dom_walker::dom_walker): Convert bool param
8089 "skip_unreachable_blocks" to enum reachability.
8090 (set_all_edges_as_executable): New decl.
8091 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8092 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8094 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8095 but converting true to REACHABLE_BLOCKS.
8096 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8098 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8099 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8100 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8102 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8103 if check_all_array_refs will be called.
8105 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8107 * tree.c (selftest::test_location_wrappers): Add more test
8110 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8112 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8113 (selftest::test_bit_in_range): Likewise.
8115 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8118 * doc/sourcebuild.texi (vect_float): Say that the selector
8119 only describes the situation when -funsafe-math-optimizations is on.
8120 (vect_float_strict): Document.
8122 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8124 PR tree-optimization/83965
8125 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8126 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8127 instead of checking only for a reduction.
8128 (vect_recog_widen_sum_pattern): Likewise.
8130 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8132 * predict.c (probably_never_executed): Only use precise profile info.
8133 (compute_function_frequency): Skip after inlining hack since we now
8134 have quality checking.
8136 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8138 * profile-count.h (profile_probability::very_unlikely,
8139 profile_probability::unlikely, profile_probability::even): Set
8140 precision to guessed.
8142 2018-01-23 Richard Biener <rguenther@suse.de>
8144 PR tree-optimization/83963
8145 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8146 Properly terminate dominator walk when crossing the exit edge not
8147 when visiting its source block.
8149 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8152 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8153 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8155 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8157 PR tree-optimization/83957
8158 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8159 semicolon after for body surrounded by braces.
8161 PR tree-optimization/83081
8162 * profile-count.h (profile_probability::split): New method.
8163 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8164 Use profile_probability::split.
8165 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8166 when splitting a single conditional jump into 2.
8168 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8170 PR tree-optimization/69452
8171 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8174 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8176 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8177 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8178 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8180 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8182 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8183 * config/rl78/rl78.md (movdi): New define_expand.
8184 * config/rl78/rl78.c (rl78_split_movdi): New function.
8186 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8189 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8191 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8192 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8193 128-bit to produce an UNSPEC move to get the double word with the
8194 signbit and then a shift directly to do signbit.
8195 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8196 implementation with a new version that just does either a direct
8197 move or a regular move. Move memory interface to separate insns.
8198 Move insns so they are next to the expander.
8199 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8200 with signbit move. Split big and little endian case.
8201 (signbit<mode>2_dm_mem_le): Likewise.
8202 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8203 (signbit<mode>2_dm2): Likewise.
8205 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8207 * config/rl78/rl78.md (anddi3): New define_expand.
8209 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8211 * config/rl78/rl78.md (umindi3): New define_expand.
8213 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8215 * config/rl78/rl78.md (smindi3): New define_expand.
8217 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8219 * config/rl78/rl78.md (smaxdi3): New define_expand.
8221 2018-01-22 Carl Love <cel@us.ibm.com>
8223 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8224 LVX_V1TI): Add macro expansion.
8225 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8226 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8227 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8228 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8229 Change check to determine if the instruction is a byte reversing
8230 entry. Fix typo in comment.
8231 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8232 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8233 Add def_builtin calls for new builtins.
8234 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8235 Add define_insn expansion.
8237 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8239 * config/rl78/rl78.md (umaxdi3): New define_expand.
8241 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8243 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8244 for non-QImode registers.
8246 2018-01-22 Richard Biener <rguenther@suse.de>
8248 PR tree-optimization/83963
8249 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8250 including the loop exit block.
8251 (scop_detection::merge_sese): Likewise.
8252 (scop_detection::add_scop): Do it here instead.
8254 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8256 * doc/sourcebuild.texi (arm_softfloat): Document.
8258 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8261 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8262 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8263 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8265 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8266 David Edelsohn <dje.gcc@gmail.com>
8269 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8270 Change "crset eq" to "crset 2".
8271 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8272 (*call_indirect_aix<mode>_nospec): Likewise.
8273 (*call_value_indirect_aix<mode>_nospec): Likewise.
8274 (*call_indirect_elfv2<mode>_nospec): Likewise.
8275 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8276 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8277 change assembly output from . to $.
8278 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8279 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8280 (*tablejump<mode>_internal1_nospec): Likewise.
8282 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8285 * config/sh/sh_optimize_sett_clrt.cc:
8286 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8288 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8290 PR tree-optimization/83940
8291 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8292 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8293 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8294 use it to pass back the definition type.
8295 (vect_check_store_rhs): Likewise rhs_dt_out.
8296 (vect_build_gather_load_calls): Add a mask_dt argument and use
8297 it instead of a call to vect_is_simple_use.
8298 (vectorizable_store): Update calls to vect_check_load_store_mask
8299 and vect_check_store_rhs. Use the dt returned by the latter instead
8300 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8301 instead of calls to vect_is_simple_use. Pass the scalar rather
8302 than the vector operand to vect_is_simple_use when handling
8303 second and subsequent copies of an rhs value.
8304 (vectorizable_load): Update calls to vect_check_load_store_mask
8305 and vect_build_gather_load_calls. Use the cached mask_dt and
8306 gs_info.offset_dt instead of calls to vect_is_simple_use.
8308 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8311 * tree-emutls.c: Include gimplify.h.
8312 (lower_emutls_2): New function.
8313 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8314 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8315 it before further processing.
8318 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8319 UINTVAL (trueop1) instead of INTVAL (op1).
8321 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8325 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8326 INCOMING_FRAME_SP_OFFSET if not defined.
8327 (scan_trace): Add ENTRY argument. If true and
8328 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8329 emit a note to adjust the CFA offset.
8330 (create_cfi_notes): Adjust scan_trace callers.
8331 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8332 INCOMING_FRAME_SP_OFFSET in the CIE.
8333 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8334 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8336 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8337 * doc/tm.texi: Regenerated.
8339 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8341 PR rtl-optimization/83147
8342 * lra-constraints.c (remove_inheritance_pseudos): Use
8343 lra_substitute_pseudo_within_insn.
8345 2018-01-19 Tom de Vries <tom@codesourcery.com>
8346 Cesar Philippidis <cesar@codesourcery.com>
8349 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8351 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8354 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8355 spaces for function labels.
8357 2018-01-19 Martin Liska <mliska@suse.cz>
8359 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8360 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8361 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8362 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8363 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8364 (PRED_CONST_RETURN): Change from 69 to 65.
8365 (PRED_NULL_RETURN): Change from 91 to 71.
8366 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8367 (PRED_LOOP_GUARD): Change from 66 to 73.
8369 2018-01-19 Martin Liska <mliska@suse.cz>
8371 * predict.c (predict_insn_def): Add new assert.
8372 (struct branch_predictor): Change type to signed integer.
8373 (test_prediction_value_range): Amend test to cover
8375 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8376 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8377 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8378 (PRED_LOOP_IV_COMPARE): Likewise.
8379 * predict.h (PROB_UNINITIALIZED): Define new constant.
8381 2018-01-19 Martin Liska <mliska@suse.cz>
8383 * predict.c (dump_prediction): Add new format for
8384 analyze_brprob.py script which is enabled with -details
8386 * profile-count.h (precise_p): New function.
8388 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8390 PR tree-optimization/83922
8391 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8392 there are no statements that need masking.
8393 (vect_active_double_reduction_p): New function.
8394 (vect_analyze_loop_operations): Use it when handling phis that
8395 are not in the loop header.
8397 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8399 PR tree-optimization/83914
8400 * tree-vect-loop.c (vectorizable_induction): Don't convert
8401 init_expr or apply the peeling adjustment for inductions
8402 that are nested within the vectorized loop.
8404 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8406 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8409 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8413 * function.h (gimplify_parameters): Add gimple_seq * argument.
8414 * function.c: Include gimple.h and options.h.
8415 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8416 for the added local temporaries if needed.
8417 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8418 if there are any parameter cleanups, wrap whole body into a
8419 try/finally with the cleanups.
8421 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8424 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8425 Use GET_MODE_CLASS for scalar floating point.
8427 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8431 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8432 Fix call of call_cgraph_insertion_hooks.
8434 2018-01-18 Martin Sebor <msebor@redhat.com>
8436 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8438 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8441 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8444 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8447 * common.opt: (-ffile-prefix-map): New option.
8448 * opts.c (common_handle_option): Defer it.
8449 * opts-global.c (handle_common_deferred_options): Handle it.
8450 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8451 * file-prefix-map.h: New file.
8452 (remap_debug_filename, add_debug_prefix_map): ...here.
8453 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8454 * final.c (debug_prefix_map, add_debug_prefix_map
8455 remap_debug_filename): Move to...
8456 * file-prefix-map.c: New file.
8457 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8458 generalize, get rid of alloca(), use strrchr() instead of strchr().
8459 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8460 Implement in terms of add_prefix_map().
8461 (remap_macro_filename, remap_debug_filename): Implement in term of
8463 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8464 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8465 * dbxout.c: Include file-prefix-map.h.
8466 * varasm.c: Likewise.
8467 * vmsdbgout.c: Likewise.
8468 * xcoffout.c: Likewise.
8469 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8470 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8471 * doc/invoke.texi (-ffile-prefix-map): Document.
8472 (-fdebug-prefix-map): Update description.
8474 2018-01-18 Martin Liska <mliska@suse.cz>
8476 * config/i386/i386.c (indirect_thunk_name): Document that also
8478 (output_indirect_thunk): Document why both instructions
8479 (pause and lfence) are generated.
8481 2018-01-18 Richard Biener <rguenther@suse.de>
8483 PR tree-optimization/83887
8484 * graphite-scop-detection.c
8485 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8486 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8487 (scop_detection::merge_sese): Re-implement with a flood-fill
8488 algorithm that properly finds a SESE region if it exists.
8490 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8493 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8494 pointer_diff optimizations use view_convert instead of convert.
8496 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8498 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8499 Generate different code for -mno-speculate-indirect-jumps.
8500 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8501 (*call_indirect_aix<mode>): Disable for
8502 -mno-speculate-indirect-jumps.
8503 (*call_indirect_aix<mode>_nospec): New define_insn.
8504 (*call_value_indirect_aix<mode>): Disable for
8505 -mno-speculate-indirect-jumps.
8506 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8507 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8508 -mno-speculate-indirect-jumps.
8509 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8511 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8513 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8514 long double type, set the flags for noting the default long double
8515 type, even if we don't pass or return a long double type.
8517 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8520 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8523 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8525 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8526 support for merge[hl].
8527 (fold_mergehl_helper): New helper function.
8528 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8529 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8530 (altivec_vmrglw_direct): Add xxmrglw insn.
8532 2018-01-17 Andrew Waterman <andrew@sifive.com>
8534 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8535 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8537 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8540 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8541 call the lto_location_cache before reading the
8542 DECL_SOURCE_LOCATION of the types.
8544 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8545 Richard Sandiford <richard.sandiford@linaro.org>
8547 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8548 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8549 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8550 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8551 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8553 * config/aarch64/constraints.md (aarch64_movti_operand):
8555 * config/aarch64/predicates.md (Uti): Add new constraint.
8557 2018-01-17 Carl Love <cel@us.ibm.com>
8559 * config/rs6000/vsx.md (define_expand xl_len_r,
8560 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8561 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8563 (define_expand, define_insn): Move the shift left from the
8564 define_insn to the define_expand for lxvl and stxvl instructions.
8565 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8566 and XL_LEN_R definitions to PURE.
8568 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8570 * config/i386/i386.c (indirect_thunk_name): Declare regno
8571 as unsigned int. Compare regno with INVALID_REGNUM.
8572 (output_indirect_thunk): Ditto.
8573 (output_indirect_thunk_function): Ditto.
8574 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8575 in the call to output_indirect_thunk_function.
8577 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8580 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8581 rather than the size of inner_type to determine the stack slot size
8582 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8584 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8587 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8590 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8592 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8593 endian Linux systems to optionally enable multilibs for selecting
8594 the long double type if the user configured an explicit type.
8595 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8596 have no long double multilibs if not defined.
8597 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8598 warn if the user used -mabi={ieee,ibm}longdouble and we built
8599 multilibs for long double.
8600 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8601 appropriate multilib option.
8602 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8604 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8605 for building long double multilibs.
8606 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8608 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8610 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8613 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8615 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8618 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8621 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8624 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8626 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8627 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8629 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8631 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8632 different rtl trees depending on TARGET_64BIT.
8633 (rs6000_gen_lvx): Likewise.
8635 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8637 * config/visium/visium.md (nop): Tweak comment.
8638 (hazard_nop): Likewise.
8640 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8642 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8643 -mspeculate-indirect-jumps.
8644 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8645 for -mno-speculate-indirect-jumps.
8646 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8647 (*call_value_indirect_elfv2<mode>): Disable for
8648 -mno-speculate-indirect-jumps.
8649 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8650 (indirect_jump): Emit different RTL for
8651 -mno-speculate-indirect-jumps.
8652 (*indirect_jump<mode>): Disable for
8653 -mno-speculate-indirect-jumps.
8654 (*indirect_jump<mode>_nospec): New define_insn.
8655 (tablejump): Emit different RTL for
8656 -mno-speculate-indirect-jumps.
8657 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8658 (tablejumpsi_nospec): New define_expand.
8659 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8660 (tablejumpdi_nospec): New define_expand.
8661 (*tablejump<mode>_internal1): Disable for
8662 -mno-speculate-indirect-jumps.
8663 (*tablejump<mode>_internal1_nospec): New define_insn.
8664 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8667 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8669 * caller-save.c (insert_save): Drop unnecessary parameter. All
8672 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8673 Richard Biener <rguenth@suse.de>
8676 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8677 return early, inline manually is_gimple_sizepos. Make sure if we
8678 call gimplify_expr we don't end up with a gimple constant.
8679 * tree.c (variably_modified_type_p): Don't return true for
8680 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8681 * gimplify.h (is_gimple_sizepos): Remove.
8683 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8685 PR tree-optimization/83857
8686 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8687 vectorizable_live_operation for pure SLP statements.
8688 (vectorizable_live_operation): Handle PHIs.
8690 2018-01-16 Richard Biener <rguenther@suse.de>
8692 PR tree-optimization/83867
8693 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8694 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8696 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8699 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8700 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8701 If off is not INTEGER_CST, issue a may not be aligned warning
8702 rather than isn't aligned. Use isn%'t rather than isn't.
8703 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8705 <case MULT_EXPR>: Improve the case when bottom and one of the
8706 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8707 operand, in that case check if the other operand is multiple of
8708 bottom divided by the INTEGER_CST operand.
8710 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8713 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8714 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8715 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8716 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8717 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8718 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8719 (pa_function_arg_size): New function.
8721 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8723 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8724 in a separate statement.
8726 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8728 PR tree-optimization/83847
8729 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8730 group gathers and scatters.
8732 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8734 PR rtl-optimization/86620
8735 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8737 PR rtl-optimization/83213
8738 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8739 to last if both are JUMP_INSNs.
8741 PR tree-optimization/83843
8742 * gimple-ssa-store-merging.c
8743 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8744 store_immediate_info for bswap/nop orig_stores.
8746 2018-01-15 Andrew Waterman <andrew@sifive.com>
8748 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8750 <UDIV>: Increase cost if !TARGET_DIV.
8752 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8754 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8755 (define_attr "cr_logical_3op"): New.
8756 (cceq_ior_compare): Adjust.
8757 (cceq_ior_compare_complement): Adjust.
8758 (*cceq_rev_compare): Adjust.
8759 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8760 (is_cracked_insn): Adjust.
8761 (insn_must_be_first_in_group): Adjust.
8762 * config/rs6000/40x.md: Adjust.
8763 * config/rs6000/440.md: Adjust.
8764 * config/rs6000/476.md: Adjust.
8765 * config/rs6000/601.md: Adjust.
8766 * config/rs6000/603.md: Adjust.
8767 * config/rs6000/6xx.md: Adjust.
8768 * config/rs6000/7450.md: Adjust.
8769 * config/rs6000/7xx.md: Adjust.
8770 * config/rs6000/8540.md: Adjust.
8771 * config/rs6000/cell.md: Adjust.
8772 * config/rs6000/e300c2c3.md: Adjust.
8773 * config/rs6000/e500mc.md: Adjust.
8774 * config/rs6000/e500mc64.md: Adjust.
8775 * config/rs6000/e5500.md: Adjust.
8776 * config/rs6000/e6500.md: Adjust.
8777 * config/rs6000/mpc.md: Adjust.
8778 * config/rs6000/power4.md: Adjust.
8779 * config/rs6000/power5.md: Adjust.
8780 * config/rs6000/power6.md: Adjust.
8781 * config/rs6000/power7.md: Adjust.
8782 * config/rs6000/power8.md: Adjust.
8783 * config/rs6000/power9.md: Adjust.
8784 * config/rs6000/rs64.md: Adjust.
8785 * config/rs6000/titan.md: Adjust.
8787 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8789 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8790 ix86_indirect_branch_register logic.
8792 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8794 * config/i386/constraints.md (Bs): Update
8795 ix86_indirect_branch_register check. Don't check
8796 ix86_indirect_branch_register with GOT_memory_operand.
8798 * config/i386/predicates.md (GOT_memory_operand): Don't check
8799 ix86_indirect_branch_register here.
8800 (GOT32_symbol_operand): Likewise.
8802 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8804 * config/i386/predicates.md (constant_call_address_operand):
8805 Rewrite ix86_indirect_branch_register logic.
8806 (sibcall_insn_operand): Likewise.
8808 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8810 * config/i386/constraints.md (Bs): Replace
8811 ix86_indirect_branch_thunk_register with
8812 ix86_indirect_branch_register.
8814 * config/i386/i386.md (indirect_jump): Likewise.
8815 (tablejump): Likewise.
8816 (*sibcall_memory): Likewise.
8817 (*sibcall_value_memory): Likewise.
8818 Peepholes of indirect call and jump via memory: Likewise.
8819 * config/i386/i386.opt: Likewise.
8820 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8821 (GOT_memory_operand): Likewise.
8822 (call_insn_operand): Likewise.
8823 (sibcall_insn_operand): Likewise.
8824 (GOT32_symbol_operand): Likewise.
8826 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8829 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8830 type rather than type addr's type points to.
8831 (expand_omp_atomic_mutex): Likewise.
8832 (expand_omp_atomic): Likewise.
8834 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8837 * config/i386/i386.c (output_indirect_thunk_function): Use
8838 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8839 for __x86_return_thunk.
8841 2018-01-15 Richard Biener <rguenther@suse.de>
8844 * expmed.c (extract_bit_field_1): Fix typo.
8846 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8849 * config/arm/iterators.md (VF): New mode iterator.
8850 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8851 Remove integer-related logic from pattern.
8852 (neon_vabd<mode>_3): Likewise.
8854 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8857 * common.opt (fstrict-overflow): No longer an alias.
8858 (fwrapv-pointer): New option.
8859 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8860 also for pointer types based on flag_wrapv_pointer.
8861 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8862 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8863 opts->x_flag_wrapv got set.
