1 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
3 * doc/sourcebuild.texi (scalar_all_fma): Document.
4 * tree.def (FMA_EXPR): Delete.
5 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
6 * internal-fn.c (ternary_direct): New macro.
7 (expand_ternary_optab_fn): Likewise.
8 (direct_ternary_optab_supported_p): Likewise.
9 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
10 * builtins.c (fold_builtin_fma): Delete.
11 (fold_builtin_3): Don't call it.
12 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
13 * expr.c (expand_expr_real_2): Likewise.
14 * fold-const.c (operand_equal_p): Likewise.
15 (fold_ternary_loc): Likewise.
16 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
17 * gimple.c (DEFTREECODE): Likewise.
18 * gimplify.c (gimplify_expr): Likewise.
19 * optabs-tree.c (optab_for_tree_code): Likewise.
20 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
21 * tree-eh.c (operation_could_trap_p): Likewise.
22 (stmt_could_throw_1_p): Likewise.
23 * tree-inline.c (estimate_operator_cost): Likewise.
24 * tree-pretty-print.c (dump_generic_node): Likewise.
25 (op_code_prio): Likewise.
26 * tree-ssa-loop-im.c (stmt_cost): Likewise.
27 * tree-ssa-operands.c (get_expr_operands): Likewise.
28 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
29 * fold-const-call.h (fold_fma): Delete.
30 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
31 CFN_FNMA and CFN_FNMS.
33 * genmatch.c (combined_fn): New enum.
34 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
35 (commutative_op): New function.
36 (commutate): Use it. Handle more than 2 operands.
37 (dt_operand::gen_gimple_expr): Use commutative_op.
38 (parser::parse_expr): Allow :c to be used with non-binary
39 operators if the commutative operand is known.
40 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
41 CFN_FMS, CFN_FNMA and CFN_FNMS.
42 (backprop::process_assign_use): Remove FMA_EXPR handling.
43 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
44 (gen_hsa_fma): New function.
45 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
46 IFN_FNMA and IFN_FNMS.
47 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
48 * gimple-fold.h (follow_all_ssa_edges): Declare.
49 * gimple-fold.c (follow_all_ssa_edges): New function.
50 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
51 gimple_build interface and use follow_all_ssa_edges to fold the result.
52 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
53 instead of checking for optabs directly.
54 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
55 rather than FMA_EXPRs.
56 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
57 call to IFN_FMA instead of an FMA_EXPR.
59 2018-05-17 Jim Wilson <jimw@sifive.com>
61 * expr.c (do_tablejump): When converting index to Pmode, if we have a
62 sign extended promoted subreg, and the range does not have the sign bit
63 set, then do a sign extend.
65 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
66 test, check for sign extended subreg and/or constant operands, and
67 do a sign extend in that case.
69 2018-05-17 Steve Ellcey <sellcey@cavium.com>
71 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
72 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
74 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
75 Change logics_shift_reg to logics_shift_imm.
76 (thunderx2t99_fp_loadpair_basic): Delete.
77 (thunderx2t99_fp_storepair_basic): Delete.
78 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
79 (thunderx2t99_asimd_polynomial): Delete.
80 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
81 and neon_fp_mul_d_scalar_q.
82 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
83 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
84 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
85 (thunderx2t99_asimd_lut): Add missing tbl types.
86 (thunderx2t99_asimd_ext): Delete.
87 (thunderx2t99_asimd_load1_1_mult): Delete.
88 (thunderx2t99_asimd_load1_2_mult): Delete.
89 (thunderx2t99_asimd_load1_ldp): New.
90 (thunderx2t99_asimd_load1): New.
91 (thunderx2t99_asimd_load2): Add missing *load2* types.
92 (thunderx2t99_asimd_load3): New.
93 (thunderx2t99_asimd_load4): New.
94 (thunderx2t99_asimd_store1_1_mult): Delete.
95 (thunderx2t99_asimd_store1_2_mult): Delete.
96 (thunderx2t99_asimd_store2_mult): Delete.
97 (thunderx2t99_asimd_store2_onelane): Delete.
98 (thunderx2t99_asimd_store_stp): New.
99 (thunderx2t99_asimd_store1): New.
100 (thunderx2t99_asimd_store2): New.
101 (thunderx2t99_asimd_store3): New.
102 (thunderx2t99_asimd_store4): New.
104 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
106 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
107 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
109 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
110 Segher Boessenkool <segher@kernel.crashing.org>
113 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
116 2018-05-17 Richard Biener <rguenther@suse.de>
118 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
119 for pruning loop and prune defs feeding only already visited PHIs.
121 2018-05-17 Richard Biener <rguenther@suse.de>
123 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
125 2018-05-17 Bin Cheng <bin.cheng@arm.com>
126 Richard Biener <rguenther@suse.de>
128 PR tree-optimization/85793
129 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
130 for VMAT_ELEMENTWISE.
132 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
134 * internal-fn.h (lookup_internal_fn): Declare
135 * internal-fn.c (lookup_internal_fn): New function.
136 * gimple.c (gimple_build_call_from_tree): Handle calls to
138 * gimple-pretty-print.c (dump_gimple_call): Print "." before
139 internal function names.
140 * tree-pretty-print.c (dump_generic_node): Likewise.
141 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
143 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
145 * gimple-fold.h (gimple_build): Make the function forms take
146 combined_fn rather than built_in_function.
147 (gimple_simplify): Likewise.
148 * gimple-match-head.c (gimple_simplify): Likewise.
149 * gimple-fold.c (gimple_build): Likewise.
150 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
151 rather than gimple_build_call_internal.
152 (get_initial_defs_for_reduction): Likewise.
153 (vect_create_epilog_for_reduction): Likewise.
154 (vectorizable_live_operation): Likewise.
156 2018-05-17 Martin Liska <mliska@suse.cz>
158 * gimple-ssa-sprintf.c (format_directive): Do not use
159 space in between 'G_' and '('.
161 2018-05-17 Jakub Jelinek <jakub@redhat.com>
164 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
165 even if the mask is not all ones.
168 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
170 (ix86_gimple_fold_builtin): Likewise.
173 * config/i386/i386.c: Include tree-vector-builder.h.
174 (ix86_vector_shift_count): New function.
175 (ix86_fold_builtin): Fold shift builtins by scalar count.
176 (ix86_gimple_fold_builtin): Likewise.
178 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
179 _mm512_setzero): New intrinsics.
181 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
182 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
184 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
185 code generation for cases where splatting a value is not useful.
186 * simplify-rtx.c (simplify_ternary_operation): Simplify
187 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
188 mode to a vec_concat.
190 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
192 * config.gcc: Support "goldmont-plus".