8864 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8865 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8866 POINTER_TYPE_OVERFLOW_UNDEFINED.
8867 * match.pd: Likewise in address comparison pattern.
8868 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8870 2018-01-15 Richard Biener <rguenther@suse.de>
8873 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8874 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8875 Reset type names to their identifier if their TYPE_DECL doesn't
8876 have linkage (and thus is used for ODR and devirt).
8877 (save_debug_info_for_decl): Remove.
8878 (save_debug_info_for_type): Likewise.
8879 (add_tree_to_fld_list): Adjust.
8880 * tree-pretty-print.c (dump_generic_node): Make dumping of
8881 type names more robust.
8883 2018-01-15 Richard Biener <rguenther@suse.de>
8885 * BASE-VER: Bump to 8.0.1.
8887 2018-01-14 Martin Sebor <msebor@redhat.com>
8890 * builtins.c (check_access): Avoid warning when the no-warning bit
8893 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8895 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8896 * ira-color (allocno_hard_regs_compare): Likewise.
8898 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8901 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8902 Use .pushsection/.popsection.
8904 2018-01-14 Martin Sebor <msebor@redhat.com>
8907 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8909 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8911 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8912 entry from extra_headers.
8913 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8914 extra_headers, make the list bitwise identical to the i?86-*-* one.
8916 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8918 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8919 -mcmodel=large with -mindirect-branch=thunk,
8920 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8921 -mfunction-return=thunk-extern.
8922 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8923 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8924 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8926 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8928 * config/i386/i386.c (print_reg): Print the name of the full
8929 integer register without '%'.
8930 (ix86_print_operand): Handle 'V'.
8931 * doc/extend.texi: Document 'V' modifier.
8933 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8935 * config/i386/constraints.md (Bs): Disallow memory operand for
8936 -mindirect-branch-register.
8938 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8939 (GOT_memory_operand): Likewise.
8940 (call_insn_operand): Likewise.
8941 (sibcall_insn_operand): Likewise.
8942 (GOT32_symbol_operand): Likewise.
8943 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8944 for -mindirect-branch-register.
8945 (tablejump): Likewise.
8946 (*sibcall_memory): Likewise.
8947 (*sibcall_value_memory): Likewise.
8948 Disallow peepholes of indirect call and jump via memory for
8949 -mindirect-branch-register.
8950 (*call_pop): Replace m with Bw.
8951 (*call_value_pop): Likewise.
8952 (*sibcall_pop_memory): Replace m with Bs.
8953 * config/i386/i386.opt (mindirect-branch-register): New option.
8954 * doc/invoke.texi: Document -mindirect-branch-register option.
8956 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8958 * config/i386/i386-protos.h (ix86_output_function_return): New.
8959 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8960 set function_return_type.
8961 (indirect_thunk_name): Add ret_p to indicate thunk for function
8963 (output_indirect_thunk_function): Pass false to
8964 indirect_thunk_name.
8965 (ix86_output_indirect_branch_via_reg): Likewise.
8966 (ix86_output_indirect_branch_via_push): Likewise.
8967 (output_indirect_thunk_function): Create alias for function
8968 return thunk if regno < 0.
8969 (ix86_output_function_return): New function.
8970 (ix86_handle_fndecl_attribute): Handle function_return.
8971 (ix86_attribute_table): Add function_return.
8972 * config/i386/i386.h (machine_function): Add
8973 function_return_type.
8974 * config/i386/i386.md (simple_return_internal): Use
8975 ix86_output_function_return.
8976 (simple_return_internal_long): Likewise.
8977 * config/i386/i386.opt (mfunction-return=): New option.
8978 (indirect_branch): Mention -mfunction-return=.
8979 * doc/extend.texi: Document function_return function attribute.
8980 * doc/invoke.texi: Document -mfunction-return= option.
8982 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8984 * config/i386/i386-opts.h (indirect_branch): New.
8985 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8986 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8987 with local indirect jump when converting indirect call and jump.
8988 (ix86_set_indirect_branch_type): New.
8989 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8990 (indirectlabelno): New.
8991 (indirect_thunk_needed): Likewise.
8992 (indirect_thunk_bnd_needed): Likewise.
8993 (indirect_thunks_used): Likewise.
8994 (indirect_thunks_bnd_used): Likewise.
8995 (INDIRECT_LABEL): Likewise.
8996 (indirect_thunk_name): Likewise.
8997 (output_indirect_thunk): Likewise.
8998 (output_indirect_thunk_function): Likewise.
8999 (ix86_output_indirect_branch_via_reg): Likewise.
9000 (ix86_output_indirect_branch_via_push): Likewise.
9001 (ix86_output_indirect_branch): Likewise.
9002 (ix86_output_indirect_jmp): Likewise.
9003 (ix86_code_end): Call output_indirect_thunk_function if needed.
9004 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9006 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9007 (ix86_attribute_table): Add indirect_branch.
9008 * config/i386/i386.h (machine_function): Add indirect_branch_type
9009 and has_local_indirect_jump.
9010 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9012 (tablejump): Likewise.
9013 (*indirect_jump): Use ix86_output_indirect_jmp.
9014 (*tablejump_1): Likewise.
9015 (simple_return_indirect_internal): Likewise.
9016 * config/i386/i386.opt (mindirect-branch=): New option.
9017 (indirect_branch): New.
9020 (thunk-inline): Likewise.
9021 (thunk-extern): Likewise.
9022 * doc/extend.texi: Document indirect_branch function attribute.
9023 * doc/invoke.texi: Document -mindirect-branch= option.
9025 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9028 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9030 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9032 * ipa-inline.c (want_inline_small_function_p): Return false if
9033 inlining has already failed with CIF_FINAL_ERROR.
9034 (update_caller_keys): Call want_inline_small_function_p before
9036 (update_callee_keys): Likewise.
9038 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9040 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9042 (rs6000_quadword_masked_address_p): Likewise.
9043 (quad_aligned_load_p): Likewise.
9044 (quad_aligned_store_p): Likewise.
9045 (const_load_sequence_p): Add comment to describe the outer-most loop.
9046 (mimic_memory_attributes_and_flags): New function.
9047 (rs6000_gen_stvx): Likewise.
9048 (replace_swapped_aligned_store): Likewise.
9049 (rs6000_gen_lvx): Likewise.
9050 (replace_swapped_aligned_load): Likewise.
9051 (replace_swapped_load_constant): Capitalize argument name in
9052 comment describing this function.
9053 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9054 and stores that access quad-word aligned addresses and replace
9055 with stvx or lvx instructions when appropriate.
9056 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9057 New function prototype.
9058 (rs6000_quadword_masked_address_p): Likewise.
9059 (rs6000_gen_lvx): Likewise.
9060 (rs6000_gen_stvx): Likewise.
9061 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9062 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9063 when memory address is aligned.
9064 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9065 this split to select lvx instruction when memory address is aligned.
9066 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9067 instruction when memory address is aligned.
9068 (*vsx_le_perm_load_v16qi): Likewise.
9069 (four unnamed splitters): Modify to select the stvx instruction
9070 when memory is aligned.
9072 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9074 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9075 which appears in the queue multiple times.
9077 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9078 Alan Hayward <alan.hayward@arm.com>
9079 David Sherwood <david.sherwood@arm.com>
9081 * tree-vectorizer.h (vec_lower_bound): New structure.
9082 (_loop_vec_info): Add check_nonzero and lower_bounds.
9083 (LOOP_VINFO_CHECK_NONZERO): New macro.
9084 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9085 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9086 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9087 fields. Make seg_len the distance travelled, not including the
9089 (dr_direction_indicator): Declare.
9090 (dr_zero_step_indicator): Likewise.
9091 (dr_known_forward_stride_p): Likewise.
9092 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9094 (runtime_alias_check_p): Allow runtime alias checks with
9096 (operator ==): Compare access_size and align.
9097 (prune_runtime_alias_test_list): Rework for new distinction between
9098 the access_size and seg_len.
9099 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9101 (get_segment_min_max): New function.
9102 (create_intersect_range_checks): Use it.
9103 (dr_step_indicator): New function.
9104 (dr_direction_indicator): Likewise.
9105 (dr_zero_step_indicator): Likewise.
9106 (dr_known_forward_stride_p): Likewise.
9107 * tree-loop-distribution.c (data_ref_segment_size): Return
9108 DR_STEP * (niters - 1).
9109 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9111 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9112 (vect_preserves_scalar_order_p): New function, split out from...
9113 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9114 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9115 (vect_vfa_access_size): New function.
9116 (vect_vfa_align): Likewise.
9117 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9118 (dump_lower_bound): New function.
9119 (vect_check_lower_bound): Likewise.
9120 (vect_small_gap_p): Likewise.
9121 (vectorizable_with_step_bound_p): Likewise.
9122 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9123 depencies if the vectorization factor is 1. Convert the checks
9124 for nonzero steps into checks on the bounds of DR_STEP. Try using
9125 a bunds check for variable steps if the minimum required step is
9126 relatively small. Update calls to the dr_with_seg_len
9127 constructor and to vect_compile_time_alias.
9128 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9130 (vect_loop_versioning): Call it.
9131 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9133 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9135 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9136 Alan Hayward <alan.hayward@arm.com>
9137 David Sherwood <david.sherwood@arm.com>
9139 * doc/sourcebuild.texi (vect_scatter_store): Document.
9140 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9142 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9144 * genopinit.c (main): Add supports_vec_scatter_store and
9145 supports_vec_scatter_store_cached to target_optabs.
9146 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9147 IFN_MASK_SCATTER_STORE.
9148 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9150 * internal-fn.h (internal_store_fn_p): Declare.
9151 (internal_fn_stored_value_index): Likewise.
9152 * internal-fn.c (scatter_store_direct): New macro.
9153 (expand_scatter_store_optab_fn): New function.
9154 (direct_scatter_store_optab_supported_p): New macro.
9155 (internal_store_fn_p): New function.
9156 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9157 IFN_MASK_SCATTER_STORE.
9158 (internal_fn_mask_index): Likewise.
9159 (internal_fn_stored_value_index): New function.
9160 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9162 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9163 * optabs-query.c (supports_vec_scatter_store_p): New function.
9164 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9165 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9166 true for scatter stores.
9167 (vect_gather_scatter_fn_p): Handle scatter stores too.
9168 (vect_check_gather_scatter): Consider using scatter stores if
9169 supports_vec_scatter_store_p.
9170 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9172 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9173 internal_fn_stored_value_index.
9174 (check_load_store_masking): Handle scatter stores too.
9175 (vect_get_store_rhs): Make public.
9176 (vectorizable_call): Use internal_store_fn_p.
9177 (vectorizable_store): Handle scatter store internal functions.
9178 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9179 when deciding whether the end of the group has been reached.
9180 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9181 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9182 (mask_scatter_store<mode>): New insns.
9184 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9185 Alan Hayward <alan.hayward@arm.com>
9186 David Sherwood <david.sherwood@arm.com>
9188 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9189 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9190 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9192 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9193 Use vect_truncate_gather_scatter_offset if we can't treat the
9194 operation as a normal gather load or scatter store.
9195 (get_group_load_store_type): Take the gather_scatter_info
9196 as argument. Try using a gather load or scatter store for
9197 single-element groups.
9198 (get_load_store_type): Update calls to get_group_load_store_type
9199 and vect_use_strided_gather_scatters_p.
9201 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9202 Alan Hayward <alan.hayward@arm.com>
9203 David Sherwood <david.sherwood@arm.com>
9205 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9206 optional tree argument.
9207 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9209 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9210 but continue to use the current value as a fallback.
9211 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9212 to compare the updates.
9213 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9214 (get_load_store_type): Use it when handling a strided access.
9215 (vect_get_strided_load_store_ops): New function.
9216 (vect_get_data_ptr_increment): Likewise.
9217 (vectorizable_load): Handle strided gather loads. Always pass
9218 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9220 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9221 Alan Hayward <alan.hayward@arm.com>
9222 David Sherwood <david.sherwood@arm.com>
9224 * doc/md.texi (gather_load@var{m}): Document.
9225 (mask_gather_load@var{m}): Likewise.
9226 * genopinit.c (main): Add supports_vec_gather_load and
9227 supports_vec_gather_load_cached to target_optabs.
9228 * optabs-tree.c (init_tree_optimization_optabs): Use
9229 ggc_cleared_alloc to allocate target_optabs.
9230 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9231 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9233 * internal-fn.h (internal_load_fn_p): Declare.
9234 (internal_gather_scatter_fn_p): Likewise.
9235 (internal_fn_mask_index): Likewise.
9236 (internal_gather_scatter_fn_supported_p): Likewise.
9237 * internal-fn.c (gather_load_direct): New macro.
9238 (expand_gather_load_optab_fn): New function.
9239 (direct_gather_load_optab_supported_p): New macro.
9240 (direct_internal_fn_optab): New function.
9241 (internal_load_fn_p): Likewise.
9242 (internal_gather_scatter_fn_p): Likewise.
9243 (internal_fn_mask_index): Likewise.
9244 (internal_gather_scatter_fn_supported_p): Likewise.
9245 * optabs-query.c (supports_at_least_one_mode_p): New function.
9246 (supports_vec_gather_load_p): Likewise.
9247 * optabs-query.h (supports_vec_gather_load_p): Declare.
9248 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9249 and memory_type field.
9250 (NUM_PATTERNS): Bump to 15.
9251 * tree-vect-data-refs.c: Include internal-fn.h.
9252 (vect_gather_scatter_fn_p): New function.
9253 (vect_describe_gather_scatter_call): Likewise.
9254 (vect_check_gather_scatter): Try using internal functions for
9255 gather loads. Recognize existing calls to a gather load function.
9256 (vect_analyze_data_refs): Consider using gather loads if
9257 supports_vec_gather_load_p.
9258 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9259 (vect_get_gather_scatter_offset_type): Likewise.
9260 (vect_convert_mask_for_vectype): Likewise.
9261 (vect_add_conversion_to_patterm): Likewise.
9262 (vect_try_gather_scatter_pattern): Likewise.
9263 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9264 (vect_vect_recog_func_ptrs): Add it.
9265 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9266 internal_fn_mask_index and internal_gather_scatter_fn_p.
9267 (check_load_store_masking): Take the gather_scatter_info as an
9268 argument and handle gather loads.
9269 (vect_get_gather_scatter_ops): New function.
9270 (vectorizable_call): Check internal_load_fn_p.
9271 (vectorizable_load): Likewise. Handle gather load internal
9273 (vectorizable_store): Update call to check_load_store_masking.
9274 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9275 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9276 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9277 (aarch64_gather_scale_operand_d): New predicates.
9278 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9279 (mask_gather_load<mode>): New insns.
9281 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9282 Alan Hayward <alan.hayward@arm.com>
9283 David Sherwood <david.sherwood@arm.com>
9285 * optabs.def (fold_left_plus_optab): New optab.
9286 * doc/md.texi (fold_left_plus_@var{m}): Document.
9287 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9288 * internal-fn.c (fold_left_direct): Define.
9289 (expand_fold_left_optab_fn): Likewise.
9290 (direct_fold_left_optab_supported_p): Likewise.
9291 * fold-const-call.c (fold_const_fold_left): New function.
9292 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9293 * tree-parloops.c (valid_reduction_p): New function.
9294 (gather_scalar_reductions): Use it.
9295 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9296 (vect_finish_replace_stmt): Declare.
9297 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9298 (needs_fold_left_reduction_p): New function, split out from...
9299 (vect_is_simple_reduction): ...here. Accept reductions that
9300 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9301 (vect_force_simple_reduction): Also store the reduction type in
9302 the assignment's STMT_VINFO_REDUC_TYPE.
9303 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9304 (merge_with_identity): New function.
9305 (vect_expand_fold_left): Likewise.
9306 (vectorize_fold_left_reduction): Likewise.
9307 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9308 scalar phi in place for it. Check for target support and reject
9309 cases that would reassociate the operation. Defer the transform
9310 phase to vectorize_fold_left_reduction.
9311 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9312 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9313 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9315 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9317 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9318 call to ifc_temp_var.
9320 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9321 Alan Hayward <alan.hayward@arm.com>
9322 David Sherwood <david.sherwood@arm.com>
9324 * target.def (legitimize_address_displacement): Take the original
9325 offset as a poly_int.
9326 * targhooks.h (default_legitimize_address_displacement): Update
9328 * targhooks.c (default_legitimize_address_displacement): Likewise.
9329 * doc/tm.texi: Regenerate.
9330 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9331 as an argument, moving assert of ad->disp == ad->disp_term to...
9332 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9333 Try calling targetm.legitimize_address_displacement before expanding
9334 the address rather than afterwards, and adjust for the new interface.
9335 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9336 Match the new hook interface. Handle SVE addresses.
9337 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9340 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9342 * Makefile.in (OBJS): Add early-remat.o.
9343 * target.def (select_early_remat_modes): New hook.
9344 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9345 * doc/tm.texi: Regenerate.
9346 * targhooks.h (default_select_early_remat_modes): Declare.
9347 * targhooks.c (default_select_early_remat_modes): New function.
9348 * timevar.def (TV_EARLY_REMAT): New timevar.
9349 * passes.def (pass_early_remat): New pass.
9350 * tree-pass.h (make_pass_early_remat): Declare.
9351 * early-remat.c: New file.
9352 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9354 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9356 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9357 Alan Hayward <alan.hayward@arm.com>
9358 David Sherwood <david.sherwood@arm.com>
9360 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9361 vfm1 with a bound_epilog parameter.
9362 (vect_do_peeling): Update calls accordingly, and move the prologue
9363 call earlier in the function. Treat the base bound_epilog as 0 for
9364 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9365 this base when peeling for gaps.
9366 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9367 with fully-masked loops.
9368 (vect_estimate_min_profitable_iters): Handle the single peeled
9369 iteration in that case.
9371 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9372 Alan Hayward <alan.hayward@arm.com>
9373 David Sherwood <david.sherwood@arm.com>
9375 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9376 single-element interleaving even if the size is not a power of 2.
9377 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9378 accesses for single-element interleaving if the group size is
9381 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9382 Alan Hayward <alan.hayward@arm.com>
9383 David Sherwood <david.sherwood@arm.com>
9385 * doc/md.texi (fold_extract_last_@var{m}): Document.
9386 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9387 * optabs.def (fold_extract_last_optab): New optab.
9388 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9389 * internal-fn.c (fold_extract_direct): New macro.
9390 (expand_fold_extract_optab_fn): Likewise.
9391 (direct_fold_extract_optab_supported_p): Likewise.
9392 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9393 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9394 EXTRACT_LAST_REDUCTION.
9395 (get_initial_def_for_reduction): Do not create an initial vector
9396 for EXTRACT_LAST_REDUCTION reductions.
9397 (vectorizable_reduction): Leave the scalar phi in place for
9398 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9399 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9400 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9401 transform phase to vectorizable_condition.
9402 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9404 (vect_finish_stmt_generation): ...here.
9405 (vect_finish_replace_stmt): New function.
9406 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9407 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9409 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9411 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9412 Alan Hayward <alan.hayward@arm.com>
9413 David Sherwood <david.sherwood@arm.com>
9415 * doc/md.texi (extract_last_@var{m}): Document.
9416 * optabs.def (extract_last_optab): New optab.
9417 * internal-fn.def (EXTRACT_LAST): New internal function.
9418 * internal-fn.c (cond_unary_direct): New macro.
9419 (expand_cond_unary_optab_fn): Likewise.
9420 (direct_cond_unary_optab_supported_p): Likewise.
9421 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9422 loops using EXTRACT_LAST.
9423 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9424 (extract_last_<mode>): ...this optab.
9425 (vec_extract<mode><Vel>): Update accordingly.
9427 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9428 Alan Hayward <alan.hayward@arm.com>
9429 David Sherwood <david.sherwood@arm.com>
9431 * target.def (empty_mask_is_expensive): New hook.
9432 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9433 * doc/tm.texi: Regenerate.
9434 * targhooks.h (default_empty_mask_is_expensive): Declare.
9435 * targhooks.c (default_empty_mask_is_expensive): New function.
9436 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9437 if the target says that empty masks are expensive.
9438 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9440 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9442 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9443 Alan Hayward <alan.hayward@arm.com>
9444 David Sherwood <david.sherwood@arm.com>
9446 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9447 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9448 (vect_use_loop_mask_for_alignment_p): New function.
9449 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9450 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9451 niters_skip argument. Make sure that the first niters_skip elements
9452 of the first iteration are inactive.
9453 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9454 Update call to vect_set_loop_masks_directly.
9455 (get_misalign_in_elems): New function, split out from...
9456 (vect_gen_prolog_loop_niters): ...here.
9457 (vect_update_init_of_dr): Take a code argument that specifies whether
9458 the adjustment should be added or subtracted.
9459 (vect_update_init_of_drs): Likewise.
9460 (vect_prepare_for_masked_peels): New function.
9461 (vect_do_peeling): Skip prologue peeling if we're using a mask
9462 instead. Update call to vect_update_inits_of_drs.
9463 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9465 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9466 alignment. Do not include the number of peeled iterations in
9467 the minimum threshold in that case.
9468 (vectorizable_induction): Adjust the start value down by
9469 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9470 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9471 Take the number of skipped iterations into account when calculating
9473 * tree-vect-stmts.c (vect_gen_while_not): New function.
9475 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9476 Alan Hayward <alan.hayward@arm.com>
9477 David Sherwood <david.sherwood@arm.com>
9479 * doc/sourcebuild.texi (vect_fully_masked): Document.
9480 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9482 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9484 (vect_analyze_loop_2): ...here. Don't check the vectorization
9485 factor against the number of loop iterations if the loop is
9488 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9489 Alan Hayward <alan.hayward@arm.com>
9490 David Sherwood <david.sherwood@arm.com>
9492 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9493 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9494 (dump_groups): Update accordingly.
9495 (iv_use::mem_type): New member variable.
9496 (address_p): New function.
9497 (record_use): Add a mem_type argument and initialize the new
9499 (record_group_use): Add a mem_type argument. Use address_p.
9500 Remove obsolete null checks of base_object. Update call to record_use.
9501 (find_interesting_uses_op): Update call to record_group_use.
9502 (find_interesting_uses_cond): Likewise.
9503 (find_interesting_uses_address): Likewise.
9504 (get_mem_type_for_internal_fn): New function.
9505 (find_address_like_use): Likewise.
9506 (find_interesting_uses_stmt): Try find_address_like_use before
9507 calling find_interesting_uses_op.
9508 (addr_offset_valid_p): Use the iv mem_type field as the type
9509 of the addressed memory.
9510 (add_autoinc_candidates): Likewise.
9511 (get_address_cost): Likewise.
9512 (split_small_address_groups_p): Use address_p.
9513 (split_address_groups): Likewise.
9514 (add_iv_candidate_for_use): Likewise.
9515 (autoinc_possible_for_pair): Likewise.
9516 (rewrite_groups): Likewise.
9517 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9518 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9519 (get_alias_ptr_type_for_ptr_address): New function.
9520 (rewrite_use_address): Rewrite address uses in calls that were
9521 identified by find_address_like_use.
9523 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9524 Alan Hayward <alan.hayward@arm.com>
9525 David Sherwood <david.sherwood@arm.com>
9527 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9529 * gimple-expr.h (is_gimple_addressable: Likewise.
9530 * gimple-expr.c (is_gimple_address): Likewise.
9531 * internal-fn.c (expand_call_mem_ref): New function.
9532 (expand_mask_load_optab_fn): Use it.
9533 (expand_mask_store_optab_fn): Likewise.
9535 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9536 Alan Hayward <alan.hayward@arm.com>
9537 David Sherwood <david.sherwood@arm.com>
9539 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9540 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9541 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9542 (cond_umax@var{mode}): Document.
9543 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9544 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9545 (cond_umin_optab, cond_umax_optab): New optabs.
9546 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9547 (COND_IOR, COND_XOR): New internal functions.
9548 * internal-fn.h (get_conditional_internal_fn): Declare.
9549 * internal-fn.c (cond_binary_direct): New macro.
9550 (expand_cond_binary_optab_fn): Likewise.
9551 (direct_cond_binary_optab_supported_p): Likewise.
9552 (get_conditional_internal_fn): New function.
9553 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9554 Cope with reduction statements that are vectorized as calls rather
9556 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9557 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9558 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9559 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9560 (UNSPEC_COND_EOR): New unspecs.
9561 (optab): Add mappings for them.
9562 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9563 (sve_int_op, sve_fp_op): New int attributes.
9565 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9566 Alan Hayward <alan.hayward@arm.com>
9567 David Sherwood <david.sherwood@arm.com>
9569 * optabs.def (while_ult_optab): New optab.
9570 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9571 * internal-fn.def (WHILE_ULT): New internal function.
9572 * internal-fn.h (direct_internal_fn_supported_p): New override
9573 that takes two types as argument.
9574 * internal-fn.c (while_direct): New macro.
9575 (expand_while_optab_fn): New function.
9576 (convert_optab_supported_p): Likewise.
9577 (direct_while_optab_supported_p): New macro.
9578 * wide-int.h (wi::udiv_ceil): New function.
9579 * tree-vectorizer.h (rgroup_masks): New structure.
9580 (vec_loop_masks): New typedef.
9581 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9583 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9584 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9585 (vect_max_vf): New function.
9586 (slpeel_make_loop_iterate_ntimes): Delete.
9587 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9588 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9589 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9590 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9591 internal-fn.h, stor-layout.h and optabs-query.h.
9592 (vect_set_loop_mask): New function.
9593 (add_preheader_seq): Likewise.
9594 (add_header_seq): Likewise.
9595 (interleave_supported_p): Likewise.
9596 (vect_maybe_permute_loop_masks): Likewise.
9597 (vect_set_loop_masks_directly): Likewise.
9598 (vect_set_loop_condition_masked): Likewise.
9599 (vect_set_loop_condition_unmasked): New function, split out from
9600 slpeel_make_loop_iterate_ntimes.
9601 (slpeel_make_loop_iterate_ntimes): Rename to..
9602 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9603 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9604 (vect_do_peeling): Update call accordingly.
9605 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9607 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9608 mask_compare_type, can_fully_mask_p and fully_masked_p.
9609 (release_vec_loop_masks): New function.
9610 (_loop_vec_info): Use it to free the loop masks.
9611 (can_produce_all_loop_masks_p): New function.
9612 (vect_get_max_nscalars_per_iter): Likewise.
9613 (vect_verify_full_masking): Likewise.
9614 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9615 retries, and free the mask rgroups before retrying. Check loop-wide
9616 reasons for disallowing fully-masked loops. Make the final decision
9617 about whether use a fully-masked loop or not.
9618 (vect_estimate_min_profitable_iters): Do not assume that peeling
9619 for the number of iterations will be needed for fully-masked loops.
9620 (vectorizable_reduction): Disable fully-masked loops.
9621 (vectorizable_live_operation): Likewise.
9622 (vect_halve_mask_nunits): New function.
9623 (vect_double_mask_nunits): Likewise.
9624 (vect_record_loop_mask): Likewise.
9625 (vect_get_loop_mask): Likewise.
9626 (vect_transform_loop): Handle the case in which the final loop
9627 iteration might handle a partial vector. Call vect_set_loop_condition
9628 instead of slpeel_make_loop_iterate_ntimes.
9629 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9630 (check_load_store_masking): New function.
9631 (prepare_load_store_mask): Likewise.
9632 (vectorizable_store): Handle fully-masked loops.
9633 (vectorizable_load): Likewise.
9634 (supportable_widening_operation): Use vect_halve_mask_nunits for
9636 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9637 (vect_gen_while): New function.
9638 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9639 (aarch64_uqdec<mode>): New insn.
9641 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9642 Alan Hayward <alan.hayward@arm.com>
9643 David Sherwood <david.sherwood@arm.com>
9645 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9646 (reduc_xor_scal_optab): New optabs.
9647 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9648 (reduc_xor_scal_@var{m}): Document.
9649 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9650 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9652 * fold-const-call.c (fold_const_call): Handle them.
9653 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9654 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9655 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9656 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9657 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9658 (UNSPEC_XORV): New unspecs.
9659 (optab): Add entries for them.
9660 (BITWISEV): New int iterator.
9661 (bit_reduc_op): New int attributes.
9663 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9664 Alan Hayward <alan.hayward@arm.com>
9665 David Sherwood <david.sherwood@arm.com>
9667 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9668 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9669 * optabs.def (vec_shl_insert_optab): New optab.
9670 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9671 (duplicate_and_interleave): Likewise.
9672 * tree-vect-loop.c: Include internal-fn.h.
9673 (neutral_op_for_slp_reduction): New function, split out from
9674 get_initial_defs_for_reduction.
9675 (get_initial_def_for_reduction): Handle option 2 for variable-length
9676 vectors by loading the neutral value into a vector and then shifting
9677 the initial value into element 0.
9678 (get_initial_defs_for_reduction): Replace the code argument with
9679 the neutral value calculated by neutral_op_for_slp_reduction.
9680 Use gimple_build_vector for constant-length vectors.
9681 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9682 but the first group_size elements have a neutral value.
9683 Use duplicate_and_interleave otherwise.
9684 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9685 Update call to get_initial_defs_for_reduction. Handle SLP
9686 reductions for variable-length vectors by creating one vector
9687 result for each scalar result, with the elements associated
9688 with other scalar results stubbed out with the neutral value.
9689 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9690 Require IFN_VEC_SHL_INSERT for double reductions on
9691 variable-length vectors, or SLP reductions that have
9692 a neutral value. Require can_duplicate_and_interleave_p
9693 support for variable-length unchained SLP reductions if there
9694 is no neutral value, such as for MIN/MAX reductions. Also require
9695 the number of vector elements to be a multiple of the number of
9696 SLP statements when doing variable-length unchained SLP reductions.
9697 Update call to vect_create_epilog_for_reduction.
9698 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9699 and remove initial values.
9700 (duplicate_and_interleave): Make public.
9701 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9702 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9704 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9705 Alan Hayward <alan.hayward@arm.com>
9706 David Sherwood <david.sherwood@arm.com>
9708 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9709 (can_duplicate_and_interleave_p): New function.
9710 (vect_get_and_check_slp_defs): Take the vector of statements
9711 rather than just the current one. Remove excess parentheses.
9712 Restriction rejectinon of vect_constant_def and vect_external_def
9713 for variable-length vectors to boolean types, or types for which
9714 can_duplicate_and_interleave_p is false.
9715 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9716 (duplicate_and_interleave): New function.
9717 (vect_get_constant_vectors): Use gimple_build_vector for
9718 constant-length vectors and suitable variable-length constant
9719 vectors. Use duplicate_and_interleave for other variable-length
9720 vectors. Don't defer the update when inserting new statements.
9722 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9723 Alan Hayward <alan.hayward@arm.com>
9724 David Sherwood <david.sherwood@arm.com>
9726 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9727 min_profitable_iters doesn't go negative.
9729 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9730 Alan Hayward <alan.hayward@arm.com>
9731 David Sherwood <david.sherwood@arm.com>
9733 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9734 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9735 * optabs.def (vec_mask_load_lanes_optab): New optab.
9736 (vec_mask_store_lanes_optab): Likewise.
9737 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9738 (MASK_STORE_LANES): Likewise.
9739 * internal-fn.c (mask_load_lanes_direct): New macro.
9740 (mask_store_lanes_direct): Likewise.
9741 (expand_mask_load_optab_fn): Handle masked operations.
9742 (expand_mask_load_lanes_optab_fn): New macro.
9743 (expand_mask_store_optab_fn): Handle masked operations.
9744 (expand_mask_store_lanes_optab_fn): New macro.
9745 (direct_mask_load_lanes_optab_supported_p): Likewise.
9746 (direct_mask_store_lanes_optab_supported_p): Likewise.
9747 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9749 (vect_load_lanes_supported): Likewise.
9750 * tree-vect-data-refs.c (strip_conversion): New function.
9751 (can_group_stmts_p): Likewise.
9752 (vect_analyze_data_ref_accesses): Use it instead of checking
9753 for a pair of assignments.
9754 (vect_store_lanes_supported): Take a masked_p parameter.
9755 (vect_load_lanes_supported): Likewise.
9756 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9757 vect_store_lanes_supported and vect_load_lanes_supported.
9758 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9759 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9760 parameter. Don't allow gaps for masked accesses.
9761 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9762 and vect_load_lanes_supported.
9763 (get_load_store_type): Take a masked_p parameter and update
9764 call to get_group_load_store_type.
9765 (vectorizable_store): Update call to get_load_store_type.
9766 Handle IFN_MASK_STORE_LANES.
9767 (vectorizable_load): Update call to get_load_store_type.
9768 Handle IFN_MASK_LOAD_LANES.
9770 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9771 Alan Hayward <alan.hayward@arm.com>
9772 David Sherwood <david.sherwood@arm.com>
9774 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9776 * config/aarch64/aarch64-protos.h
9777 (aarch64_sve_struct_memory_operand_p): Declare.
9778 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9779 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9780 (VPRED, vpred): Handle SVE structure modes.
9781 * config/aarch64/constraints.md (Utx): New constraint.
9782 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9783 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9784 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9785 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9786 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9787 structure modes. Split into pieces after RA.
9788 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9789 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9791 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9792 SVE structure modes.
9793 (aarch64_classify_address): Likewise.
9794 (sizetochar): Move earlier in file.
9795 (aarch64_print_operand): Handle SVE register lists.
9796 (aarch64_array_mode): New function.
9797 (aarch64_sve_struct_memory_operand_p): Likewise.
9798 (TARGET_ARRAY_MODE): Redefine.
9800 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9801 Alan Hayward <alan.hayward@arm.com>
9802 David Sherwood <david.sherwood@arm.com>
9804 * target.def (array_mode): New target hook.
9805 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9806 * doc/tm.texi: Regenerate.
9807 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9808 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9809 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9811 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9814 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9815 Alan Hayward <alan.hayward@arm.com>
9816 David Sherwood <david.sherwood@arm.com>
9818 * fold-const.c (fold_binary_loc): Check the argument types
9819 rather than the result type when testing for a vector operation.
9821 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9823 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9824 * doc/tm.texi: Regenerate.
9826 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9827 Alan Hayward <alan.hayward@arm.com>
9828 David Sherwood <david.sherwood@arm.com>
9830 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9831 (sve): Document new AArch64 extension.
9832 * doc/md.texi (w): Extend the description of the AArch64
9833 constraint to include SVE vectors.
9834 (Upl, Upa): Document new AArch64 predicate constraints.
9835 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9837 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9838 (msve-vector-bits=): New option.
9839 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9840 SVE when these are disabled.
9841 (sve): New extension.
9842 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9843 modes. Adjust their number of units based on aarch64_sve_vg.
9844 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9845 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9846 aarch64_addr_query_type.
9847 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9848 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9849 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9850 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9851 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9852 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9853 (aarch64_simd_imm_zero_p): Delete.
9854 (aarch64_check_zero_based_sve_index_immediate): Declare.
9855 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9856 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9857 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9858 (aarch64_sve_float_mul_immediate_p): Likewise.
9859 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9861 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9862 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9863 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9864 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9865 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9866 (aarch64_regmode_natural_size): Likewise.
9867 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9868 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9870 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9871 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9872 for VG and the SVE predicate registers.
9873 (V_ALIASES): Add a "z"-prefixed alias.
9874 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9875 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9876 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9877 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9878 (REG_CLASS_NAMES): Add entries for them.
9879 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9880 and the predicate registers.
9881 (aarch64_sve_vg): Declare.
9882 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9883 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9884 (REGMODE_NATURAL_SIZE): Define.
9885 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9887 * config/aarch64/aarch64.c: Include cfgrtl.h.
9888 (simd_immediate_info): Add a constructor for series vectors,
9889 and an associated step field.
9890 (aarch64_sve_vg): New variable.
9891 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9892 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9893 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9894 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9895 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9896 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9897 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9898 (aarch64_get_mask_mode): New functions.
9899 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9900 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9901 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9902 predicate modes and predicate registers. Explicitly restrict
9903 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9904 to store a vector mode if it is recognized by
9905 aarch64_classify_vector_mode.
9906 (aarch64_regmode_natural_size): New function.
9907 (aarch64_hard_regno_caller_save_mode): Return the original mode
9909 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9910 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9911 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9912 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9914 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9915 does not overlap dest if the function is frame-related. Handle
9917 (aarch64_split_add_offset): New function.
9918 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9919 them aarch64_add_offset.
9920 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9921 and update call to aarch64_sub_sp.
9922 (aarch64_add_cfa_expression): New function.
9923 (aarch64_expand_prologue): Pass extra temporary registers to the
9924 functions above. Handle the case in which we need to emit new
9925 DW_CFA_expressions for registers that were originally saved
9926 relative to the stack pointer, but now have to be expressed
9927 relative to the frame pointer.
9928 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9930 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9931 IP0 and IP1 values for SVE frames.
9932 (aarch64_expand_vec_series): New function.
9933 (aarch64_expand_sve_widened_duplicate): Likewise.
9934 (aarch64_expand_sve_const_vector): Likewise.
9935 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9936 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9937 into the register, rather than emitting a SET directly.
9938 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9939 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9940 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9941 (offset_9bit_signed_scaled_p): New functions.
9942 (aarch64_replicate_bitmask_imm): New function.
9943 (aarch64_bitmask_imm): Use it.
9944 (aarch64_cannot_force_const_mem): Reject expressions involving
9945 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9946 (aarch64_classify_index): Handle SVE indices, by requiring
9947 a plain register index with a scale that matches the element size.
9948 (aarch64_classify_address): Handle SVE addresses. Assert that
9949 the mode of the address is VOIDmode or an integer mode.