193 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
195 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
196 PROCESSOR_GOLDMONT_PLUS.
197 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
198 (processor_target_table): Add "goldmont-plus".
199 (PTA_GOLDMONT_PLUS): Define.
200 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
201 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
202 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
203 (fold_builtin_cpu): Add "goldmont-plus".
204 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
205 (ix86_option_override_internal): Add "goldmont-plus".
206 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
207 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
208 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
209 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
211 2018-05-17 Richard Biener <rguenther@suse.de>
213 PR tree-optimization/85757
214 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
215 remove defs that only feed that PHI from further processing.
217 2018-05-16 Jim Wilson <jimw@sifive.com>
219 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
221 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
223 2018-05-16 Mark Wielaard <mark@klomp.org>
225 * dwarf2out.c (count_index_strings): New function.
226 (output_indirect_strings): Call count_index_strings and generate
227 header for dwarf_version >= 5.
229 2018-05-16 Mark Wielaard <mark@klomp.org>
231 * dwarf2out.c (dwarf_FORM): New function.
232 (set_indirect_string): Use dwarf_FORM.
233 (reset_indirect_string): Likewise.
234 (size_of_die): Likewise.
235 (value_format): Likewise.
236 (output_die): Likewise.
237 (add_skeleton_AT_string): Likewise.
238 (output_macinfo_op): Likewise.
239 (index_string): Likewise.
240 (output_index_string_offset): Likewise.
241 (output_index_string): Likewise.
242 (count_index_strings): Likewise.
244 2018-05-16 Carl Love <cel@us.ibm.com>
246 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
247 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
249 2018-05-16 Martin Jambor <mjambor@suse.cz>
251 * ipa-prop.c (ipa_free_all_edge_args): Remove.
252 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
254 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
256 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
257 (fnma<mode>4): Likewise.
258 (fms<mode>4): Likewise.
259 (fnms<mode>4): Likewise.
260 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
261 (aarch64_fnma<mode>4): Likewise.
262 (aarch64_fms<mode>4): Likewise.
263 (aarch64_fnms<mode>4): Likewise.
264 (aarch64_fnmadd<mode>4): Likewise.
266 2018-05-16 Jason Merrill <jason@redhat.com>
268 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
270 2018-05-16 Richard Biener <rguenther@suse.de>
272 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
273 (dump_stmt_cost): Declare.
274 (add_stmt_cost): Dump cost we add.
275 (add_stmt_costs): New function.
276 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
278 (vect_analyze_stmt): Adjust prototype.
279 (vectorizable_condition): Likewise.
280 (vectorizable_live_operation): Likewise.
281 (vectorizable_reduction): Likewise.
282 (vectorizable_induction): Likewise.
283 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
284 cost vector to pass to vectorizable_ and record afterwards.
285 (vect_model_reduction_cost): Take cost vector argument and adjust.
286 (vect_model_induction_cost): Likewise.
287 (vectorizable_reduction): Likewise.
288 (vectorizable_induction): Likewise.
289 (vectorizable_live_operation): Likewise.
290 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
291 SLP_TREE_NUMBER_OF_VEC_STMTS.
292 (vect_analyze_slp_cost_1): Remove.
293 (vect_analyze_slp_cost): Likewise.
294 (vect_slp_analyze_node_operations): Take visited args and
295 a target cost vector. Avoid processing already visited stmt sets.
296 (vect_slp_analyze_operations): Use a local cost vector to gather
297 costs and register those of non-discarded instances.
298 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
299 (vect_schedule_slp_instance): Remove copying of
300 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
302 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
303 adding cost. Record cost entry location.
304 (vect_prologue_cost_for_slp_op): Function to compute cost of
305 a constant or invariant generated for SLP vect in the prologue,
306 split out from vect_analyze_slp_cost_1.
307 (vect_model_simple_cost): Make static. Adjust for SLP costing.
308 (vect_model_promotion_demotion_cost): Likewise.
309 (vect_model_store_cost): Likewise, make static.
310 (vect_model_load_cost): Likewise.
311 (vectorizable_bswap): Add cost vector arg and adjust.
312 (vectorizable_call): Likewise.
313 (vectorizable_simd_clone_call): Likewise.
314 (vectorizable_conversion): Likewise.
315 (vectorizable_assignment): Likewise.
316 (vectorizable_shift): Likewise.
317 (vectorizable_operation): Likewise.
318 (vectorizable_store): Likewise.
319 (vectorizable_load): Likewise.
320 (vectorizable_condition): Likewise.
321 (vectorizable_comparison): Likewise.
322 (can_vectorize_live_stmts): Likewise.
323 (vect_analyze_stmt): Likewise.
324 (vect_transform_stmt): Adjust calls to vectorizable_*.
325 * tree-vectorizer.c: Include gimple-pretty-print.h.
326 (dump_stmt_cost): New function.
328 2018-05-16 Richard Biener <rguenther@suse.de>
330 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
331 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
332 * tree-ssa-dse.c: Include tree-ssa-loop.h.
333 (check_name): New callback.
334 (dse_classify_store): Track cycles via a visited bitmap of PHI
335 defs and simplify handling of in-loop and across loop dead stores
336 and properly fail for loop-variant refs. Handle byte-tracking with
337 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
340 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
342 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
343 (vect_get_mask_type_for_stmt): Likewise.
344 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
346 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
347 to determine the statement's vector type and the vector type that
348 should be used for calculating nunits. Deal with cases in which
349 the type has to be deferred.
350 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
351 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
352 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
353 (vect_determine_vf_for_stmt): New functions, split out from...
354 (vect_determine_vectorization_factor): ...here.
355 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
356 (vect_get_mask_type_for_stmt): New functions, split out from
357 vect_determine_vectorization_factor.
359 2018-05-16 Richard Biener <rguenther@suse.de>
361 * tree-cfg.c (verify_gimple_assign_ternary): Properly
362 verify the [VEC_]COND_EXPR embedded comparison.
364 2018-05-15 Martin Sebor <msebor@redhat.com>
366 PR tree-optimization/85753
367 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
368 RECORD_TYPE in addition to ARRAY_TYPE.
370 2018-05-15 Martin Sebor <msebor@redhat.com>
373 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
375 2018-05-15 Richard Biener <rguenther@suse.de>
377 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
378 add by_clobber_p one. Change algorithm to collect all defs
379 representing uses we need to walk and try reducing them to
380 a single one before failing.
381 (dse_dom_walker::dse_optimize_stmt): Adjust.
383 2018-05-13 Mark Wielaard <mark@klomp.org>
385 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
386 (size_of_loc_descr): Likewise.
387 (output_loc_operands): Likewise.
388 (output_loc_operands_raw): Likewise.
389 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
390 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
391 (hash_loc_operands): Likewise.
392 (compare_loc_operands): Likewise.
394 2018-05-14 Mark Wielaard <mark@klomp.org>
396 * dwarf2out.c (count_index_addrs): New function.
397 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
399 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
401 PR tree-optimization/83648
402 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
403 return value as malloc candidate.
405 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
408 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
409 as true in call to suggest_attribute.
411 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
413 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
416 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
418 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
420 (aarch64_crypto_aesd_fused): Likewise.
422 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
424 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
425 (movsi_aarch64): Likewise.
426 (load_pairsi): Likewise.
427 (load_pairdi): Likewise.
428 (store_pairsi): Likewise.
429 (store_pairdi): Likewise.
430 (load_pairsf): Likewise.
431 (load_pairdf): Likewise.
432 (store_pairsf): Likewise.
433 (store_pairdf): Likewise.
434 (zero_extend): Likewise.
435 (trunc): Swap alternatives.
436 (fcvt_target): Add '?' to prefer w over r.
438 2018-05-14 Jakub Jelinek <jakub@redhat.com>
441 * config/i386/i386.md: Disallow non-commutative arithmetics in
442 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
443 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
444 in the peephole2 before it.
446 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
448 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
449 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
450 (ix86_handle_option): Handle -mcldemote.
451 * config.gcc: New header.
452 * config/i386/cldemoteintrin.h: New file.
453 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
454 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
456 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
457 OPTION_MASK_ISA_CLDEMOTE.
458 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
459 (ix86_valid_target_attribute_inner_p): Ditto.
460 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
461 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
462 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
463 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
464 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
466 * config/i386/i386.opt: Add -mcldemote.
467 * config/i386/x86intrin.h: New header.
468 * doc/invoke.texi: Add -mcldemote.
470 2018-05-14 Richard Biener <rguenther@suse.de>
472 * doc/match-and-simplify.texi: Adjust :s documentation.
474 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
476 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
477 intended memcpy size.
478 (REORDER_45): Likewise.
480 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
483 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
484 * vec.c (qsort_chk): Use gcc_qsort.
485 * Makefile.in (OBJS-libcommon): Add sort.o.
486 (build/sort.o): New target. Use it...
487 (BUILD_RTL): ... here, and...
488 (build/gencfn-macros): ... here, and...
489 (build/genmatch): ... here.
491 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
492 Chung-Ju Wu <jasonwucj@gmail.com>
494 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
495 * config/nds32/nds32-graywolf.md: New file.
496 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
497 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
499 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
500 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
501 * config/nds32/nds32.md (pipeline_model): Add graywolf.
502 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
503 * config/nds32/pipelines.md: Include n15 settings.
505 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
506 Chung-Ju Wu <jasonwucj@gmail.com>
508 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
509 * config/nds32/nds32-n13.md: New file.
510 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
511 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
513 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
514 * config/nds32/nds32.md (pipeline_model): Add n13.
515 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
516 * config/nds32/pipelines.md: Include n13 settings.
518 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
519 Chung-Ju Wu <jasonwucj@gmail.com>
521 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
522 * config/nds32/nds32-n10.md: New file.
523 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
524 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
526 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
527 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
528 * config/nds32/nds32.md (pipeline_model): Add n10.
529 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
530 * config/nds32/pipelines.md: Include n10 settings.
532 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
533 Kito Cheng <kito.cheng@gmail.com>
534 Chung-Ju Wu <jasonwucj@gmail.com>
536 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
537 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
538 Add enum values for DSP extension instructions.
539 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
541 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
542 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
544 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
545 * config/nds32/nds32-dspext.md: New file for DSP implementation.
546 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
547 * config/nds32/nds32-intrinsic.md: Likewise.
548 * config/nds32/nds32_intrinsic.h: Likewise.
549 * config/nds32/nds32-md-auxiliary.c: Likewise.
550 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
551 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
552 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
553 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
554 * config/nds32/nds32-protos.h: New declarations for DSP extension.
555 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
556 TYPE_DMAC in switch statement.
557 * config/nds32/nds32.c: New checking and implementation for DSP
558 extension instructions.
559 * config/nds32/nds32.h: Likewise.
560 * config/nds32/nds32.md: Likewise.
561 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
562 * config/nds32/predicates.md: Implement new predicates for DSP
565 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
567 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
568 Reformat alternatives and attributes so it is easier to identify
569 which constraints/attributes go with which instruction.
570 (mov<mode>_hardfloat32, FMOVE64): Likewise.
571 (mov<mode>_softfloat32, FMOVE64): Likewise.
572 (mov<mode>_hardfloat64, FMOVE64): Likewise.
573 (mov<mode>_softfloat64, FMOVE64): Likewise.
575 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
577 * doc/extend.texi (PowerPC Built-in Functions): Rename this
579 (Basic PowerPC Built-in Functions): The new name of the
580 subsection previously known as "PowerPC Built-in Functions".
581 (Basic PowerPC Built-in Functions Available on all Configurations):
583 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
584 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
585 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
586 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
588 2018-05-11 Martin Jambor <mjambor@suse.cz>
591 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
594 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
597 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
599 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
601 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
602 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
603 (ix86_handle_option): Handle -mwaitpkg.
604 * config.gcc: New header.
605 * config/i386/cpuid.h (bit_WAITPKG): New bit.
606 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
607 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
609 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
610 OPTION_MASK_ISA_WAITPKG.
611 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
612 (ix86_option_override_internal): Add PTA_WAITPKG.
613 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
614 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
616 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
617 __builtin_ia32_umwait and __builtin_ia32_tpause.
618 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
619 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
620 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
621 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
622 UNSPECV_TPAUSE): New.
623 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
624 * config/i386/i386.opt: Add -mwaitpkg.
625 * config/i386/waitpkgintrin.h: New file.
626 * config/i386/x86intrin.h: New header.
627 * doc/invoke.texi: Add -mwaitpkg.
629 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
632 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
634 (cortex-m0): Use armv6s-m isa.
635 (cortex-m0plus): Likewise.
636 (cortex-m1): Likewise.
637 (cortex-m0.small-multiply): Likewise.
638 (cortex-m0plus.small-multiply): Likewise.
639 (cortex-m1.small-multiply): Likewise.
641 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
642 Jakub Jelinek <jakub@redhat.com>
644 PR tree-optimization/85692
645 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
646 source permute as well.
648 2018-05-11 Martin Liska <mliska@suse.cz>
651 * doc/extend.texi: Document LLVM style format for no_sanitize
654 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
656 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
657 mode_supports_vsx_dform_quad to mode_supports_dq_form.
658 (mode_supports_vsx_dform_quad): Likewise.