9950 Update call to aarch64_classify_symbol.
9951 (aarch64_classify_symbolic_expression): Update call to
9952 aarch64_classify_symbol.
9953 (aarch64_const_vec_all_in_range_p): New function.
9954 (aarch64_print_vector_float_operand): Likewise.
9955 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9956 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9957 and the FP immediates 1.0 and 0.5.
9958 (aarch64_print_address_internal): Handle SVE addresses.
9959 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9960 (aarch64_regno_regclass): Handle predicate registers.
9961 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9963 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9964 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9965 (aarch64_convert_sve_vector_bits): New function.
9966 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9967 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9969 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9970 Handle SVE vector and predicate modes. Accept VL-based constants
9971 that need only one temporary register, and VL offsets that require
9972 no temporary registers.
9973 (aarch64_conditional_register_usage): Mark the predicate registers
9974 as fixed if SVE isn't available.
9975 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9976 Return true for SVE vector and predicate modes.
9977 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9978 rather than an unsigned int. Handle SVE modes.
9979 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9981 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9983 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9984 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9985 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9986 (aarch64_sve_float_mul_immediate_p): New functions.
9987 (aarch64_sve_valid_immediate): New function.
9988 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9989 Explicitly reject structure modes. Check for INDEX constants.
9990 Handle PTRUE and PFALSE constants.
9991 (aarch64_check_zero_based_sve_index_immediate): New function.
9992 (aarch64_simd_imm_zero_p): Delete.
9993 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9994 vector modes. Accept constants in the range of CNT[BHWD].
9995 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9996 ask for an Advanced SIMD mode.
9997 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9998 (aarch64_simd_vector_alignment): Handle SVE predicates.
9999 (aarch64_vectorize_preferred_vector_alignment): New function.
10000 (aarch64_simd_vector_alignment_reachable): Use it instead of
10002 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10003 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10005 (MAX_VECT_LEN): Delete.
10006 (expand_vec_perm_d): Add a vec_flags field.
10007 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10008 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10009 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10011 (aarch64_evpc_rev): Rename to...
10012 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10013 (aarch64_evpc_rev_global): New function.
10014 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10015 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10017 (aarch64_evpc_sve_tbl): New function.
10018 (aarch64_expand_vec_perm_const_1): Update after rename of
10019 aarch64_evpc_rev. Handle SVE permutes too, trying
10020 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10021 than aarch64_evpc_tbl.
10022 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10023 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10024 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10025 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10026 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10027 (aarch64_expand_sve_vcond): New functions.
10028 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10029 of aarch64_vector_mode_p.
10030 (aarch64_dwarf_poly_indeterminate_value): New function.
10031 (aarch64_compute_pressure_classes): Likewise.
10032 (aarch64_can_change_mode_class): Likewise.
10033 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10034 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10035 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10036 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10037 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10038 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10039 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10040 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10042 (Dn, Dl, Dr): Accept const as well as const_vector.
10043 (Dz): Likewise. Compare against CONST0_RTX.
10044 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10045 of "vector" where appropriate.
10046 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10047 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10048 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10049 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10050 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10051 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10052 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10053 (v_int_equiv): Extend to SVE modes.
10054 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10056 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10057 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10058 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10059 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10060 (SVE_COND_FP_CMP): New int iterators.
10061 (perm_hilo): Handle the new unpack unspecs.
10062 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10064 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10065 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10066 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10067 (aarch64_equality_operator, aarch64_constant_vector_operand)
10068 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10069 (aarch64_sve_nonimmediate_operand): Likewise.
10070 (aarch64_sve_general_operand): Likewise.
10071 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10072 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10073 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10074 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10075 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10076 (aarch64_sve_float_arith_immediate): Likewise.
10077 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10078 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10079 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10080 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10081 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10082 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10083 (aarch64_sve_float_arith_operand): Likewise.
10084 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10085 (aarch64_sve_float_mul_operand): Likewise.
10086 (aarch64_sve_vec_perm_operand): Likewise.
10087 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10088 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10089 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10090 as well as const_vector.
10091 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10092 in file. Use CONST0_RTX and CONSTM1_RTX.
10093 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10094 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10095 Use aarch64_simd_imm_zero.
10096 * config/aarch64/aarch64-sve.md: New file.
10097 * config/aarch64/aarch64.md: Include it.
10098 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10099 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10100 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10101 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10102 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10103 (sve): New attribute.
10104 (enabled): Disable instructions with the sve attribute unless
10106 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10107 aarch64_expand_mov_immediate.
10108 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10109 CNT[BHSD] immediates.
10110 (movti): Split CONST_POLY_INT moves into two halves.
10111 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10112 Split additions that need a temporary here if the destination
10113 is the stack pointer.
10114 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10115 (*add<mode>3_poly_1): New instruction.
10116 (set_clobber_cc): New expander.
10118 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10120 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10121 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10122 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10123 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10124 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10125 Change innermode from fixed_mode_size to machine_mode.
10126 (simplify_subreg): Update call accordingly. Handle a constant-sized
10127 subreg of a variable-length CONST_VECTOR.
10129 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10130 Alan Hayward <alan.hayward@arm.com>
10131 David Sherwood <david.sherwood@arm.com>
10133 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10134 (add_offset_to_base): New function, split out from...
10135 (create_mem_ref): ...here. When handling a scale other than 1,
10136 check first whether the address is valid without the offset.
10137 Add it into the base if so, leaving the index and scale as-is.
10139 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10142 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10143 fold_for_warn before checking if arg2 is INTEGER_CST.
10145 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10147 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10148 (store_multiple_operation): Delete.
10149 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10150 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10151 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10152 guarded by TARGET_STRING.
10153 (rs6000_output_load_multiple): Delete.
10154 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10155 OPTION_MASK_STRING / TARGET_STRING handling.
10156 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10157 (const rs6000_opt_masks) <"string">: Change mask to 0.
10158 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10159 (MASK_STRING): Delete.
10160 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10162 (load_multiple): Delete.
10169 (store_multiple): Delete.
10176 (movmemsi_8reg): Delete.
10177 (corresponding unnamed define_insn): Delete.
10178 (movmemsi_6reg): Delete.
10179 (corresponding unnamed define_insn): Delete.
10180 (movmemsi_4reg): Delete.
10181 (corresponding unnamed define_insn): Delete.
10182 (movmemsi_2reg): Delete.
10183 (corresponding unnamed define_insn): Delete.
10184 (movmemsi_1reg): Delete.
10185 (corresponding unnamed define_insn): Delete.
10186 * config/rs6000/rs6000.opt (mno-string): New.
10187 (mstring): Replace by deprecation warning stub.
10188 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10190 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10192 * regrename.c (regrename_do_replace): If replacing the same
10193 reg multiple times, try to reuse last created gen_raw_REG.
10196 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10197 main to workaround a bug in GDB.
10199 2018-01-12 Tom de Vries <tom@codesourcery.com>
10202 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10204 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10206 PR rtl-optimization/80481
10207 * ira-color.c (get_cap_member): New function.
10208 (allocnos_conflict_by_live_ranges_p): Use it.
10209 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10210 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10212 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10215 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10216 (*saddl_se_1): Ditto.
10217 (*ssubsi_1): Ditto.
10218 (*ssubl_se_1): Ditto.
10220 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10222 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10223 rather than wi::to_widest for DR_INITs.
10224 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10225 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10226 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10228 (vect_analyze_group_access_1): Note that here.
10230 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10232 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10233 polynomial type sizes.
10235 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10237 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10238 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10239 (gimple_add_tmp_var): Likewise.
10241 2018-01-12 Martin Liska <mliska@suse.cz>
10243 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10244 (gimple_alloc_sizes): Likewise.
10245 (dump_gimple_statistics): Use PRIu64 in printf format.
10246 * gimple.h: Change uint64_t to int.
10248 2018-01-12 Martin Liska <mliska@suse.cz>
10250 * tree-core.h: Use uint64_t instead of int.
10251 * tree.c (tree_node_counts): Likewise.
10252 (tree_node_sizes): Likewise.
10253 (dump_tree_statistics): Use PRIu64 in printf format.
10255 2018-01-12 Martin Liska <mliska@suse.cz>
10257 * Makefile.in: As qsort_chk is implemented in vec.c, add
10258 vec.o to linkage of gencfn-macros.
10259 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10260 passing the info to record_node_allocation_statistics.
10261 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10263 * ggc-common.c (struct ggc_usage): Add operator== and use
10264 it in operator< and compare function.
10265 * mem-stats.h (struct mem_usage): Likewise.
10266 * vec.c (struct vec_usage): Remove operator< and compare
10267 function. Can be simply inherited.
10269 2018-01-12 Martin Jambor <mjambor@suse.cz>
10272 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10273 * tree-ssa-math-opts.c: Include domwalk.h.
10274 (convert_mult_to_fma_1): New function.
10275 (fma_transformation_info): New type.
10276 (fma_deferring_state): Likewise.
10277 (cancel_fma_deferring): New function.
10278 (result_of_phi): Likewise.
10279 (last_fma_candidate_feeds_initial_phi): Likewise.
10280 (convert_mult_to_fma): Added deferring logic, split actual
10281 transformation to convert_mult_to_fma_1.
10282 (math_opts_dom_walker): New type.
10283 (math_opts_dom_walker::after_dom_children): New method, body moved
10284 here from pass_optimize_widening_mul::execute, added deferring logic
10286 (pass_optimize_widening_mul::execute): Moved most of code to
10287 math_opts_dom_walker::after_dom_children.
10288 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10289 * config/i386/i386.c (ix86_option_override_internal): Added
10290 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10292 2018-01-12 Richard Biener <rguenther@suse.de>
10295 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10296 inline instance vars.
10298 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10301 * config/rx/rx.c (rx_is_restricted_memory_address):
10302 Handle SUBREG case.
10304 2018-01-12 Richard Biener <rguenther@suse.de>
10306 PR tree-optimization/80846
10307 * target.def (split_reduction): New target hook.
10308 * targhooks.c (default_split_reduction): New function.
10309 * targhooks.h (default_split_reduction): Declare.
10310 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10311 target requests first reduce vectors by combining low and high
10313 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10314 (get_vectype_for_scalar_type_and_size): Export.
10315 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10316 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10317 * doc/tm.texi: Regenerate.
10318 * config/i386/i386.c (ix86_split_reduction): Implement
10319 TARGET_VECTORIZE_SPLIT_REDUCTION.
10321 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10324 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10325 in PIC mode except for TARGET_VXWORKS_RTP.
10326 * config/sparc/sparc.c: Include cfgrtl.h.
10327 (TARGET_INIT_PIC_REG): Define.
10328 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10329 (sparc_pic_register_p): New predicate.
10330 (sparc_legitimate_address_p): Use it.
10331 (sparc_legitimize_pic_address): Likewise.
10332 (sparc_delegitimize_address): Likewise.
10333 (sparc_mode_dependent_address_p): Likewise.
10334 (gen_load_pcrel_sym): Remove 4th parameter.
10335 (load_got_register): Adjust call to above. Remove obsolete stuff.
10336 (sparc_expand_prologue): Do not call load_got_register here.
10337 (sparc_flat_expand_prologue): Likewise.
10338 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10339 (sparc_use_pseudo_pic_reg): New function.
10340 (sparc_init_pic_reg): Likewise.
10341 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10342 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10344 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10346 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10347 Add item for branch_cost.
10349 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10351 PR rtl-optimization/83565
10352 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10353 not extend the result to a larger mode for rotate operations.
10354 (num_sign_bit_copies1): Likewise.
10356 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10359 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10361 Use values-Xc.o for -pedantic.
10362 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10364 2018-01-12 Martin Liska <mliska@suse.cz>
10367 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10369 (possible_polymorphic_call_targets): Use it.
10370 (ipa_devirt): Likewise.
10372 2018-01-12 Martin Liska <mliska@suse.cz>
10374 * profile-count.h (enum profile_quality): Use 0 as invalid
10375 enum value of profile_quality.
10377 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10379 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10380 -mext-string options.
10382 2018-01-12 Richard Biener <rguenther@suse.de>
10384 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10385 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10386 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10388 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10390 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10392 * configure.ac (--with-long-double-format): Add support for the
10393 configuration option to change the default long double format on
10395 * config.gcc (powerpc*-linux*-*): Likewise.
10396 * configure: Regenerate.
10397 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10398 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10399 used without modification.
10401 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10403 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10404 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10405 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10406 MISC_BUILTIN_SPEC_BARRIER.
10407 (rs6000_init_builtins): Likewise.
10408 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10410 (speculation_barrier): New define_insn.
10411 * doc/extend.texi: Document __builtin_speculation_barrier.
10413 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10416 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10417 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10418 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10420 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10421 integral modes instead of "ss" and "sd".
10422 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10423 vectors with 32-bit and 64-bit elements.
10424 (vecdupssescalarmodesuffix): New mode attribute.
10425 (vec_dup<mode>): Use it.
10427 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10430 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10431 frame if argument is passed on stack.
10433 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10436 * ree.c (combine_reaching_defs): Optimize also
10437 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10438 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10440 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10442 PR middle-end/83189
10443 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10445 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10447 PR middle-end/83718
10448 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10449 after they are computed.
10451 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10453 PR tree-optimization/83695
10454 * gimple-loop-linterchange.cc
10455 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10456 reset cached scev information after interchange.
10457 (pass_linterchange::execute): Remove call to scev_reset_htab.
10459 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10461 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10462 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10463 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10464 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10465 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10466 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10467 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10468 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10469 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10470 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10471 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10472 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10473 (V_lane_reg): Likewise.
10474 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10476 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10477 (vfmal_lane_low<mode>_intrinsic,
10478 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10479 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10480 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10481 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10482 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10483 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10485 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10487 * config/arm/arm-cpus.in (fp16fml): New feature.
10488 (ALL_SIMD): Add fp16fml.
10489 (armv8.2-a): Add fp16fml as an option.
10490 (armv8.3-a): Likewise.
10491 (armv8.4-a): Add fp16fml as part of fp16.
10492 * config/arm/arm.h (TARGET_FP16FML): Define.
10493 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10495 * config/arm/arm-modes.def (V2HF): Define.
10496 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10497 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10498 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10499 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10500 vfmsl_low, vfmsl_high): New set of builtins.
10501 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10502 (vfml_op): New code attribute.
10503 (VFMLHALVES): New int iterator.
10504 (VFML, VFMLSEL): New mode attributes.
10505 (V_reg): Define mapping for V2HF.
10506 (V_hi, V_lo): New mode attributes.
10507 (VF_constraint): Likewise.
10508 (vfml_half, vfml_half_selector): New int attributes.
10509 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10511 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10512 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10514 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10515 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10516 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10517 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10519 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10520 Document new effective target and option set.
10522 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10524 * config/arm/arm-cpus.in (armv8_4): New feature.
10525 (ARMv8_4a): New fgroup.
10526 (armv8.4-a): New arch.
10527 * config/arm/arm-tables.opt: Regenerate.
10528 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10529 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10530 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10531 Add matching rules for -march=armv8.4-a and extensions.
10532 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10534 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10537 * config/rx/rx.md (BW): New mode attribute.
10538 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10540 2018-01-11 Richard Biener <rguenther@suse.de>
10542 PR tree-optimization/83435
10543 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10544 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10545 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10547 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10548 Alan Hayward <alan.hayward@arm.com>
10549 David Sherwood <david.sherwood@arm.com>
10551 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10553 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10554 (aarch64_print_address_internal): Use it to check for a zero offset.
10556 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10557 Alan Hayward <alan.hayward@arm.com>
10558 David Sherwood <david.sherwood@arm.com>
10560 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10561 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10562 Return a poly_int64 rather than a HOST_WIDE_INT.
10563 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10564 rather than a HOST_WIDE_INT.
10565 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10566 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10567 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10568 final_offset from HOST_WIDE_INT to poly_int64.
10569 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10570 to_constant when getting the number of units in an Advanced SIMD
10572 (aarch64_builtin_vectorized_function): Check for a constant number
10574 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10576 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10577 attribute instead of GET_MODE_NUNITS.
10578 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10579 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10580 GET_MODE_SIZE for fixed-size registers.
10581 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10582 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10583 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10584 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10585 (aarch64_print_operand, aarch64_print_address_internal)
10586 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10587 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10588 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10589 Handle polynomial GET_MODE_SIZE.
10590 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10591 wider than SImode without modification.
10592 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10593 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10594 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10595 passing and returning SVE modes.
10596 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10597 rather than GEN_INT.
10598 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10599 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10600 (aarch64_allocate_and_probe_stack_space): Likewise.
10601 (aarch64_layout_frame): Cope with polynomial offsets.
10602 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10603 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10604 polynomial offsets.
10605 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10606 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10607 poly_int64 rather than a HOST_WIDE_INT.
10608 (aarch64_get_separate_components, aarch64_process_components)
10609 (aarch64_expand_prologue, aarch64_expand_epilogue)
10610 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10611 (aarch64_anchor_offset): New function, split out from...
10612 (aarch64_legitimize_address): ...here.
10613 (aarch64_builtin_vectorization_cost): Handle polynomial
10614 TYPE_VECTOR_SUBPARTS.
10615 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10617 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10618 number of elements from the PARALLEL rather than the mode.
10619 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10620 rather than GET_MODE_BITSIZE.
10621 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10622 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10623 (aarch64_expand_vec_perm_const_1): Handle polynomial
10624 d->perm.length () and d->perm elements.
10625 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10626 Apply to_constant to d->perm elements.
10627 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10628 polynomial CONST_VECTOR_NUNITS.
10629 (aarch64_move_pointer): Take amount as a poly_int64 rather
10631 (aarch64_progress_pointer): Avoid temporary variable.
10632 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10633 the mode attribute instead of GET_MODE.
10635 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10636 Alan Hayward <alan.hayward@arm.com>
10637 David Sherwood <david.sherwood@arm.com>
10639 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10640 x exists before using it.
10641 (aarch64_add_constant_internal): Rename to...
10642 (aarch64_add_offset_1): ...this. Replace regnum with separate
10643 src and dest rtxes. Handle the case in which they're different,
10644 including when the offset is zero. Replace scratchreg with an rtx.
10645 Use 2 additions if there is no spare register into which we can
10646 move a 16-bit constant.
10647 (aarch64_add_constant): Delete.
10648 (aarch64_add_offset): Replace reg with separate src and dest
10649 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10650 Use aarch64_add_offset_1.
10651 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10652 an rtx rather than an int. Take the delta as a poly_int64
10653 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10654 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10655 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10656 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10657 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10658 and aarch64_add_sp.
10659 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10660 aarch64_add_constant.
10662 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10664 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10665 Use scalar_float_mode.
10667 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10669 * config/aarch64/aarch64-simd.md
10670 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10671 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10672 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10673 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10674 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10675 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10676 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10677 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10678 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10679 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10681 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10684 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10685 targ_options->x_arm_arch_string is non NULL.
10687 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10689 * config/aarch64/aarch64.h
10690 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10692 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10695 * expmed.c (emit_store_flag_force): Swap if const op0
10696 and change VOIDmode to mode of op0.