659 (mode_supports_vmx_dform): Move these functions to be next to the
660 other mode_supports functions.
661 (mode_supports_dq_form): Likewise.
662 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
663 mode_supports_dq_form.
664 (reg_offset_addressing_ok_p): Likewise.
665 (offsettable_ok_by_alignment): Likewise.
666 (rs6000_legitimate_offset_address_p): Likewise.
667 (legitimate_lo_sum_address_p): Likewise.
668 (rs6000_legitimize_address): Likewise.
669 (rs6000_legitimize_reload_address): Likewise.
670 (rs6000_secondary_reload_inner): Likewise.
671 (rs6000_preferred_reload_class): Likewise.
672 (rs6000_output_move_128bit): Likewise.
674 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
676 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
677 Generate SImode target register for null target.
678 <case IX86_BUILTIN_XGETBV>: Ditto.
679 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
680 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
682 2018-05-10 Carl Love <cel@us.ibm.com>
684 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
685 dcbtt and dcbtstt if operands[2] is 0.
687 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
690 * config/i386/sse.md (usadv64qi): New expander.
692 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
694 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
695 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
696 -maltivec=be support.
697 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
698 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
699 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
700 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
701 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
702 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
703 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
704 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
705 altivec_vsumsws): Adjust.
706 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
707 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
708 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
710 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
711 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
712 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
713 (altivec_lve<VI_char>x): Delete expand.
714 (*altivec_lve<VI_char>x_internal): Rename to...
715 (altivec_lve<VI_char>x): ... this.
716 (altivec_lvxl_<mode>): Delete expand.
717 (*altivec_lvxl_<mode>_internal): Rename to ...
718 (altivec_lvxl_<mode>): ... this.
719 (altivec_stvxl_<mode>): Delete expand.
720 (*altivec_stvxl_<mode>_internal): Rename to ...
721 (altivec_stvxl_<mode>): ... this.
722 (altivec_stve<VI_char>x): Delete expand.
723 (*altivec_stve<VI_char>x_internal): Rename to ...
724 (altivec_stve<VI_char>x): ... this.
725 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
726 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
727 reduc_plus_scal_<mode>): Adjust.
728 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
730 (rs6000_cpu_cpp_builtins): Adjust.
731 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
732 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
733 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
734 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
735 -maltivec=be support.
736 (rs6000_split_vec_extract_var): Adjust.
737 (rs6000_split_v4si_init): Adjust.
738 (swap_selector_for_mode): Delete.
739 (altivec_expand_lvx_be, altivec_expand_stvx_be,
740 altivec_expand_stvex_be): Delete.
741 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
742 -maltivec=be support.
743 (rs6000_gimple_fold_builtin): Ditto.
744 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
746 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
747 (TARGET_DIRECT_MOVE_64BIT): Adjust.
748 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
749 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
750 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
751 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
752 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
753 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
754 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
755 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
756 anonymous split): Adjust.
757 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
758 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
760 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
762 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
763 when --with-gxx-include-dir is also specified.
764 * configure: Regenerate.
766 2018-05-09 Jim Wilson <jimw@sifive.com>
769 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
770 * config/riscv/t-withmultilib: New.
771 * config/riscv/withmultilib.h: New.
772 * doc/install.texi: Document RISC-V --with-multilib-list support.
774 2018-05-09 Richard Biener <rguenther@suse.de>
776 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
778 (vect_bb_vectorization_profitable_p): Adjust. Compute
779 actual scalar cost using the cost vector and the add_stmt_cost
782 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
784 PR rtl-optimization/85645
785 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
786 in the REG_CFA_REGISTER note for LR, don't leave it empty.
788 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
790 PR rtl-optimization/85645
791 * shrink-wrap.c (spread_components): Return a boolean saying if
792 anything was changed.
793 (try_shrink_wrapping_separate): Iterate spread_components until
794 nothing changes anymore.
796 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
798 PR rtl-optimization/85645
799 * regrename.c (build_def_use): Also kill the chains that include the
800 destination of a REG_CFA_REGISTER note.
802 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
804 PR rtl-optimization/85645
805 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
806 insn that has a REG_CFA_REGISTER note.
808 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
810 * cfgexpand.c (expand_clobber): New function.
811 (expand_gimple_stmt_1): Use it.
812 * tree-vect-stmts.c (vect_clobber_variable): New function,
814 (vectorizable_simd_clone_call): ...here.
815 (vectorizable_store): Emit a clobber either side of an
816 IFN_STORE_LANES sequence.
817 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
819 2018-05-09 Tom de Vries <tom@codesourcery.com>
822 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
823 (define_insn "trap_if_false"): Add exit after trap.
825 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
827 PR rtl-optimization/85638
828 * bb-reorder.c: Include common/common-target.h.
829 (create_forwarder_block): New function extracted from...
830 (fix_up_crossing_landing_pad): ...here. Rename into...
831 (dw2_fix_up_crossing_landing_pad): ...this.
832 (sjlj_fix_up_crossing_landing_pad): New function.
833 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
834 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
835 from both partitions and exit the loop after one iteration.
837 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
840 * doc/extend.texi (PowerPC Built-in Functions): Rename this
842 (Basic PowerPC Built-in Functions): The new name of the
843 subsection previously known as "PowerPC Built-in Functions".
844 (Basic PowerPC Built-in Functions Available on all Configurations):
846 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
848 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
850 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
852 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
855 2018-05-08 Jim Wilson <jimw@sifive.com>
857 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
858 (LD_EMUL_SUFFIX): New.
861 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
863 * doc/extend.texi (PowerPC Built-in Functions): Rename this
865 (Basic PowerPC Built-in Functions): The new name of the
866 subsection previously known as "PowerPC Built-in Functions".
867 (Basic PowerPC Built-in Functions Available on all Configurations):
869 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
871 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
873 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
875 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
878 2018-05-08 Jakub Jelinek <jakub@redhat.com>
881 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
882 after cmpelim optimization.
884 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
886 * config.gcc: Support "goldmont".
887 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
888 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
890 * config/i386/i386.c (m_GOLDMONT): Define.
891 (processor_target_table): Add "goldmont".
892 (PTA_GOLDMONT): Define.
893 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
894 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
895 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
896 (fold_builtin_cpu): Add "goldmont".
897 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
898 (ix86_option_override_internal): Add "goldmont".
899 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
900 (processor_type): Add PROCESSOR_GOLDMONT.
901 * config/i386/i386.md: Add CPU "glm".
902 * config/i386/glm.md: New file.
903 * config/i386/x86-tune.def: Add m_GOLDMONT.
904 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
906 2018-05-08 Jakub Jelinek <jakub@redhat.com>
909 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
911 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
912 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
913 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
916 * config/i386/i386.c (ix86_fold_builtin): Handle
917 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
920 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
921 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
923 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
926 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
927 (check_arch): Likewise.
928 (check_fpu): Return the result rather than printing it.
929 (end arch): Fix operator precedence.
931 (END): Print the result from check_fpu.
933 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
934 Alan Hayward <alan.hayward@arm.com>
935 David Sherwood <david.sherwood@arm.com>
937 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
938 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
939 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
940 (*fcmuo<mode>_and): New patterns.
942 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
944 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
945 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
946 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
947 (cmp_op, sve_imm_con): New code attributes.
948 (SVE_COND_INT_CMP, imm_con): Delete.
949 (cmp_op): Remove above unspecs from int attribute.
950 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
952 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
953 comparison-specific unspecs.
954 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
955 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
956 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
957 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
958 (*vec_fcm<cmp_op><mode>): Rename to...
959 (*fcm<cmp_op><mode>): ...this and adjust likewise.
960 (*vec_fcmuo<mode>): Rename to...
961 (*fcmuo<mode>): ...this and adjust likewise.
962 (*pred_fcm<cmp_op><mode>): New pattern.
963 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
964 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
966 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
968 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
969 (aarch64_emit_sve_predicated_cond): New function.
970 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
971 (aarch64_emit_unspec_cond_or): Replace with...
972 (aarch64_emit_sve_or_conds): ...this new function. Use
973 aarch64_emit_sve_ptrue_op for the individual comparisons and
974 aarch64_emit_binop to OR them together.
975 (aarch64_emit_inverted_unspec_cond): Replace with...
976 (aarch64_emit_sve_inverted_cond): ...this new function. Use
977 aarch64_emit_sve_ptrue_op for the comparison and
978 aarch64_emit_unop to invert the result.
979 (aarch64_expand_sve_vec_cmp_float): Update after the above
980 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
982 2018-05-07 Nathan Sidwell <nathan@acm.org>
984 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
985 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
986 (Backwards Compatibility): Likewise.
988 2018-05-07 Luis Machado <luis.machado@linaro.org>
992 2018-05-07 Luis Machado <luis.machado@linaro.org>
994 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
995 <prefetch_dynamic_strides>: New const bool field.
996 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
997 prefetch_dynamic_strides.
998 (exynosm1_prefetch_tune): Likewise.
999 (thunderxt88_prefetch_tune): Likewise.
1000 (thunderx_prefetch_tune): Likewise.
1001 (thunderx2t99_prefetch_tune): Likewise.
1002 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1003 (aarch64_override_options_internal): Update to set
1004 PARAM_PREFETCH_DYNAMIC_STRIDES.
1005 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1006 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1007 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1008 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1009 prefetch-dynamic-strides setting.
1011 2018-05-07 Luis Machado <luis.machado@linaro.org>
1013 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1014 <minimum_stride>: New const int field.
1015 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1016 minimum_stride field.
1017 (exynosm1_prefetch_tune): Likewise.
1018 (thunderxt88_prefetch_tune): Likewise.
1019 (thunderx_prefetch_tune): Likewise.
1020 (thunderx2t99_prefetch_tune): Likewise.
1021 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1022 (aarch64_override_options_internal): Update to set
1023 PARAM_PREFETCH_MINIMUM_STRIDE.
1024 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1025 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1026 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1027 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1028 stride is constant and is below the minimum stride threshold.
1030 2018-05-07 Luis Machado <luis.machado@linaro.org>
1032 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1035 2018-05-07 Luis Machado <luis.machado@linaro.org>
1037 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1038 <prefetch_dynamic_strides>: New const bool field.
1039 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1040 prefetch_dynamic_strides.
1041 (exynosm1_prefetch_tune): Likewise.
1042 (thunderxt88_prefetch_tune): Likewise.
1043 (thunderx_prefetch_tune): Likewise.
1044 (thunderx2t99_prefetch_tune): Likewise.
1045 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1046 (aarch64_override_options_internal): Update to set
1047 PARAM_PREFETCH_DYNAMIC_STRIDES.
1048 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1049 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1050 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1051 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1052 prefetch-dynamic-strides setting.
1054 2018-05-07 Luis Machado <luis.machado@linaro.org>
1056 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1057 <minimum_stride>: New const int field.
1058 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1059 minimum_stride field.
1060 (exynosm1_prefetch_tune): Likewise.
1061 (thunderxt88_prefetch_tune): Likewise.
1062 (thunderx_prefetch_tune): Likewise.
1063 (thunderx2t99_prefetch_tune): Likewise.
1064 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1065 (aarch64_override_options_internal): Update to set
1066 PARAM_PREFETCH_MINIMUM_STRIDE.
1067 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1068 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1069 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1070 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1071 stride is constant and is below the minimum stride threshold.
1073 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1076 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1077 the type is addressable. Don't force op into register if it has
1080 2018-05-05 Roland McGrath <mcgrathr@google.com>
1083 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1084 any section for which we don't know a specific type it should have,
1085 regardless of name. Previously this was done only for the exact
1086 names ".init_array", ".fini_array", and ".preinit_array".
1087 (default_elf_asm_named_section): Add comment about
1088 relationship with default_section_type_flags and SECTION_NOTYPE.
1089 (get_section): Don't consider it a type conflict if one side has
1090 SECTION_NOTYPE and the other doesn't, as long as neither has the
1091 SECTION_BSS et al used in the default_section_type_flags logic.
1093 2018-05-05 Tom de Vries <tom@codesourcery.com>
1096 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1097 (workaround_barsyncs): New function.
1098 (nvptx_reorg): Use workaround_barsyncs.
1099 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1100 (define_expand "nvptx_membar_cta"): New define_expand.
1101 (define_insn "*nvptx_membar_cta"): New insn.
1103 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1105 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1106 To improve optimization opportunities.
1107 * builtin-types.def: The new needed builtin types for the above.
1109 2018-05-04 Richard Biener <rguenther@suse.de>
1111 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1112 * gimple-ssa-store-merging.c
1113 (imm_store_chain_info::output_merged_store): Remove redundant create,
1114 release split_store vector contents on failure.
1115 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1116 scalar stmt vector on cache hit.
1118 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1120 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1122 * config.gcc (powerpc-xilinx-eabi*): Remove.
1123 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1125 (fusion_addis_mem_combo_load): Ditto.
1126 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1128 (rs6000_cpu_cpp_builtins): Ditto.
1129 * config/rs6000/rs6000-linux.c
1130 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1131 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1132 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1134 (rs6000_setup_reg_addr_masks): Ditto.
1135 (rs6000_init_hard_regno_mode_ok): Ditto.
1136 (rs6000_option_override_internal): Ditto.