10698 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10700 PR rtl-optimization/83761
10701 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10702 than bytes to mode_for_size.
10704 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10706 PR middle-end/83189
10707 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10708 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10711 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10713 PR middle-end/83575
10714 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10715 when in layout mode.
10716 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10717 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10720 2018-01-10 Michael Collison <michael.collison@arm.com>
10722 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10723 * config/aarch64/aarch64-option-extension.def: Add
10724 AARCH64_OPT_EXTENSION of 'fp16fml'.
10725 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10726 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10727 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10728 * config/aarch64/constraints.md (Ui7): New constraint.
10729 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10730 (VFMLA_SEL_W): Ditto.
10733 (VFMLA16_LOW): New int iterator.
10734 (VFMLA16_HIGH): Ditto.
10735 (UNSPEC_FMLAL): New unspec.
10736 (UNSPEC_FMLSL): Ditto.
10737 (UNSPEC_FMLAL2): Ditto.
10738 (UNSPEC_FMLSL2): Ditto.
10739 (f16mac): New code attribute.
10740 * config/aarch64/aarch64-simd-builtins.def
10741 (aarch64_fmlal_lowv2sf): Ditto.
10742 (aarch64_fmlsl_lowv2sf): Ditto.
10743 (aarch64_fmlalq_lowv4sf): Ditto.
10744 (aarch64_fmlslq_lowv4sf): Ditto.
10745 (aarch64_fmlal_highv2sf): Ditto.
10746 (aarch64_fmlsl_highv2sf): Ditto.
10747 (aarch64_fmlalq_highv4sf): Ditto.
10748 (aarch64_fmlslq_highv4sf): Ditto.
10749 (aarch64_fmlal_lane_lowv2sf): Ditto.
10750 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10751 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10752 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10753 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10754 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10755 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10756 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10757 (aarch64_fmlal_lane_highv2sf): Ditto.
10758 (aarch64_fmlsl_lane_highv2sf): Ditto.
10759 (aarch64_fmlal_laneq_highv2sf): Ditto.
10760 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10761 (aarch64_fmlalq_lane_highv4sf): Ditto.
10762 (aarch64_fmlsl_lane_highv4sf): Ditto.
10763 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10764 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10765 * config/aarch64/aarch64-simd.md:
10766 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10767 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10768 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10769 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10770 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10771 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10772 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10773 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10774 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10775 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10776 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10777 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10778 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10779 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10780 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10781 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10782 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10783 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10784 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10785 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10786 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10787 (vfmlsl_low_u32): Ditto.
10788 (vfmlalq_low_u32): Ditto.
10789 (vfmlslq_low_u32): Ditto.
10790 (vfmlal_high_u32): Ditto.
10791 (vfmlsl_high_u32): Ditto.
10792 (vfmlalq_high_u32): Ditto.
10793 (vfmlslq_high_u32): Ditto.
10794 (vfmlal_lane_low_u32): Ditto.
10795 (vfmlsl_lane_low_u32): Ditto.
10796 (vfmlal_laneq_low_u32): Ditto.
10797 (vfmlsl_laneq_low_u32): Ditto.
10798 (vfmlalq_lane_low_u32): Ditto.
10799 (vfmlslq_lane_low_u32): Ditto.
10800 (vfmlalq_laneq_low_u32): Ditto.
10801 (vfmlslq_laneq_low_u32): Ditto.
10802 (vfmlal_lane_high_u32): Ditto.
10803 (vfmlsl_lane_high_u32): Ditto.
10804 (vfmlal_laneq_high_u32): Ditto.
10805 (vfmlsl_laneq_high_u32): Ditto.
10806 (vfmlalq_lane_high_u32): Ditto.
10807 (vfmlslq_lane_high_u32): Ditto.
10808 (vfmlalq_laneq_high_u32): Ditto.
10809 (vfmlslq_laneq_high_u32): Ditto.
10810 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10811 (AARCH64_FL_FOR_ARCH8_4): New.
10812 (AARCH64_ISA_F16FML): New ISA flag.
10813 (TARGET_F16FML): New feature flag for fp16fml.
10814 (doc/invoke.texi): Document new fp16fml option.
10816 2018-01-10 Michael Collison <michael.collison@arm.com>
10818 * config/aarch64/aarch64-builtins.c:
10819 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10820 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10821 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10822 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10823 (AARCH64_ISA_SHA3): New ISA flag.
10824 (TARGET_SHA3): New feature flag for sha3.
10825 * config/aarch64/iterators.md (sha512_op): New int attribute.
10826 (CRYPTO_SHA512): New int iterator.
10827 (UNSPEC_SHA512H): New unspec.
10828 (UNSPEC_SHA512H2): Ditto.
10829 (UNSPEC_SHA512SU0): Ditto.
10830 (UNSPEC_SHA512SU1): Ditto.
10831 * config/aarch64/aarch64-simd-builtins.def
10832 (aarch64_crypto_sha512hqv2di): New builtin.
10833 (aarch64_crypto_sha512h2qv2di): Ditto.
10834 (aarch64_crypto_sha512su0qv2di): Ditto.
10835 (aarch64_crypto_sha512su1qv2di): Ditto.
10836 (aarch64_eor3qv8hi): Ditto.
10837 (aarch64_rax1qv2di): Ditto.
10838 (aarch64_xarqv2di): Ditto.
10839 (aarch64_bcaxqv8hi): Ditto.
10840 * config/aarch64/aarch64-simd.md:
10841 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10842 (aarch64_crypto_sha512su0qv2di): Ditto.
10843 (aarch64_crypto_sha512su1qv2di): Ditto.
10844 (aarch64_eor3qv8hi): Ditto.
10845 (aarch64_rax1qv2di): Ditto.
10846 (aarch64_xarqv2di): Ditto.
10847 (aarch64_bcaxqv8hi): Ditto.
10848 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10849 (vsha512h2q_u64): Ditto.
10850 (vsha512su0q_u64): Ditto.
10851 (vsha512su1q_u64): Ditto.
10852 (veor3q_u16): Ditto.
10853 (vrax1q_u64): Ditto.
10854 (vxarq_u64): Ditto.
10855 (vbcaxq_u16): Ditto.
10856 * config/arm/types.md (crypto_sha512): New type attribute.
10857 (crypto_sha3): Ditto.
10858 (doc/invoke.texi): Document new sha3 option.
10860 2018-01-10 Michael Collison <michael.collison@arm.com>
10862 * config/aarch64/aarch64-builtins.c:
10863 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10864 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10865 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10866 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10867 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10868 (AARCH64_ISA_SM4): New ISA flag.
10869 (TARGET_SM4): New feature flag for sm4.
10870 * config/aarch64/aarch64-simd-builtins.def
10871 (aarch64_sm3ss1qv4si): Ditto.
10872 (aarch64_sm3tt1aq4si): Ditto.
10873 (aarch64_sm3tt1bq4si): Ditto.
10874 (aarch64_sm3tt2aq4si): Ditto.
10875 (aarch64_sm3tt2bq4si): Ditto.
10876 (aarch64_sm3partw1qv4si): Ditto.
10877 (aarch64_sm3partw2qv4si): Ditto.
10878 (aarch64_sm4eqv4si): Ditto.
10879 (aarch64_sm4ekeyqv4si): Ditto.
10880 * config/aarch64/aarch64-simd.md:
10881 (aarch64_sm3ss1qv4si): Ditto.
10882 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10883 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10884 (aarch64_sm4eqv4si): Ditto.
10885 (aarch64_sm4ekeyqv4si): Ditto.
10886 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10887 (sm3part_op): Ditto.
10888 (CRYPTO_SM3TT): Ditto.
10889 (CRYPTO_SM3PART): Ditto.
10890 (UNSPEC_SM3SS1): New unspec.
10891 (UNSPEC_SM3TT1A): Ditto.
10892 (UNSPEC_SM3TT1B): Ditto.
10893 (UNSPEC_SM3TT2A): Ditto.
10894 (UNSPEC_SM3TT2B): Ditto.
10895 (UNSPEC_SM3PARTW1): Ditto.
10896 (UNSPEC_SM3PARTW2): Ditto.
10897 (UNSPEC_SM4E): Ditto.
10898 (UNSPEC_SM4EKEY): Ditto.
10899 * config/aarch64/constraints.md (Ui2): New constraint.
10900 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10901 * config/arm/types.md (crypto_sm3): New type attribute.
10902 (crypto_sm4): Ditto.
10903 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10904 (vsm3tt1aq_u32): Ditto.
10905 (vsm3tt1bq_u32): Ditto.
10906 (vsm3tt2aq_u32): Ditto.
10907 (vsm3tt2bq_u32): Ditto.
10908 (vsm3partw1q_u32): Ditto.
10909 (vsm3partw2q_u32): Ditto.
10910 (vsm4eq_u32): Ditto.
10911 (vsm4ekeyq_u32): Ditto.
10912 (doc/invoke.texi): Document new sm4 option.
10914 2018-01-10 Michael Collison <michael.collison@arm.com>
10916 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10917 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10918 (AARCH64_FL_FOR_ARCH8_4): New.
10919 (AARCH64_FL_V8_4): New flag.
10920 (doc/invoke.texi): Document new armv8.4-a option.
10922 2018-01-10 Michael Collison <michael.collison@arm.com>
10924 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10925 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10926 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10927 * config/aarch64/aarch64-option-extension.def: Add
10928 AARCH64_OPT_EXTENSION of 'sha2'.
10929 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10930 (crypto): Disable sha2 and aes if crypto disabled.
10931 (crypto): Enable aes and sha2 if enabled.
10932 (simd): Disable sha2 and aes if simd disabled.
10933 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10935 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10936 (TARGET_SHA2): New feature flag for sha2.
10937 (TARGET_AES): New feature flag for aes.
10938 * config/aarch64/aarch64-simd.md:
10939 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10940 conditional on TARGET_AES.
10941 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10942 (aarch64_crypto_sha1hsi): Make pattern conditional
10944 (aarch64_crypto_sha1hv4si): Ditto.
10945 (aarch64_be_crypto_sha1hv4si): Ditto.
10946 (aarch64_crypto_sha1su1v4si): Ditto.
10947 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10948 (aarch64_crypto_sha1su0v4si): Ditto.
10949 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10950 (aarch64_crypto_sha256su0v4si): Ditto.
10951 (aarch64_crypto_sha256su1v4si): Ditto.
10952 (doc/invoke.texi): Document new aes and sha2 options.
10954 2018-01-10 Martin Sebor <msebor@redhat.com>
10956 PR tree-optimization/83781
10957 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10960 2018-01-11 Martin Sebor <msebor@gmail.com>
10961 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10963 PR tree-optimization/83501
10964 PR tree-optimization/81703
10966 * tree-ssa-strlen.c (get_string_cst): Rename...
10967 (get_string_len): ...to this. Handle global constants.
10968 (handle_char_store): Adjust.
10970 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10971 Jim Wilson <jimw@sifive.com>
10973 * config/riscv/riscv-protos.h (riscv_output_return): New.
10974 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10975 (riscv_attribute_table, riscv_output_return),
10976 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10977 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10978 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10980 (riscv_expand_prologue): Add early return for naked function.
10981 (riscv_expand_epilogue): Likewise.
10982 (riscv_function_ok_for_sibcall): Return false for naked function.
10983 (riscv_set_current_function): New.
10984 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10985 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10986 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10987 * doc/extend.texi (RISC-V Function Attributes): New.
10989 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10991 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10992 check for 128-bit long double before checking TCmode.
10993 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10994 128-bit long doubles before checking TFmode or TCmode.
10995 (FLOAT128_IBM_P): Likewise.
10997 2018-01-10 Martin Sebor <msebor@redhat.com>
10999 PR tree-optimization/83671
11000 * builtins.c (c_strlen): Unconditionally return zero for the empty
11002 Use -Warray-bounds for warnings.
11003 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11004 for non-constant array indices with COMPONENT_REF, arrays of
11005 arrays, and pointers to arrays.
11006 (gimple_fold_builtin_strlen): Determine and set length range for
11007 non-constant character arrays.
11009 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11011 PR middle-end/81897
11012 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11015 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11017 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11019 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11022 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11023 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11024 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11025 indexed_or_indirect_operand predicate.
11026 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11027 (*vsx_le_perm_load_v8hi): Likewise.
11028 (*vsx_le_perm_load_v16qi): Likewise.
11029 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11030 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11031 (*vsx_le_perm_store_v8hi): Likewise.
11032 (*vsx_le_perm_store_v16qi): Likewise.
11033 (eight unnamed splitters): Likewise.
11035 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11037 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11038 * config/rs6000/emmintrin.h: Likewise.
11039 * config/rs6000/mmintrin.h: Likewise.
11040 * config/rs6000/xmmintrin.h: Likewise.
11042 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11045 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11047 * tree.c (tree_nop_conversion): Return true for location wrapper
11049 (maybe_wrap_with_location): New function.
11050 (selftest::check_strip_nops): New function.
11051 (selftest::test_location_wrappers): New function.
11052 (selftest::tree_c_tests): Call it.
11053 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11054 (maybe_wrap_with_location): New decl.
11055 (EXPR_LOCATION_WRAPPER_P): New macro.
11056 (location_wrapper_p): New inline function.
11057 (tree_strip_any_location_wrapper): New inline function.
11059 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11062 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11063 stack_realign_offset for the largest alignment of stack slot
11065 (ix86_find_max_used_stack_alignment): New function.
11066 (ix86_finalize_stack_frame_flags): Use it. Set
11067 max_used_stack_alignment if we don't realign stack.
11068 * config/i386/i386.h (machine_function): Add
11069 max_used_stack_alignment.
11071 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11073 * config/arm/arm.opt (-mbranch-cost): New option.
11074 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11077 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11080 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11081 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11083 2018-01-10 Richard Biener <rguenther@suse.de>
11086 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11087 early out so it also covers the case where we have a non-NULL
11090 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11092 PR tree-optimization/83753
11093 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11094 for non-strided grouped accesses if the number of elements is 1.
11096 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11099 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11100 * i386.h (TARGET_USE_GATHER): Define.
11101 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11103 2018-01-10 Martin Liska <mliska@suse.cz>
11106 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11107 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11109 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11110 CLEANUP_NO_PARTITIONING is not set.
11112 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11114 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11115 for vectors, as a partial revert of r254296.
11116 * rtl.h (const_vec_p): Delete.
11117 (const_vec_duplicate_p): Don't test for vector CONSTs.
11118 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11119 * expmed.c (make_tree): Likewise.
11122 * common.md (E, F): Use CONSTANT_P instead of checking for
11124 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11125 checking for CONST_VECTOR.
11127 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11129 PR middle-end/83575
11130 * predict.c (force_edge_cold): Handle in more sane way edges
11131 with no prediction.
11133 2018-01-09 Carl Love <cel@us.ibm.com>
11135 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11137 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11138 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11139 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11140 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11141 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11142 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11143 * config/rs6000/rs6000-protos.h: Add extern defition for
11144 rs6000_generate_float2_double_code.
11145 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11147 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11148 (float2_v2df): Add define_expand.
11150 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11153 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11154 op_mode in the force_to_mode call.
11156 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11158 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11159 instead of checking each element individually.
11160 (aarch64_evpc_uzp): Likewise.
11161 (aarch64_evpc_zip): Likewise.
11162 (aarch64_evpc_ext): Likewise.
11163 (aarch64_evpc_rev): Likewise.
11164 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11165 instead of checking each element individually. Return true without
11167 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11168 whether all selected elements come from the same input, instead of
11169 checking each element individually. Remove calls to gen_rtx_REG,
11170 start_sequence and end_sequence and instead assert that no rtl is
11173 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11175 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11176 order of HIGH and CONST checks.
11178 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11180 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11181 if the destination isn't an SSA_NAME.
11183 2018-01-09 Richard Biener <rguenther@suse.de>
11185 PR tree-optimization/83668
11186 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11188 (canonicalize_loop_form): ... here, renamed from ...
11189 (canonicalize_loop_closed_ssa_form): ... this and amended to
11190 swap successor edges for loop exit blocks to make us use
11191 the RPO order we need for initial schedule generation.
11193 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11195 PR tree-optimization/64811
11196 * match.pd: When optimizing comparisons with Inf, avoid
11197 introducing or losing exceptions from comparisons with NaN.
11199 2018-01-09 Martin Liska <mliska@suse.cz>
11202 * asan.c (shadow_mem_size): Add gcc_assert.
11204 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11206 Don't save registers in main().
11209 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11210 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11211 * config/avr/avr.c (avr_set_current_function): Don't error if
11212 naked, OS_task or OS_main are specified at the same time.
11213 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11215 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11217 * common/config/avr/avr-common.c (avr_option_optimization_table):
11218 Switch on -mmain-is-OS_task for optimizing compilations.
11220 2018-01-09 Richard Biener <rguenther@suse.de>
11222 PR tree-optimization/83572
11223 * graphite.c: Include cfganal.h.
11224 (graphite_transform_loops): Connect infinite loops to exit
11225 and remove fake edges at the end.
11227 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11229 * ipa-inline.c (edge_badness): Revert accidental checkin.
11231 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11234 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11235 symbols; not inline clones.
11237 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11240 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11241 hard registers. Formatting fixes.
11243 PR preprocessor/83722
11244 * gcc.c (try_generate_repro): Pass
11245 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11246 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11249 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11250 Kito Cheng <kito.cheng@gmail.com>
11252 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11253 (riscv_leaf_function_p): Delete.
11254 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11256 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11258 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11260 (do_ifelse): New function.
11261 (do_isel): New function.
11262 (do_sub3): New function.
11263 (do_add3): New function.
11264 (do_load_mask_compare): New function.
11265 (do_overlap_load_compare): New function.
11266 (expand_compare_loop): New function.
11267 (expand_block_compare): Call expand_compare_loop() when appropriate.
11268 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11269 option description.
11270 (-mblock-compare-inline-loop-limit): New option.
11272 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11275 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11276 Reverse order of second and third operands in first alternative.
11277 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11278 of first and second elements in UNSPEC_VPERMR vector.
11279 (altivec_expand_vec_perm_le): Likewise.
11281 2018-01-08 Jeff Law <law@redhat.com>
11283 PR rtl-optimizatin/81308
11284 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11285 (process_switch): If group_case_labels makes a change, then set
11287 (pass_convert_switch::execute): If a switch is converted, then
11288 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11290 PR rtl-optimization/81308
11291 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11294 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11296 PR target/83663 - Revert r255946
11297 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11298 generation for cases where splatting a value is not useful.
11299 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11300 across a vec_duplicate and a paradoxical subreg forming a vector
11301 mode to a vec_concat.
11303 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11305 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11306 -march=armv8.3-a variants.
11307 * config/arm/t-multilib: Likewise.
11308 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11310 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11312 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11314 (cceq_ior_compare_complement): Give it a name so I can use it, and
11315 change boolean_or_operator predicate to boolean_operator so it can
11316 be used to generate a crand.