1137 (legitimate_lo_sum_address_p): Ditto.
1138 (rs6000_legitimize_address): Ditto.
1139 (rs6000_legitimize_reload_address): Ditto.
1140 (rs6000_legitimate_address_p): Ditto.
1141 (abi_v4_pass_in_fpr): Ditto.
1142 (setup_incoming_varargs): Ditto.
1143 (rs6000_gimplify_va_arg): Ditto.
1144 (rs6000_split_multireg_move): Ditto.
1145 (rs6000_savres_strategy): Ditto.
1146 (rs6000_emit_prologue_components): Ditto.
1147 (rs6000_emit_epilogue_components): Ditto.
1148 (rs6000_emit_prologue): Ditto.
1149 (rs6000_emit_epilogue): Ditto.
1150 (rs6000_elf_file_end): Ditto.
1151 (rs6000_function_value): Ditto.
1152 (rs6000_libcall_value): Ditto.
1153 * config/rs6000/rs6000.h: Ditto.
1154 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1155 (TARGET_MINMAX): ... this. New.
1156 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1157 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1158 (*movsi_internal1_single): Delete.
1159 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1160 mfpu=, mxilinx-fpu): Delete.
1161 * config/rs6000/singlefp.h: Delete.
1162 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1163 * config/rs6000/t-rs6000: Ditto.
1164 * config/rs6000/t-xilinx: Delete.
1165 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1166 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1167 (VStype_simple): Delete.
1168 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1169 * config/rs6000/xfpu.h: Delete.
1170 * config/rs6000/xfpu.md: Delete.
1171 * config/rs6000/xilinx.h: Delete.
1172 * config/rs6000/xilinx.opt: Delete.
1173 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1174 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1176 2018-05-04 Tom de Vries <tom@codesourcery.com>
1179 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1182 2018-05-04 Richard Biener <rguenther@suse.de>
1185 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1186 (expand_complex_div_wide): Likewise.
1187 (expand_complex_operations_1): Likewise.
1188 (expand_complex_libcall): Preserve EH info of the original stmt.
1189 (tree_lower_complex): Handle removed blocks.
1190 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1191 on complex multiplication and division libcall builtins.
1193 2018-05-04 Richard Biener <rguenther@suse.de>
1196 * fold-const.c (negate_expr_p): Restrict negation of operand
1197 zero of a division to when we know that can happen without
1199 (fold_negate_expr_1): Likewise.
1201 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1204 * real.h (real_nextafter): Declare.
1205 * real.c (real_nextafter): New function.
1206 * fold-const-call.c (fold_const_nextafter): New function.
1207 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1208 CASE_CFN_NEXTTOWARD.
1209 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1210 even when arg1_mode is different from arg0_mode.
1212 2018-05-03 Nathan Sidwell <nathan@acm.org>
1214 * doc/extend.texi (Deprecated Features): Remove
1216 (Backwards Compatibility): Likewise.
1217 * doc/invoke.texi (C++ Language Options): Likewise.
1218 (C++ Dialect Options): Likewise.
1220 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1223 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1224 _mm512_mask_mullox_epi64): New intrinsics.
1226 2018-05-03 Tom de Vries <tom@codesourcery.com>
1229 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1230 dump files): Add offload-tree.
1232 2018-05-03 Richard Biener <rguenther@suse.de>
1234 PR tree-optimization/85615
1235 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1236 to loops not nested in BBs loop father to avoid creating multi-entry
1239 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1241 PR tree-optimization/70291
1242 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1243 arguments. Change return type to tree. Emit libcall as a new
1244 statement rather than replacing existing one when inplace_p is true.
1245 (expand_complex_multiplication_components): New function.
1246 (expand_complex_multiplication): Expand floating-point complex
1247 multiplication using the above.
1248 (expand_complex_division): Rename inner_type parameter to type.
1249 Update expand_complex_libcall call-site.
1250 (expand_complex_operations_1): Update expand_complex_multiplication
1251 and expand_complex_division call-sites.
1253 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1256 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1257 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1258 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1259 the highest significant bit of the shift count mask is clear. In
1260 check whether and[sq]i3 is needed verify that all significant bits
1261 of the shift count other than the highest are set.
1263 2018-05-02 Tom de Vries <tom@codesourcery.com>
1266 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1267 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1268 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1269 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1270 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1271 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1272 __builtin_goacc_parlevel_size.
1274 2018-05-02 Richard Biener <rguenther@suse.de>
1276 PR tree-optimization/85597
1277 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1278 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1281 2018-05-02 Tom de Vries <tom@codesourcery.com>
1284 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1285 dump files): Add ltrans-tree.
1287 2018-05-02 Tom de Vries <tom@codesourcery.com>
1290 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1291 dump files): Add wpa-ipa.
1293 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1295 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1296 powerpc*-*-linux*paired* target.
1297 * config/rs6000/750cl.h: Delete.
1298 * config/rs6000/paired.h: Delete.
1299 * config/rs6000/paired.md: Delete.
1300 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1302 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1303 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1304 comment. Remove paired float support.
1305 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1306 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1308 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1309 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1311 * config/rs6000/rs6000.c: Remove paired float support.
1312 (paired_expand_vector_init, paired_expand_vector_move,
1313 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1314 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1315 paired_expand_builtin, paired_expand_predicate_builtin,
1316 paired_init_builtins): Delete.
1317 * config/rs6000/rs6000.h: Remove paired float support.
1318 * config/rs6000/rs6000.md: Remove paired float support.
1319 (move_from_CR_ov_bit): Delete.
1320 * config/rs6000/rs6000.opt (mpaired): Delete.
1321 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1322 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1324 2018-05-02 Richard Biener <rguenther@suse.de>
1327 * gimplify.c (gimplify_save_expr): When in SSA form allow
1328 SAVE_EXPRs to compute to SSA vars.
1330 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1333 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1334 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1335 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1336 clobber operands[2], instead use a new pseudo. Formatting fixes.
1338 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1340 PR tree-optimization/85586
1341 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1342 exit early for statements in the same group if the accesses are
1345 2018-05-02 Tom de Vries <tom@codesourcery.com>
1348 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1351 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1353 PR tree-optimization/85143
1354 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1356 2018-05-01 Tom de Vries <tom@codesourcery.com>
1359 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1360 not found" error message.
1362 2018-05-01 Tom de Vries <tom@codesourcery.com>
1365 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1366 * vec.c (test_ordered_remove_if): New function.
1367 (vec_c_tests): Call test_ordered_remove_if.
1368 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1369 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1370 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1371 VEC_ORDERED_REMOVE_IF.
1373 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1375 PR tree-optimization/82665
1376 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1377 pointer subtraction where arguments come from a memchr call.