11317 (eqne): New code iterator.
11318 (bd/bd_neg): New code_attrs.
11319 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11320 a single define_insn.
11321 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11322 decrement (bdnzt/bdnzf/bdzt/bdzf).
11323 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11324 with the new names of the branch decrement patterns, and added the
11325 names of the branch decrement conditional patterns.
11327 2018-01-08 Richard Biener <rguenther@suse.de>
11329 PR tree-optimization/83563
11330 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11333 2018-01-08 Richard Biener <rguenther@suse.de>
11335 PR middle-end/83713
11336 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11338 2018-01-08 Richard Biener <rguenther@suse.de>
11340 PR tree-optimization/83685
11341 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11342 references to abnormals.
11344 2018-01-08 Richard Biener <rguenther@suse.de>
11347 * dwarf2out.c (output_indirect_strings): Handle empty
11348 skeleton_debug_str_hash.
11349 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11351 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11353 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11354 (emit_store_direct): Likewise.
11355 (arc_trampoline_adjust_address): Likewise.
11356 (arc_asm_trampoline_template): New function.
11357 (arc_initialize_trampoline): Use asm_trampoline_template.
11358 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11359 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11360 * config/arc/arc.md (flush_icache): Delete pattern.
11362 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11364 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11365 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11368 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11371 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11372 by not USED_FOR_TARGET.
11373 (make_pass_resolve_sw_modes): Likewise.
11375 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11377 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11380 2018-01-08 Richard Biener <rguenther@suse.de>
11382 PR middle-end/83580
11383 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11385 2018-01-08 Richard Biener <rguenther@suse.de>
11387 PR middle-end/83517
11388 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11390 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11392 PR middle-end/81897
11393 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11394 basic blocks with a small number of successors.
11395 (convert_control_dep_chain_into_preds): Improve handling of
11397 (dump_predicates): Split apart into...
11398 (dump_pred_chain): ...here...
11399 (dump_pred_info): ...and here.
11400 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11401 (can_chain_union_be_invalidated_p): Improve check for invalidation
11403 (uninit_uses_cannot_happen): Avoid unnecessary if
11404 convert_control_dep_chain_into_preds yielded nothing.
11406 2018-01-06 Martin Sebor <msebor@redhat.com>
11408 PR tree-optimization/83640
11409 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11410 subtracting negative offset from size.
11411 (builtin_access::overlap): Adjust offset bounds of the access to fall
11412 within the size of the object if possible.
11414 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11416 PR rtl-optimization/83699
11417 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11418 extract_bit_field_as_subreg to cases in which the extracted
11419 value is also a vector.
11421 * lra-constraints.c (process_alt_operands): Test for the equivalence
11422 substitutions when detecting a possible reload cycle.
11424 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11427 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11428 by default if flag_selective_schedling{,2}. Formatting fixes.
11430 PR rtl-optimization/83682
11431 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11432 if it has non-VECTOR_MODE element mode.
11433 (vec_duplicate_p): Likewise.
11435 PR middle-end/83694
11436 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11437 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11439 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11442 * config/i386/i386-builtin.def
11443 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11444 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11445 Require also OPTION_MASK_ISA_AVX512F in addition to
11446 OPTION_MASK_ISA_GFNI.
11447 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11448 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11449 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11450 to OPTION_MASK_ISA_GFNI.
11451 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11452 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11453 OPTION_MASK_ISA_AVX512BW.
11454 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11455 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11456 addition to OPTION_MASK_ISA_GFNI.
11457 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11458 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11459 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11460 to OPTION_MASK_ISA_GFNI.
11461 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11462 a requirement for all ISAs rather than any of them with a few
11464 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11466 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11467 bitmasks to be enabled with 3 exceptions, instead of requiring any
11468 enabled ISA with lots of exceptions.
11469 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11470 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11471 Change avx512bw in isa attribute to avx512f.
11472 * config/i386/sgxintrin.h: Add license boilerplate.
11473 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11474 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11475 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11476 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11478 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11479 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11480 temporarily sse2 rather than sse if not enabled already.
11483 * config/i386/sse.md (VI248_VLBW): Rename to ...
11484 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11485 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11486 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11487 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11488 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11489 mode iterator instead of VI248_VLBW.
11491 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11493 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11494 (record_modified): Skip clobbers; add debug output.
11495 (param_change_prob): Use sreal frequencies.
11497 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11499 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11500 punt for user-aligned variables.
11502 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11504 * tree-chrec.c (chrec_contains_symbols): Return true for
11507 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11510 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11511 of (x|y) == x for BICS pattern.
11513 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11515 PR tree-optimization/83605
11516 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11517 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11520 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11522 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11523 * config/epiphany/rtems.h: New file.
11525 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11526 Uros Bizjak <ubizjak@gmail.com>
11529 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11530 QIreg_operand instead of register_operand predicate.
11531 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11532 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11533 comments instead of -fmitigate[-_]rop.
11535 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11538 * cgraphunit.c (symbol_table::compile): Switch to text_section
11539 before calling assembly_start debug hook.
11540 * run-rtl-passes.c (run_rtl_passes): Likewise.
11543 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11545 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11546 range_int_cst_p rather than !symbolic_range_p before calling
11547 extract_range_from_multiplicative_op_1.
11549 2018-01-04 Jeff Law <law@redhat.com>
11551 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11552 redundant test in assertion.
11554 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11556 * doc/rtl.texi: Document machine_mode wrapper classes.
11558 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11560 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11561 using tree_to_uhwi.
11563 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11565 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11566 the VEC_PERM_EXPR fold to fail.
11568 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11571 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11572 to switched_sections.
11574 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11577 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11578 test for d.testing.
11580 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11583 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11584 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11586 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11589 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11590 is BLKmode and bitpos not zero or mode change is needed.
11592 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11595 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11598 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11601 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11602 instead of MULT rtx. Update all corresponding splitters.
11603 (*saddl_se): Ditto.
11604 (*ssub<modesuffix>): Ditto.
11605 (*ssubl_se): Ditto.
11606 (*cmp_sadd_di): Update split patterns.
11607 (*cmp_sadd_si): Ditto.
11608 (*cmp_sadd_sidi): Ditto.
11609 (*cmp_ssub_di): Ditto.
11610 (*cmp_ssub_si): Ditto.
11611 (*cmp_ssub_sidi): Ditto.
11612 * config/alpha/predicates.md (const23_operand): New predicate.
11613 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11614 Look for ASHIFT, not MULT inner operand.
11615 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11617 2018-01-04 Martin Liska <mliska@suse.cz>
11619 PR gcov-profile/83669
11620 * gcov.c (output_intermediate_file): Add version to intermediate
11622 * doc/gcov.texi: Document new field 'version' in intermediate
11623 file format. Fix location of '-k' option of gcov command.
11625 2018-01-04 Martin Liska <mliska@suse.cz>
11628 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11630 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11632 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11634 2018-01-03 Martin Sebor <msebor@redhat.com>
11636 PR tree-optimization/83655
11637 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11638 checking calls with invalid arguments.
11640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11642 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11643 (vectorizable_mask_load_store): Delete.
11644 (vectorizable_call): Return false for masked loads and stores.
11645 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11646 instead of gimple_assign_rhs1.
11647 (vectorizable_load): Handle IFN_MASK_LOAD.
11648 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11652 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11654 (vectorizable_mask_load_store): ...here.
11655 (vectorizable_load): ...and here.
11657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11659 * tree-vect-stmts.c (vect_build_all_ones_mask)
11660 (vect_build_zero_merge_argument): New functions, split out from...
11661 (vectorizable_load): ...here.
11663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11665 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11667 (vectorizable_mask_load_store): ...here.
11668 (vectorizable_store): ...and here.
11670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11672 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11674 (vectorizable_mask_load_store): ...here.
11676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11678 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11679 (vect_model_store_cost): Take a vec_load_store_type instead of a
11681 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11682 (vect_model_store_cost): Take a vec_load_store_type instead of a
11684 (vectorizable_mask_load_store): Update accordingly.
11685 (vectorizable_store): Likewise.
11686 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11690 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11691 IFN_MASK_LOAD calls here rather than...
11692 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11694 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11695 Alan Hayward <alan.hayward@arm.com>
11696 David Sherwood <david.sherwood@arm.com>
11698 * expmed.c (extract_bit_field_1): For vector extracts,
11699 fall back to extract_bit_field_as_subreg if vec_extract
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 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11707 they are variable or constant sized.
11708 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11709 slots for constant-sized data.
11711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11712 Alan Hayward <alan.hayward@arm.com>
11713 David Sherwood <david.sherwood@arm.com>
11715 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11716 handling COND_EXPRs with boolean comparisons, try to find a better
11717 basis for the mask type than the boolean itself.
11719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11721 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11722 is calculated and how it can be overridden.
11723 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11724 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11726 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11730 Alan Hayward <alan.hayward@arm.com>
11731 David Sherwood <david.sherwood@arm.com>
11733 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11734 Remove the mode argument.
11735 (aarch64_simd_valid_immediate): Remove the mode and inverse
11737 * config/aarch64/iterators.md (bitsize): New iterator.
11738 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11739 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11740 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11741 aarch64_simd_valid_immediate.
11742 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11743 (aarch64_reg_or_bic_imm): Likewise.
11744 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11745 with an insn_type enum and msl with a modifier_type enum.
11746 Replace element_width with a scalar_mode. Change the shift
11747 to unsigned int. Add constructors for scalar_float_mode and
11748 scalar_int_mode elements.
11749 (aarch64_vect_float_const_representable_p): Delete.
11750 (aarch64_can_const_movi_rtx_p)
11751 (aarch64_simd_scalar_immediate_valid_for_move)
11752 (aarch64_simd_make_constant): Update call to
11753 aarch64_simd_valid_immediate.
11754 (aarch64_advsimd_valid_immediate_hs): New function.
11755 (aarch64_advsimd_valid_immediate): Likewise.
11756 (aarch64_simd_valid_immediate): Remove mode and inverse
11757 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11758 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11759 and aarch64_float_const_representable_p on the result.
11760 (aarch64_output_simd_mov_immediate): Remove mode argument.
11761 Update call to aarch64_simd_valid_immediate and use of
11762 simd_immediate_info.
11763 (aarch64_output_scalar_simd_mov_immediate): Update call
11766 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11767 Alan Hayward <alan.hayward@arm.com>
11768 David Sherwood <david.sherwood@arm.com>
11770 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11771 (mode_nunits): Likewise CONST_MODE_NUNITS.
11772 * machmode.def (ADJUST_NUNITS): Document.
11773 * genmodes.c (mode_data::need_nunits_adj): New field.
11774 (blank_mode): Update accordingly.
11775 (adj_nunits): New variable.
11776 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11778 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11779 listed in adj_nunits.
11780 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11781 listed in adj_nunits. Don't emit case statements for such modes.
11782 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11783 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11784 nothing if adj_nunits is nonnull.
11785 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11786 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11787 (emit_mode_fbit): Update use of print_maybe_const_decl.
11788 (emit_move_size): Likewise. Treat the array as non-const
11790 (emit_mode_adjustments): Handle adj_nunits.
11792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11794 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11795 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11796 (VECTOR_MODES): Use it.
11797 (make_vector_modes): Take the prefix as an argument.
11799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11800 Alan Hayward <alan.hayward@arm.com>
11801 David Sherwood <david.sherwood@arm.com>
11803 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11804 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11805 for MODE_VECTOR_BOOL.
11806 * machmode.def (VECTOR_BOOL_MODE): Document.
11807 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11808 (make_vector_bool_mode): New function.
11809 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11811 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11812 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11814 * stor-layout.c (int_mode_for_mode): Likewise.
11815 * tree.c (build_vector_type_for_mode): Likewise.
11816 * varasm.c (output_constant_pool_2): Likewise.
11817 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11818 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11819 for MODE_VECTOR_BOOL.
11820 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11821 of mode class checks.
11822 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11823 instead of a list of mode class checks.
11824 (expand_vector_scalar_condition): Likewise.
11825 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11828 Alan Hayward <alan.hayward@arm.com>
11829 David Sherwood <david.sherwood@arm.com>
11831 * machmode.h (mode_size): Change from unsigned short to
11833 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11834 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11835 or if measurement_type is not polynomial.
11836 (fixed_size_mode::includes_p): Check for constant-sized modes.
11837 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11838 return a poly_uint16 rather than an unsigned short.
11839 (emit_mode_size): Change the type of mode_size from unsigned short
11840 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11841 (emit_mode_adjustments): Cope with polynomial vector sizes.
11842 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11844 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11846 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11847 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11848 * caller-save.c (setup_save_areas): Likewise.
11849 (replace_reg_with_saved_mem): Likewise.
11850 * calls.c (emit_library_call_value_1): Likewise.
11851 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11852 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11853 (gen_lowpart_for_combine): Likewise.
11854 * convert.c (convert_to_integer_1): Likewise.
11855 * cse.c (equiv_constant, cse_insn): Likewise.
11856 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11857 (cselib_subst_to_values): Likewise.
11858 * dce.c (word_dce_process_block): Likewise.
11859 * df-problems.c (df_word_lr_mark_ref): Likewise.
11860 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11861 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11862 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11863 (rtl_for_decl_location): Likewise.
11864 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11865 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11866 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11867 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11868 (expand_expr_real_1): Likewise.
11869 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11870 (pad_below): Likewise.
11871 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11872 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11873 * ira.c (get_subreg_tracking_sizes): Likewise.
11874 * ira-build.c (ira_create_allocno_objects): Likewise.
11875 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11876 (ira_sort_regnos_for_alter_reg): Likewise.
11877 * ira-costs.c (record_operand_costs): Likewise.
11878 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11879 (resolve_simple_move): Likewise.
11880 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11881 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11882 (lra_constraints): Likewise.
11883 (CONST_POOL_OK_P): Reject variable-sized modes.
11884 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11885 (add_pseudo_to_slot, lra_spill): Likewise.
11886 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11887 * optabs-query.c (get_best_extraction_insn): Likewise.
11888 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11889 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11890 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11891 * recog.c (offsettable_address_addr_space_p): Likewise.
11892 * regcprop.c (maybe_mode_change): Likewise.
11893 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11894 * regrename.c (build_def_use): Likewise.
11895 * regstat.c (dump_reg_info): Likewise.
11896 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11897 (find_reloads, find_reloads_subreg_address): Likewise.
11898 * reload1.c (eliminate_regs_1): Likewise.
11899 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11900 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11901 (simplify_binary_operation_1, simplify_subreg): Likewise.
11902 * targhooks.c (default_function_arg_padding): Likewise.
11903 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11904 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11905 (verify_gimple_assign_ternary): Likewise.
11906 * tree-inline.c (estimate_move_cost): Likewise.
11907 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11908 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11909 (get_address_cost_ainc): Likewise.
11910 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11911 (vect_supportable_dr_alignment): Likewise.
11912 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11913 (vectorizable_reduction): Likewise.
11914 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11915 (vectorizable_operation, vectorizable_load): Likewise.
11916 * tree.c (build_same_sized_truth_vector_type): Likewise.
11917 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11918 * var-tracking.c (emit_note_insn_var_location): Likewise.
11919 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11920 (ADDR_VEC_ALIGN): Likewise.
11922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11923 Alan Hayward <alan.hayward@arm.com>
11924 David Sherwood <david.sherwood@arm.com>
11926 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11928 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11929 or if measurement_type is polynomial.
11930 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11931 * combine.c (make_extraction): Likewise.
11932 * dse.c (find_shift_sequence): Likewise.
11933 * dwarf2out.c (mem_loc_descriptor): Likewise.
11934 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11935 (extract_bit_field, extract_low_bits): Likewise.
11936 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11937 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11938 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11939 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11940 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11941 * reload.c (find_reloads): Likewise.
11942 * reload1.c (alter_reg): Likewise.
11943 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11944 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11945 * tree-if-conv.c (predicate_mem_writes): Likewise.
11946 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11947 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11948 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11949 * valtrack.c (dead_debug_insert_temp): Likewise.
11950 * varasm.c (mergeable_constant_section): Likewise.
11951 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11953 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11954 Alan Hayward <alan.hayward@arm.com>
11955 David Sherwood <david.sherwood@arm.com>
11957 * expr.c (expand_assignment): Cope with polynomial mode sizes
11958 when assigning to a CONCAT.
11960 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11961 Alan Hayward <alan.hayward@arm.com>
11962 David Sherwood <david.sherwood@arm.com>
11964 * machmode.h (mode_precision): Change from unsigned short to
11966 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11968 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11969 or if measurement_type is not polynomial.
11970 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11971 in which the mode is already known to be a scalar_int_mode.
11972 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11973 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11975 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11976 for GET_MODE_PRECISION.
11977 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11978 for GET_MODE_PRECISION.
11979 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11981 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11982 (expand_field_assignment, make_extraction): Likewise.
11983 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11984 (get_last_value): Likewise.
11985 * convert.c (convert_to_integer_1): Likewise.
11986 * cse.c (cse_insn): Likewise.
11987 * expr.c (expand_expr_real_1): Likewise.
11988 * lra-constraints.c (simplify_operand_subreg): Likewise.
11989 * optabs-query.c (can_atomic_load_p): Likewise.
11990 * optabs.c (expand_atomic_load): Likewise.
11991 (expand_atomic_store): Likewise.
11992 * ree.c (combine_reaching_defs): Likewise.
11993 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11994 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11995 * tree.h (type_has_mode_precision_p): Likewise.
11996 * ubsan.c (instrument_si_overflow): Likewise.
11998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11999 Alan Hayward <alan.hayward@arm.com>
12000 David Sherwood <david.sherwood@arm.com>
12002 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12003 polynomial numbers of units.
12004 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12005 (valid_vector_subparts_p): New function.
12006 (build_vector_type): Remove temporary shim and take the number
12007 of units as a poly_uint64 rather than an int.
12008 (build_opaque_vector_type): Take the number of units as a
12009 poly_uint64 rather than an int.
12010 * tree.c (build_vector_from_ctor): Handle polynomial
12011 TYPE_VECTOR_SUBPARTS.
12012 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12013 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12014 (build_vector_from_val): If the number of units is variable,
12015 use build_vec_duplicate_cst for constant operands and
12016 VEC_DUPLICATE_EXPR otherwise.
12017 (make_vector_type): Remove temporary is_constant ().
12018 (build_vector_type, build_opaque_vector_type): Take the number of
12019 units as a poly_uint64 rather than an int.
12020 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12022 * cfgexpand.c (expand_debug_expr): Likewise.
12023 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12024 (store_constructor, expand_expr_real_1): Likewise.
12025 (const_scalar_mask_from_tree): Likewise.
12026 * fold-const-call.c (fold_const_reduction): Likewise.
12027 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12028 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12029 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12030 (fold_relational_const): Likewise.
12031 (native_interpret_vector): Likewise. Change the size from an
12032 int to an unsigned int.
12033 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12034 TYPE_VECTOR_SUBPARTS.