1379 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1381 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1382 --push-state --as-needed and --pop-state instead of --as-needed and
1383 --no-as-needed if ld supports it.
1384 * configure: Regenerated.
1387 * doc/install.texi2html: Replace _002d with - and _002a with * in
1388 generated html files using sed.
1390 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1393 * gcc-rich-location.c (blank_line_before_p): New function.
1394 (use_new_line): New function.
1395 (gcc_rich_location::add_fixit_insert_formatted): New function.
1396 * gcc-rich-location.h
1397 (gcc_rich_location::add_fixit_insert_formatted): New function.
1399 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1401 * selftest.c (assert_streq): Rename "expected" and "actual" to
1402 "val1" and "val2". Extend NULL-handling to cover both inputs
1403 symmetrically, while still requiring both to be non-NULL for a pass.
1404 * selftest.h (assert_streq): Rename "expected" and "actual" to
1406 (ASSERT_EQ): Likewise.
1407 (ASSERT_EQ_AT): Likewise.
1408 (ASSERT_KNOWN_EQ): Likewise.
1409 (ASSERT_KNOWN_EQ_AT): Likewise.
1410 (ASSERT_NE): Likewise.
1411 (ASSERT_MAYBE_NE): Likewise.
1412 (ASSERT_MAYBE_NE_AT): Likewise.
1413 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1414 the assertion to pass.
1415 (ASSERT_STREQ_AT): Likewise.
1417 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1419 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1420 interaction with -pie.
1422 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1424 * selftest.h: Fix alphabetization of per-source-file selftest
1427 2018-04-30 Jason Merrill <jason@redhat.com>
1429 PR c++/61982 - dead stores to destroyed objects.
1430 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1433 2018-04-30 Jason Merrill <jason@redhat.com>
1435 * tree.c (build_clobber): New.
1436 * tree.h: Declare it.
1437 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1439 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1441 * diagnostic-show-locus.c (layout::layout): Update for
1442 location_get_source_line returning a char_span.
1443 (struct char_span): Move to input.h.
1444 (struct correction): Update for fields in char_span becoming
1446 (struct source_line): Update for location_get_source_line
1447 returning a char_span.
1448 (layout::print_line): Likewise.
1449 * edit-context.c (edited_file::print_content): Likewise.
1450 (edited_file::print_diff_hunk): Likewise.
1451 (edited_file::print_run_of_changed_lines): Likewise.
1452 (edited_file::get_num_lines): Likewise.
1453 (edited_line::edited_line): Likewise.
1454 * final.c (asm_show_source): Likewise.
1455 * input.c (location_get_source_line): Convert return type
1456 from const char * to char_span, losing the final "line_len"
1458 (dump_location_info): Update for the above.
1459 (get_substring_ranges_for_loc): Likewise. Use a char_span
1460 when handling the literal within the line.
1461 (test_reading_source_line): Update for location_get_source_line
1462 returning a char_span.
1463 * input.h (class char_span): Move here from
1464 diagnostic-show-locus.c, converting from a struct to a class.
1465 Make data members private.
1466 (char_span::operator bool): New.
1467 (char_span::length): New.
1468 (char_span::get_buffer): New.
1469 (char_span::operator[]): New.
1470 (char_span::subspan): Make const.
1471 (char_span::xstrdup): New.
1472 (location_get_source_line): Convert return type from const char *
1473 to char_span, losing the final "line_size" param.
1475 2018-04-30 Jan Hubicka <jh@suse.cz>
1477 * lto-wrapper.c (ltrans_priorities): New static var.
1478 (cmp_priority): New.
1479 (run_gcc): Read priorities and if doing parallel build order
1480 the Makefile by them.
1482 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1484 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1486 2018-04-30 Richard Biener <rguenther@suse.de>
1488 * tree-cfg.c (verify_address): Remove base argument, add
1489 flag whether to check TREE_ADDRESSABLE and do that.
1490 (verify_expr): Remove.
1491 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1492 (verify_gimple_assign_single): Likewise.
1493 (verify_gimple_switch): Likewise.
1494 (verify_expr_location_1): Dereference tp once. Add (disabled)
1495 piece from verify_expr.
1496 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1498 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1500 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1502 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1504 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1505 (small_data_pattern): Likewise.
1506 (arc_rewrite_small_data): Likewise.
1507 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1508 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1509 (get_symbol_alignment): New function.
1510 (legitimate_small_data_address_p): Likewise.
1511 (legitimate_scaled_address): Update, call
1512 legitimate_small_data_address_p.
1513 (output_sdata): New static variable.
1514 (arc_print_operand): Update how we handle small data operands.
1515 (arc_print_operand_address): Likewise.
1516 (arc_legitimate_address_p): Update, use
1517 legitimate_small_data_address_p.
1518 (arc_rewrite_small_data_p): Remove.
1519 (arc_rewrite_small_data_1): Likewise.
1520 (arc_rewrite_small_data): Likewise.
1521 (small_data_pattern): Likewise.
1522 (compact_sda_memory_operand): Update to use
1523 legitimate_small_data_address_p and get_symbol_alignment.
1524 (prepare_move_operands): Don't rewite sdata pattern.
1525 (prepare_extend_operands): Remove.
1526 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1528 (zero_extendqisi2): Likewise.
1529 (zero_extendhisi2): Likewise.
1530 (extendqihi2): Likewise.
1531 (extendqisi2): Likewise.
1532 (extendhisi2): Likewise.
1536 * config/arc/constraints.md (Usd): Change it to memory constraint.
1538 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1540 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1541 as source of std instructions.
1542 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1543 allow 6-bit constants as source for store instructions.
1544 (movdi_insn): Update instruction pattern to allow 6-bit constants
1545 as source for store instructions.
1547 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1549 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1551 2018-04-30 Nathan Sidwell <nathan@acm.org>
1552 Sandra Loosemore <sandra@codesourcery.com>
1554 * dumpfile.c (dump_open): Allow '-' for stdout.
1555 * doc/invoke.texi (Developer Options): Document dump filename
1556 determination early. Document stdin/stdout selection.
1558 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1560 Microblaze Target: PIC data text relative
1562 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1563 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1565 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1566 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1567 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1568 New addressing mode for data-text relative position indepenedent code.
1569 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1570 'ADDRESS_SYMBOLIC_TXT_REL'.
1571 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1572 (microblaze_legitimate_pic_operand): Exclude function calls from
1573 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1574 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1576 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1577 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1578 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1579 for 'address + offset'.
1580 (microblaze_expand_prologue): Add new function prologue call for
1582 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1583 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1584 table in case of TARGET_PIC_DATA_TEXT_REL.