12035 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12036 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12037 duplicating a non-constant operand into a variable-length vector.
12038 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12039 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12040 * ipa-icf.c (sem_variable::equals): Likewise.
12041 * match.pd: Likewise.
12042 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12043 * print-tree.c (print_node): Likewise.
12044 * stor-layout.c (layout_type): Likewise.
12045 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12046 * tree-cfg.c (verify_gimple_comparison): Likewise.
12047 (verify_gimple_assign_binary): Likewise.
12048 (verify_gimple_assign_ternary): Likewise.
12049 (verify_gimple_assign_single): Likewise.
12050 * tree-pretty-print.c (dump_generic_node): Likewise.
12051 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12052 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12053 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12054 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12055 (vect_shift_permute_load_chain): Likewise.
12056 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12057 (expand_vector_condition, optimize_vector_constructor): Likewise.
12058 (lower_vec_perm, get_compute_type): Likewise.
12059 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12060 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12061 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12062 (vect_recog_mask_conversion_pattern): Likewise.
12063 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12064 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12065 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12066 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12067 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12068 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12069 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12070 (supportable_widening_operation): Likewise.
12071 (supportable_narrowing_operation): Likewise.
12072 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12074 * varasm.c (output_constant): Likewise.
12076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12077 Alan Hayward <alan.hayward@arm.com>
12078 David Sherwood <david.sherwood@arm.com>
12080 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12081 so that both the length == 3 and length != 3 cases set up their
12082 own permute vectors. Add comments explaining why we know the
12083 number of elements is constant.
12084 (vect_permute_load_chain): Likewise.
12086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12087 Alan Hayward <alan.hayward@arm.com>
12088 David Sherwood <david.sherwood@arm.com>
12090 * machmode.h (mode_nunits): Change from unsigned char to
12092 (ONLY_FIXED_SIZE_MODES): New macro.
12093 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12094 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12095 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12097 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12098 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12099 or if measurement_type is not polynomial.
12100 * genmodes.c (ZERO_COEFFS): New macro.
12101 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12103 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12104 Use ZERO_COEFFS when emitting initializers.
12105 * data-streamer.h (bp_pack_poly_value): New function.
12106 (bp_unpack_poly_value): Likewise.
12107 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12108 for GET_MODE_NUNITS.
12109 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12110 for GET_MODE_NUNITS.
12111 * tree.c (make_vector_type): Remove temporary shim and make
12112 the real function take the number of units as a poly_uint64
12113 rather than an int.
12114 (build_vector_type_for_mode): Handle polynomial nunits.
12115 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12116 * emit-rtl.c (const_vec_series_p_1): Likewise.
12117 (gen_rtx_CONST_VECTOR): Likewise.
12118 * fold-const.c (test_vec_duplicate_folding): Likewise.
12119 * genrecog.c (validate_pattern): Likewise.
12120 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12121 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12122 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12123 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12124 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12125 * rtlanal.c (subreg_get_info): Likewise.
12126 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12127 (vect_grouped_load_supported): Likewise.
12128 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12129 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12130 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12131 (simplify_const_unary_operation, simplify_binary_operation_1)
12132 (simplify_const_binary_operation, simplify_ternary_operation)
12133 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12134 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12135 instead of CONST_VECTOR_NUNITS.
12136 * varasm.c (output_constant_pool_2): Likewise.
12137 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12138 explicit-encoded elements in the XVEC for variable-length vectors.
12140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12142 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12145 Alan Hayward <alan.hayward@arm.com>
12146 David Sherwood <david.sherwood@arm.com>
12148 * coretypes.h (fixed_size_mode): Declare.
12149 (fixed_size_mode_pod): New typedef.
12150 * builtins.h (target_builtins::x_apply_args_mode)
12151 (target_builtins::x_apply_result_mode): Change type to
12152 fixed_size_mode_pod.
12153 * builtins.c (apply_args_size, apply_result_size, result_vector)
12154 (expand_builtin_apply_args_1, expand_builtin_apply)
12155 (expand_builtin_return): Update accordingly.
12157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12159 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12160 * cselib.c (cselib_hash_rtx): Likewise.
12161 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12162 CONST_VECTOR encoding.
12164 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12165 Jeff Law <law@redhat.com>
12168 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12169 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12170 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12171 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12174 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12175 explicitly probe *sp in a noreturn function if there were any callee
12176 register saves or frame pointer is needed.
12178 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12181 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12182 BLKmode for ternary, binary or unary expressions.
12185 * var-tracking.c (delete_vta_debug_insn): New inline function.
12186 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12187 insns from get_insns () to NULL instead of each bb separately.
12188 Use delete_vta_debug_insn. No longer static.
12189 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12190 delete_vta_debug_insns callers.
12191 * rtl.h (delete_vta_debug_insns): Declare.
12192 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12193 instead of variable_tracking_main.
12195 2018-01-03 Martin Sebor <msebor@redhat.com>
12197 PR tree-optimization/83603
12198 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12199 arguments past the endof the argument list in functions declared
12200 without a prototype.
12201 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12202 Avoid checking when arguments are null.
12204 2018-01-03 Martin Sebor <msebor@redhat.com>
12207 * doc/extend.texi (attribute const): Fix a typo.
12208 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12209 issuing -Wsuggest-attribute for void functions.
12211 2018-01-03 Martin Sebor <msebor@redhat.com>
12213 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12214 offset_int::from instead of wide_int::to_shwi.
12215 (maybe_diag_overlap): Remove assertion.
12216 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12217 * gimple-ssa-sprintf.c (format_directive): Same.
12218 (parse_directive): Same.
12219 (sprintf_dom_walker::compute_format_length): Same.
12220 (try_substitute_return_value): Same.
12222 2018-01-03 Jeff Law <law@redhat.com>
12224 PR middle-end/83654
12225 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12226 non-constant residual for zero at runtime and avoid probing in
12227 that case. Reorganize code for trailing problem to mirror handling
12230 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12232 PR tree-optimization/83501
12233 * tree-ssa-strlen.c (get_string_cst): New.
12234 (handle_char_store): Call get_string_cst.
12236 2018-01-03 Martin Liska <mliska@suse.cz>
12238 PR tree-optimization/83593
12239 * tree-ssa-strlen.c: Include tree-cfg.h.
12240 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12241 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12242 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12244 (strlen_dom_walker::before_dom_children): Call
12245 gimple_purge_dead_eh_edges. Dump tranformation with details
12247 (strlen_dom_walker::before_dom_children): Update call by adding
12248 new argument cleanup_eh.
12249 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12251 2018-01-03 Martin Liska <mliska@suse.cz>
12254 * cif-code.def (VARIADIC_THUNK): New enum value.
12255 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12258 2018-01-03 Jan Beulich <jbeulich@suse.com>
12260 * sse.md (mov<mode>_internal): Tighten condition for when to use
12261 vmovdqu<ssescalarsize> for TI and OI modes.
12263 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12265 Update copyright years.
12267 2018-01-03 Martin Liska <mliska@suse.cz>
12270 * ipa-visibility.c (function_and_variable_visibility): Skip
12271 functions with noipa attribure.
12273 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12275 * gcc.c (process_command): Update copyright notice dates.
12276 * gcov-dump.c (print_version): Ditto.
12277 * gcov.c (print_version): Ditto.
12278 * gcov-tool.c (print_version): Ditto.
12279 * gengtype.c (create_file): Ditto.
12280 * doc/cpp.texi: Bump @copying's copyright year.
12281 * doc/cppinternals.texi: Ditto.
12282 * doc/gcc.texi: Ditto.
12283 * doc/gccint.texi: Ditto.
12284 * doc/gcov.texi: Ditto.
12285 * doc/install.texi: Ditto.
12286 * doc/invoke.texi: Ditto.
12288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12290 * vector-builder.h (vector_builder::m_full_nelts): Change from
12291 unsigned int to poly_uint64.
12292 (vector_builder::full_nelts): Update prototype accordingly.
12293 (vector_builder::new_vector): Likewise.
12294 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12295 (vector_builder::operator ==): Likewise.
12296 (vector_builder::finalize): Likewise.
12297 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12298 Take the number of elements as a poly_uint64 rather than an
12300 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12301 from unsigned int to poly_uint64.
12302 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12303 (vec_perm_indices::new_vector): Likewise.
12304 (vec_perm_indices::length): Likewise.
12305 (vec_perm_indices::nelts_per_input): Likewise.
12306 (vec_perm_indices::input_nelts): Likewise.
12307 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12308 number of elements per input as a poly_uint64 rather than an
12309 unsigned int. Use the original encoding for variable-length
12310 vectors, rather than clamping each individual element.
12311 For the second and subsequent elements in each pattern,
12312 clamp the step and base before clamping their sum.
12313 (vec_perm_indices::series_p): Handle polynomial element counts.
12314 (vec_perm_indices::all_in_range_p): Likewise.
12315 (vec_perm_indices_to_tree): Likewise.
12316 (vec_perm_indices_to_rtx): Likewise.
12317 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12318 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12319 (tree_vector_builder::new_binary_operation): Handle polynomial
12320 element counts. Return false if we need to know the number
12321 of elements at compile time.
12322 * fold-const.c (fold_vec_perm): Punt if the number of elements
12323 isn't known at compile time.
12325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12327 * vec-perm-indices.h (vec_perm_builder): Change element type
12328 from HOST_WIDE_INT to poly_int64.
12329 (vec_perm_indices::element_type): Update accordingly.
12330 (vec_perm_indices::clamp): Handle polynomial element_types.
12331 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12332 (vec_perm_indices::all_in_range_p): Likewise.
12333 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12335 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12336 polynomial vec_perm_indices element types.
12337 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12338 * fold-const.c (fold_vec_perm): Likewise.
12339 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12340 * tree-vect-generic.c (lower_vec_perm): Likewise.
12341 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12342 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12343 element type to HOST_WIDE_INT.
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 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12350 rather than an int. Use plus_constant.
12351 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12352 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12354 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12355 Alan Hayward <alan.hayward@arm.com>
12356 David Sherwood <david.sherwood@arm.com>
12358 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12359 a HOST_WIDE_INT to a poly_int64.
12361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12362 Alan Hayward <alan.hayward@arm.com>
12363 David Sherwood <david.sherwood@arm.com>
12365 * calls.c (load_register_parameters): Cope with polynomial
12366 mode sizes. Require a constant size for BLKmode parameters
12367 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12368 forces a parameter to be padded at the lsb end in order to
12369 fill a complete number of words, require the parameter size
12370 to be ordered wrt UNITS_PER_WORD.
12372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12373 Alan Hayward <alan.hayward@arm.com>
12374 David Sherwood <david.sherwood@arm.com>
12376 * reload1.c (spill_stack_slot_width): Change element type
12377 from unsigned int to poly_uint64_pod.
12378 (alter_reg): Treat mode sizes as polynomial.
12380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12381 Alan Hayward <alan.hayward@arm.com>
12382 David Sherwood <david.sherwood@arm.com>
12384 * reload.c (complex_word_subreg_p): New function.
12385 (reload_inner_reg_of_subreg, push_reload): Use it.
12387 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12388 Alan Hayward <alan.hayward@arm.com>
12389 David Sherwood <david.sherwood@arm.com>
12391 * lra-constraints.c (process_alt_operands): Reject matched
12392 operands whose sizes aren't ordered.
12393 (match_reload): Refer to this check here.
12395 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12396 Alan Hayward <alan.hayward@arm.com>
12397 David Sherwood <david.sherwood@arm.com>
12399 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12400 that the mode size is in the set {1, 2, 4, 8, 16}.
12402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12403 Alan Hayward <alan.hayward@arm.com>
12404 David Sherwood <david.sherwood@arm.com>
12406 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12407 Use plus_constant instead of gen_rtx_PLUS.
12409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12410 Alan Hayward <alan.hayward@arm.com>
12411 David Sherwood <david.sherwood@arm.com>
12413 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12414 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12415 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12416 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12417 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12418 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12419 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12420 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12421 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12422 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12424 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12425 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12426 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12427 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12428 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12429 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12430 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12431 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12432 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12433 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12435 * expr.c (emit_move_resolve_push): Treat the input and result
12436 of PUSH_ROUNDING as a poly_int64.
12437 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12438 (emit_push_insn): Likewise.
12439 * lra-eliminations.c (mark_not_eliminable): Likewise.
12440 * recog.c (push_operand): Likewise.
12441 * reload1.c (elimination_effects): Likewise.
12442 * rtlanal.c (nonzero_bits1): Likewise.
12443 * calls.c (store_one_arg): Likewise. Require the padding to be
12444 known at compile time.
12446 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12447 Alan Hayward <alan.hayward@arm.com>
12448 David Sherwood <david.sherwood@arm.com>
12450 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12451 Use plus_constant instead of gen_rtx_PLUS.
12453 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12454 Alan Hayward <alan.hayward@arm.com>
12455 David Sherwood <david.sherwood@arm.com>
12457 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12458 rather than an int.
12460 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12461 Alan Hayward <alan.hayward@arm.com>
12462 David Sherwood <david.sherwood@arm.com>
12464 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12465 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12466 via stack temporaries. Treat the mode size as polynomial too.
12468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12469 Alan Hayward <alan.hayward@arm.com>
12470 David Sherwood <david.sherwood@arm.com>
12472 * expr.c (expand_expr_real_2): When handling conversions involving
12473 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12474 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12475 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
12488 REGMODE_NATURAL_SIZE.
12489 * expmed.c (store_bit_field_1): Likewise.
12490 * expr.c (store_constructor): Likewise.
12491 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12492 and polynomial REGMODE_NATURAL_SIZE.
12493 (gen_lowpart_common): Likewise.
12494 * reginfo.c (record_subregs_of_mode): Likewise.
12495 * rtlanal.c (read_modify_subreg_p): Likewise.
12497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12498 Alan Hayward <alan.hayward@arm.com>
12499 David Sherwood <david.sherwood@arm.com>
12501 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12502 numbers of elements.
12504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12505 Alan Hayward <alan.hayward@arm.com>
12506 David Sherwood <david.sherwood@arm.com>
12508 * match.pd: Cope with polynomial numbers of vector elements.
12510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12511 Alan Hayward <alan.hayward@arm.com>
12512 David Sherwood <david.sherwood@arm.com>
12514 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12515 in a POINTER_PLUS_EXPR.
12517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12518 Alan Hayward <alan.hayward@arm.com>
12519 David Sherwood <david.sherwood@arm.com>
12521 * omp-simd-clone.c (simd_clone_subparts): New function.
12522 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12523 (ipa_simd_modify_function_body): Likewise.
12525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12526 Alan Hayward <alan.hayward@arm.com>
12527 David Sherwood <david.sherwood@arm.com>
12529 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12530 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12531 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12532 (expand_vector_condition, vector_element): Likewise.
12533 (subparts_gt): New function.
12534 (get_compute_type): Use subparts_gt.
12535 (count_type_subparts): Delete.
12536 (expand_vector_operations_1): Use subparts_gt instead of
12537 count_type_subparts.
12539 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12540 Alan Hayward <alan.hayward@arm.com>
12541 David Sherwood <david.sherwood@arm.com>
12543 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12544 (vect_compile_time_alias): ...this new function. Do the calculation
12545 on poly_ints rather than trees.
12546 (vect_prune_runtime_alias_test_list): Update call accordingly.
12548 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12549 Alan Hayward <alan.hayward@arm.com>
12550 David Sherwood <david.sherwood@arm.com>
12552 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12554 (vect_schedule_slp_instance): Likewise.
12556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12557 Alan Hayward <alan.hayward@arm.com>
12558 David Sherwood <david.sherwood@arm.com>
12560 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12561 constant and extern definitions for variable-length vectors.
12562 (vect_get_constant_vectors): Note that the number of units
12563 is known to be constant.
12565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12566 Alan Hayward <alan.hayward@arm.com>
12567 David Sherwood <david.sherwood@arm.com>
12569 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12570 of units as polynomial. Choose between WIDE and NARROW based
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-stmts.c (simd_clone_subparts): New function.
12578 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12581 Alan Hayward <alan.hayward@arm.com>
12582 David Sherwood <david.sherwood@arm.com>
12584 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12585 vectors as polynomial. Use build_index_vector for
12586 IFN_GOMP_SIMD_LANE.
12588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12589 Alan Hayward <alan.hayward@arm.com>
12590 David Sherwood <david.sherwood@arm.com>
12592 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12593 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12594 for variable-length vectors.
12595 (vectorizable_mask_load_store): Treat the number of units as
12596 polynomial, asserting that it is constant if the condition has
12597 already been enforced.
12598 (vectorizable_store, vectorizable_load): Likewise.
12600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12601 Alan Hayward <alan.hayward@arm.com>
12602 David Sherwood <david.sherwood@arm.com>
12604 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12605 of units as polynomial. Punt if we can't tell at compile time
12606 which vector contains the final result.
12608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12609 Alan Hayward <alan.hayward@arm.com>
12610 David Sherwood <david.sherwood@arm.com>
12612 * tree-vect-loop.c (vectorizable_induction): Treat the number
12613 of units as polynomial. Punt on SLP inductions. Use an integer
12614 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12615 cast of such a series for variable-length floating-point
12618 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12619 Alan Hayward <alan.hayward@arm.com>
12620 David Sherwood <david.sherwood@arm.com>
12622 * tree.h (build_index_vector): Declare.
12623 * tree.c (build_index_vector): New function.
12624 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12625 of units as polynomial, forcibly converting it to a constant if
12626 vectorizable_reduction has already enforced the condition.
12627 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12628 to create a {1,2,3,...} vector.
12629 (vectorizable_reduction): Treat the number of units as polynomial.
12630 Choose vectype_in based on the largest scalar element size rather
12631 than the smallest number of units. Enforce the restrictions
12634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12635 Alan Hayward <alan.hayward@arm.com>
12636 David Sherwood <david.sherwood@arm.com>
12638 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12639 number of units as polynomial.
12641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12642 Alan Hayward <alan.hayward@arm.com>
12643 David Sherwood <david.sherwood@arm.com>
12645 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12646 * target.def (autovectorize_vector_sizes): Return the vector sizes
12647 by pointer, using vector_sizes rather than a bitmask.
12648 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12649 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12650 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12652 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12653 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12654 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12655 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12656 * omp-general.c (omp_max_vf): Likewise.
12657 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12658 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12659 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12660 * tree-vect-slp.c (vect_slp_bb): Likewise.
12661 * doc/tm.texi: Regenerate.
12662 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12664 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12665 the vector size as a poly_uint64 rather than an unsigned int.
12666 (current_vector_size): Change from an unsigned int to a poly_uint64.
12667 (get_vectype_for_scalar_type): Update accordingly.
12668 * tree.h (build_truth_vector_type): Take the size and number of
12669 units as a poly_uint64 rather than an unsigned int.
12670 (build_vector_type): Add a temporary overload that takes
12671 the number of units as a poly_uint64 rather than an unsigned int.
12672 * tree.c (make_vector_type): Likewise.
12673 (build_truth_vector_type): Take the number of units as a poly_uint64
12674 rather than an unsigned int.
12676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12677 Alan Hayward <alan.hayward@arm.com>
12678 David Sherwood <david.sherwood@arm.com>
12680 * target.def (get_mask_mode): Take the number of units and length
12681 as poly_uint64s rather than unsigned ints.
12682 * targhooks.h (default_get_mask_mode): Update accordingly.
12683 * targhooks.c (default_get_mask_mode): Likewise.
12684 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12685 * doc/tm.texi: Regenerate.
12687 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12688 Alan Hayward <alan.hayward@arm.com>
12689 David Sherwood <david.sherwood@arm.com>
12691 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12692 * omp-general.c (omp_max_vf): Likewise.
12693 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12694 (expand_omp_simd): Handle polynomial safelen.
12695 * omp-low.c (omplow_simd_context): Add a default constructor.
12696 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12697 (lower_rec_simd_input_clauses): Update accordingly.
12698 (lower_rec_input_clauses): Likewise.
12700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12701 Alan Hayward <alan.hayward@arm.com>
12702 David Sherwood <david.sherwood@arm.com>
12704 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12705 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12706 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12707 (vect_analyze_slp_cost): Likewise.
12708 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12709 (vect_model_load_cost): Likewise.
12711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12712 Alan Hayward <alan.hayward@arm.com>
12713 David Sherwood <david.sherwood@arm.com>
12715 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12716 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12717 from an unsigned int * to a poly_uint64_pod *.
12718 (calculate_unrolling_factor): New function.
12719 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12722 Alan Hayward <alan.hayward@arm.com>
12723 David Sherwood <david.sherwood@arm.com>
12725 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12726 from an unsigned int to a poly_uint64.
12727 (_loop_vec_info::slp_unrolling_factor): Likewise.
12728 (_loop_vec_info::vectorization_factor): Change from an int
12730 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12731 (vect_get_num_vectors): New function.
12732 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12733 (vect_get_num_copies): Use vect_get_num_vectors.
12734 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12735 to an unsigned int *.
12736 (vect_analyze_data_refs): Change min_vf from an int * to a
12738 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12739 than an unsigned HOST_WIDE_INT.
12740 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12741 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12742 to an unsigned int *.
12743 (vect_analyze_data_ref_dependences): Likewise.
12744 (vect_compute_data_ref_alignment): Handle polynomial vf.
12745 (vect_enhance_data_refs_alignment): Likewise.
12746 (vect_prune_runtime_alias_test_list): Likewise.
12747 (vect_shift_permute_load_chain): Likewise.
12748 (vect_supportable_dr_alignment): Likewise.
12749 (dependence_distance_ge_vf): Take the vectorization factor as a
12750 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12751 (vect_analyze_data_refs): Change min_vf from an int * to a
12753 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12754 vfm1 as a poly_uint64 rather than an int. Make the same change
12755 for the returned bound_scalar.
12756 (vect_gen_vector_loop_niters): Handle polynomial vf.
12757 (vect_do_peeling): Likewise. Update call to
12758 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12759 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12761 * tree-vect-loop.c (vect_determine_vectorization_factor)
12762 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12763 (vect_get_known_peeling_cost): Likewise.
12764 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12765 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12766 (vect_transform_loop): Likewise. Use the lowest possible VF when
12767 updating the upper bounds of the loop.
12768 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12769 rather than an int.
12770 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12771 polynomial unroll factors.
12772 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12773 (vect_make_slp_decision): Likewise.
12774 (vect_supported_load_permutation_p): Likewise, and polynomial
12776 (vect_analyze_slp_cost): Handle polynomial vf.
12777 (vect_slp_analyze_node_operations): Likewise.
12778 (vect_slp_analyze_bb_1): Likewise.
12779 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12780 than an unsigned HOST_WIDE_INT.
12781 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12782 (vectorizable_load): Handle polynomial vf.
12783 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12785 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12788 Alan Hayward <alan.hayward@arm.com>
12789 David Sherwood <david.sherwood@arm.com>
12791 * match.pd: Handle bit operations involving three constants
12792 and try to fold one pair.
12794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12796 * tree-vect-loop-manip.c: Include gimple-fold.h.
12797 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12798 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12799 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12800 Add a path that uses a step of VF instead of 1, but disable it
12802 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12803 and niters_no_overflow parameters. Update calls to
12804 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12805 Create a new SSA name if the latter choses to use a ste other
12806 than zero, and return it via niters_vector_mult_vf_var.
12807 * tree-vect-loop.c (vect_transform_loop): Update calls to
12808 vect_do_peeling, vect_gen_vector_loop_niters and
12809 slpeel_make_loop_iterate_ntimes.
12810 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12811 (vect_gen_vector_loop_niters): Update declarations after above changes.
12813 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12815 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12816 128-bit round to integer instructions.
12817 (ceil<mode>2): Likewise.
12818 (btrunc<mode>2): Likewise.
12819 (round<mode>2): Likewise.
12821 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12823 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12824 unaligned VSX load/store on P8/P9.
12825 (expand_block_clear): Allow the use of unaligned VSX
12826 load/store on P8/P9.
12828 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12830 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12832 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12833 swap associated with both a load and a store.
12835 2018-01-02 Andrew Waterman <andrew@sifive.com>
12837 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12838 * config/riscv/riscv.md (clear_cache): Use it.
12840 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12842 * web.c: Remove out-of-date comment.
12844 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12846 * expr.c (fixup_args_size_notes): Check that any existing
12847 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12848 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12849 (emit_single_push_insn): ...here.
12851 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12853 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12854 (const_vector_encoded_nelts): New function.
12855 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12856 (const_vector_int_elt, const_vector_elt): Declare.
12857 * emit-rtl.c (const_vector_int_elt_1): New function.
12858 (const_vector_elt): Likewise.
12859 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12860 of CONST_VECTOR_ELT.
12862 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12864 * expr.c: Include rtx-vector-builder.h.
12865 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12866 directly on the tree encoding.
12867 (const_vector_from_tree): Likewise.
12868 * optabs.c: Include rtx-vector-builder.h.
12869 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12870 sequence of "u" values.
12871 * vec-perm-indices.c: Include rtx-vector-builder.h.
12872 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12873 directly on the vec_perm_indices encoding.
12875 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12877 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12878 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12879 * rtx-vector-builder.h: New file.
12880 * rtx-vector-builder.c: Likewise.
12881 * rtl.h (rtx_def::u2): Add a const_vector field.
12882 (CONST_VECTOR_NPATTERNS): New macro.
12883 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12884 (CONST_VECTOR_DUPLICATE_P): Likewise.
12885 (CONST_VECTOR_STEPPED_P): Likewise.
12886 (CONST_VECTOR_ENCODED_ELT): Likewise.
12887 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12888 (unwrap_const_vec_duplicate): Likewise.
12889 (const_vec_series_p): Check for a non-duplicated vector encoding.
12890 Say that the function only returns true for integer vectors.
12891 * emit-rtl.c: Include rtx-vector-builder.h.
12892 (gen_const_vec_duplicate_1): Delete.
12893 (gen_const_vector): Call gen_const_vec_duplicate instead of
12894 gen_const_vec_duplicate_1.
12895 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12896 (gen_const_vec_duplicate): Use rtx_vector_builder.
12897 (gen_const_vec_series): Likewise.
12898 (gen_rtx_CONST_VECTOR): Likewise.
12899 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12900 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12901 Build a new vector rather than modifying a CONST_VECTOR in-place.
12902 (handle_special_swappables): Update call accordingly.
12903 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12904 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12905 Build a new vector rather than modifying a CONST_VECTOR in-place.
12906 (handle_special_swappables): Update call accordingly.
12908 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12910 * simplify-rtx.c (simplify_const_binary_operation): Use
12911 CONST_VECTOR_ELT instead of XVECEXP.
12913 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12915 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12916 the selector elements to be different from the data elements
12917 if the selector is a VECTOR_CST.
12918 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12919 ssizetype for the selector.
12921 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12923 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12924 before testing each element individually.
12925 * tree-vect-generic.c (lower_vec_perm): Likewise.
12927 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12929 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12930 * selftest-run-tests.c (selftest::run_tests): Call it.
12931 * vector-builder.h (vector_builder::operator ==): New function.
12932 (vector_builder::operator !=): Likewise.
12933 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12934 (vec_perm_indices::all_from_input_p): New function.
12935 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12936 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12937 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12938 instead of reading the VECTOR_CST directly. Detect whether both
12939 vector inputs are the same before constructing the vec_perm_indices,
12940 and update the number of inputs argument accordingly. Use the
12941 utility functions added above. Only construct sel2 if we need to.
12943 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12945 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12946 the broadcast of the low byte.
12947 (expand_mult_highpart): Use an explicit encoding for the permutes.
12948 * optabs-query.c (can_mult_highpart_p): Likewise.
12949 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12950 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12951 (vectorizable_bswap): Likewise.
12952 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12953 explicit encoding for the power-of-2 permutes.
12954 (vect_permute_store_chain): Likewise.
12955 (vect_grouped_load_supported): Likewise.
12956 (vect_permute_load_chain): Likewise.
12958 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12960 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12961 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12962 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12963 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12964 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12965 (vect_gen_perm_mask_any): Likewise.
12967 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12969 * int-vector-builder.h: New file.
12970 * vec-perm-indices.h: Include int-vector-builder.h.
12971 (vec_perm_indices): Redefine as an int_vector_builder.
12972 (auto_vec_perm_indices): Delete.
12973 (vec_perm_builder): Redefine as a stand-alone class.
12974 (vec_perm_indices::vec_perm_indices): New function.
12975 (vec_perm_indices::clamp): Likewise.
12976 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12977 (vec_perm_indices::new_vector): New function.
12978 (vec_perm_indices::new_expanded_vector): Update for new
12979 vec_perm_indices class.
12980 (vec_perm_indices::rotate_inputs): New function.
12981 (vec_perm_indices::all_in_range_p): Operate directly on the
12982 encoded form, without computing elided elements.
12983 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12984 encoding. Update for new vec_perm_indices class.
12985 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12986 the given vec_perm_builder.
12987 (expand_vec_perm_var): Update vec_perm_builder constructor.
12988 (expand_mult_highpart): Use vec_perm_builder instead of
12989 auto_vec_perm_indices.
12990 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12991 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12992 or double series encoding as appropriate.
12993 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12994 vec_perm_indices instead of auto_vec_perm_indices.
12995 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12996 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12997 (vect_permute_store_chain): Likewise.
12998 (vect_grouped_load_supported): Likewise.
12999 (vect_permute_load_chain): Likewise.
13000 (vect_shift_permute_load_chain): Likewise.
13001 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13002 (vect_transform_slp_perm_load): Likewise.
13003 (vect_schedule_slp_instance): Likewise.
13004 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13005 (vectorizable_mask_load_store): Likewise.
13006 (vectorizable_bswap): Likewise.
13007 (vectorizable_store): Likewise.
13008 (vectorizable_load): Likewise.
13009 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13010 vec_perm_indices instead of auto_vec_perm_indices. Use
13011 tree_to_vec_perm_builder to read the vector from a tree.
13012 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13013 vec_perm_builder instead of a vec_perm_indices.
13014 (have_whole_vector_shift): Use vec_perm_builder and
13015 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13016 truncation to calc_vec_perm_mask_for_shift.
13017 (vect_create_epilog_for_reduction): Likewise.
13018 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13019 from auto_vec_perm_indices to vec_perm_indices.
13020 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13021 instead of changing individual elements.
13022 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13023 the vector in d.perm.
13024 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13025 from auto_vec_perm_indices to vec_perm_indices.
13026 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13027 instead of changing individual elements.
13028 (arm_vectorize_vec_perm_const): Use new_vector to install
13029 the vector in d.perm.
13030 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13031 Update vec_perm_builder constructor.
13032 (rs6000_expand_interleave): Likewise.
13033 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13034 (rs6000_expand_interleave): Likewise.
13036 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13038 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13039 to qimode could truncate the indices.
13040 * optabs.c (expand_vec_perm_var): Likewise.
13042 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13044 * Makefile.in (OBJS): Add vec-perm-indices.o.
13045 * vec-perm-indices.h: New file.
13046 * vec-perm-indices.c: Likewise.
13047 * target.h (vec_perm_indices): Replace with a forward class
13049 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13050 * optabs.h: Include vec-perm-indices.h.
13051 (expand_vec_perm): Delete.
13052 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13053 (expand_vec_perm_const): Declare.
13054 * target.def (vec_perm_const_ok): Replace with...
13055 (vec_perm_const): ...this new hook.
13056 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13057 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13058 * doc/tm.texi: Regenerate.
13059 * optabs.def (vec_perm_const): Delete.
13060 * doc/md.texi (vec_perm_const): Likewise.
13061 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13062 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13063 expand_vec_perm for constant permutation vectors. Assert that
13064 the mode of variable permutation vectors is the integer equivalent
13065 of the mode that is being permuted.
13066 * optabs-query.h (selector_fits_mode_p): Declare.
13067 * optabs-query.c: Include vec-perm-indices.h.
13068 (selector_fits_mode_p): New function.
13069 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13070 is defined, instead of checking whether the vec_perm_const_optab
13071 exists. Use targetm.vectorize.vec_perm_const instead of
13072 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13073 fit in the vector mode before using a variable permute.
13074 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13075 vec_perm_indices instead of an rtx.
13076 (expand_vec_perm): Replace with...
13077 (expand_vec_perm_const): ...this new function. Take the selector
13078 as a vec_perm_indices rather than an rtx. Also take the mode of
13079 the selector. Update call to shift_amt_for_vec_perm_mask.
13080 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13081 Use vec_perm_indices::new_expanded_vector to expand the original
13082 selector into bytes. Check whether the indices fit in the vector
13083 mode before using a variable permute.
13084 (expand_vec_perm_var): Make global.
13085 (expand_mult_highpart): Use expand_vec_perm_const.
13086 * fold-const.c: Includes vec-perm-indices.h.
13087 * tree-ssa-forwprop.c: Likewise.
13088 * tree-vect-data-refs.c: Likewise.
13089 * tree-vect-generic.c: Likewise.
13090 * tree-vect-loop.c: Likewise.
13091 * tree-vect-slp.c: Likewise.
13092 * tree-vect-stmts.c: Likewise.
13093 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13095 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13096 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13097 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13098 (aarch64_vectorize_vec_perm_const): ...this new function.
13099 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13100 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13101 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13102 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13103 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13104 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13105 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13107 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13108 check for NEON modes.
13109 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13110 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13111 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13112 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13114 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13115 the old VEC_PERM_CONST conditions.
13116 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13117 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13118 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13119 (ia64_vectorize_vec_perm_const_ok): Merge into...
13120 (ia64_vectorize_vec_perm_const): ...this new function.
13121 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13122 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13123 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13124 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13125 * config/mips/mips.c (mips_expand_vec_perm_const)
13126 (mips_vectorize_vec_perm_const_ok): Merge into...
13127 (mips_vectorize_vec_perm_const): ...this new function.
13128 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13129 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13130 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13131 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13132 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13133 (rs6000_expand_vec_perm_const): Delete.
13134 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13136 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13137 (altivec_expand_vec_perm_const_le): Take each operand individually.
13138 Operate on constant selectors rather than rtxes.
13139 (altivec_expand_vec_perm_const): Likewise. Update call to
13140 altivec_expand_vec_perm_const_le.
13141 (rs6000_expand_vec_perm_const): Delete.
13142 (rs6000_vectorize_vec_perm_const_ok): Delete.
13143 (rs6000_vectorize_vec_perm_const): New function.
13144 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13145 an element count and rtx array.
13146 (rs6000_expand_extract_even): Update call accordingly.
13147 (rs6000_expand_interleave): Likewise.
13148 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13149 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13150 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13151 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13152 (rs6000_expand_vec_perm_const): Delete.
13153 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13154 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13155 (altivec_expand_vec_perm_const_le): Take each operand individually.
13156 Operate on constant selectors rather than rtxes.
13157 (altivec_expand_vec_perm_const): Likewise. Update call to
13158 altivec_expand_vec_perm_const_le.
13159 (rs6000_expand_vec_perm_const): Delete.
13160 (rs6000_vectorize_vec_perm_const_ok): Delete.
13161 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13162 reference to the SPE evmerge intructions.
13163 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13164 an element count and rtx array.
13165 (rs6000_expand_extract_even): Update call accordingly.
13166 (rs6000_expand_interleave): Likewise.
13167 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13168 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13170 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13172 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13174 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13175 vector mode and that that mode matches the mode of the data
13177 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13178 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13179 directly using expand_vec_perm_1 when forcing selectors into
13181 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13183 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13185 * optabs-query.h (can_vec_perm_p): Delete.
13186 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13187 * optabs-query.c (can_vec_perm_p): Split into...
13188 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13189 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13190 particular selector is valid.
13191 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13192 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13193 (vect_grouped_load_supported): Likewise.
13194 (vect_shift_permute_load_chain): Likewise.
13195 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13196 (vect_transform_slp_perm_load): Likewise.
13197 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13198 (vectorizable_bswap): Likewise.
13199 (vect_gen_perm_mask_checked): Likewise.
13200 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13201 implementations of variable permutation vectors into account
13202 when deciding which selector to use.
13203 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13204 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13205 with a false third argument.
13206 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13207 to test whether the constant selector is valid and can_vec_perm_var_p
13208 to test whether a variable selector is valid.
13210 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13212 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13213 * optabs-query.c (can_vec_perm_p): Likewise.
13214 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13215 instead of vec_perm_indices.
13216 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13217 (vect_gen_perm_mask_checked): Likewise,
13218 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13219 (vect_gen_perm_mask_checked): Likewise,
13221 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13223 * optabs-query.h (qimode_for_vec_perm): Declare.
13224 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13225 (qimode_for_vec_perm): ...this new function.
13226 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13228 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13230 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13231 does not have a conditional at the top.
13233 2018-01-02 Richard Biener <rguenther@suse.de>
13235 * ipa-inline.c (big_speedup_p): Fix expression.
13237 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13240 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13243 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13247 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13248 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13249 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13250 cond_taken_branch_cost 3->4.
13252 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13254 PR tree-optimization/83581
13255 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13256 TODO_cleanup_cfg if any changes have been made.
13258 PR middle-end/83608
13259 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13260 convert_modes if target mode has the right side, but different mode
13263 PR middle-end/83609
13264 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13265 last argument when extracting from CONCAT. If either from_real or
13266 from_imag is NULL, use expansion through memory. If result is not
13267 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13268 the parts directly to inner mode, if even that fails, use expansion
13271 PR middle-end/83623
13272 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13273 check for bswap in mode rather than HImode and use that in expand_unop
13276 Copyright (C) 2018 Free Software Foundation, Inc.
13278 Copying and distribution of this file, with or without modification,
13279 are permitted in any medium without royalty provided the copyright
13280 notice and this notice are preserved.