1585 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1586 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1587 Add new macros 'UNSPEC_TEXT',
1588 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1589 + exclude function calls from 'UNSPEC_PLT' in case of data text
1591 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1592 new target hook for generating address diff vector tables in case of
1594 * doc/tm.texi : Regenerate.
1595 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1596 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1597 of addr diff vector generation.
1598 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1599 target hook definition.
1600 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1601 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1602 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1603 option for data text relative.
1605 2018-04-30 Richard Biener <rguenther@suse.de>
1607 * tree-chrec.h (evolution_function_is_constant_p): Remove
1609 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1611 2018-04-30 Richard Biener <rguenther@suse.de>
1614 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1616 2018-04-30 Richard Biener <rguenther@suse.de>
1618 PR tree-optimization/28364
1619 PR tree-optimization/85275
1620 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1621 copying first exit test.
1623 2018-04-28 Mark Wielaard <mark@klomp.org>
1625 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1627 (dwarf_AT): Handle DW_AT_addr_base.
1628 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1630 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1633 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1634 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1635 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1636 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1638 2018-04-28 Richard Biener <rguenther@suse.de>
1640 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1641 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1642 to reflect use. Only add interesting stmts.
1644 2018-04-27 Martin Jambor <mjambor@suse.cz>
1647 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1648 the jump function allows for passing through aggregate values.
1650 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1652 * input.h (in_system_header_at): Convert from macro to inline
1654 (from_macro_expansion_at): Likewise.
1655 (from_macro_definition_at): Likewise.
1657 2018-04-27 Jeff Law <law@redhat.com>
1659 * config.gcc: Mark tile* targets as deprecated/obsolete.
1661 2018-04-27 Richard Biener <rguenther@suse.de>
1663 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1666 2018-04-27 Richard Biener <rguenther@suse.de>
1668 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1670 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1672 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1673 with Yd constraint. Set "preferred_for_speed" attribute from
1674 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1676 (*movdi_internal): Ditto.
1677 (movti_interunit splitters): Remove
1678 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1679 (movdi_interunit splitters): Ditto.
1680 * config/i386/constraints.md (Ye): Remove.
1681 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1683 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1686 * config/aarch64/constraints.md (Usg): Limit to 31.
1689 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1691 PR tree-optimization/85529
1692 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1693 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1694 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1695 zero extension or masking of the MSB bit.
1696 (optimize_range_tests): Add FIRST_BB argument, pass it through
1697 to optimize_range_tests_var_bound.
1698 (maybe_optimize_range_tests, reassociate_bb): Adjust
1699 optimize_range_tests callers.
1701 2018-04-26 Richard Biener <rguenther@suse.de>
1702 Jakub Jelinek <jakub@redhat.com>
1704 * cgraph.h (symbol_table): Just declare debug method here.
1705 * symtab.c (symbol_table::debug): Define.
1707 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1709 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1711 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1713 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1714 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1715 (*movdi_internal): Substitute Yi and Yj constraint with x
1716 and Ym and Yn constraint with y constraint. Update "isa"
1717 attribute and set "preferred_for_speed" attribute from
1718 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1719 (*movsi_internal): Ditto.
1720 (*movdf_internal): Ditto.
1721 (*movsf_internal): Ditto.
1722 (*zero_extendsidi2): Ditto.
1723 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1724 (sse2_loadld): Ditto.
1725 (*vec_extract<ssevecmodelower>_0): Ditto.
1726 (*vec_extractv4si_0_zext_sse4): Ditto.
1727 (vec_concatv2di): Ditto.
1728 (*vec_dup<mode>): Ditto.
1729 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1730 * config/i386/constraints.md (Yi): Remove.
1735 2018-04-26 Nathan Sidwell <nathan@acm.org>
1737 * dumpfile.c (dump_open): New.
1738 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1739 (dump_finish): Detect stdio/stderr by value not name.
1741 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1743 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1745 2018-04-26 Tom de Vries <tom@codesourcery.com>
1748 * config/nvptx/nvptx.c (verify_neutering_jumps)
1749 (verify_neutering_labels): New function
1750 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1752 2018-04-26 Tom de Vries <tom@codesourcery.com>
1755 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1756 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1759 2018-04-26 Richard Biener <rguenther@suse.de>
1760 Tom de Vries <tom@codesourcery.com>
1763 * lto-streamer-out.c (output_function): Fixup loops if required to match
1764 discovery done in the reader.
1766 2018-04-26 Richard Biener <rguenther@suse.de>
1768 PR tree-optimization/85116
1769 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1770 have a loop exit from the single latch predecessor. Remove
1771 case of header with just condition.
1772 (ch_base::copy_headers): Exclude infinite loops from any
1774 (pass_ch::execute): Record exits.
1776 2018-04-26 Richard Biener <rguenther@suse.de>
1778 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1779 prologue cost vector and pass it to vect_get_load_cost.
1780 (vect_get_peeling_costs_all_drs): Likewise.
1781 (vect_peeling_hash_get_lowest_cost): Likewise.
1782 (vect_enhance_data_refs_alignment): Likewise.
1784 2018-04-26 Richard Biener <rguenther@suse.de>
1787 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1788 checking of integer<->pointer conversions.
1789 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1790 sign-/zero-extending pointer types.
1791 (expand_omp_for_static_chunk): Likewise.
1793 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1794 Jean Lee <xiaoyur347@gmail.com>
1796 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1797 (TARGET_ASAN_SHADOW_OFFSET): Define.
1798 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1799 true for -fsanitize=address.
1801 2018-04-25 Mark Wielaard <mark@klomp.org>
1803 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1806 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1808 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1809 than "alu", remove explicit "memory" and "imm_disp" attributes.
1810 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1813 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1814 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1815 gen_lowpart_no_emit.
1817 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1820 * config/i386/i386.c (ix86_expand_builtin): Change memory
1821 operand to XI, extend p0 to Pmode.
1822 * config/i386/i386.md: Change unspec volatile and operand
1823 1 mode to XI, change operand 0 mode to P.
1825 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1827 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1828 GET_MODE_MASK before any checking.
1829 (nds32_can_use_bset_p): Likewise.
1830 (nds32_can_use_btgl_p): Likewise.
1832 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1834 * config/nds32/nds32-doubleword.md: New define_split pattern for
1835 illegal register number.
1837 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1839 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1841 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1843 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1845 2018-04-25 Richard Biener <rguenther@suse.de>
1847 * lto-streamer.h (LTO_major_version): Bump to 8.
1849 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1851 * BASE-VER: Set to 9.0.0.
1853 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1855 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1856 in __abskf2 and __powikf2.
1858 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1861 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1862 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1863 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1864 Use the above on operand 2. Reindent.
1865 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1867 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1870 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1872 * config/i386/i386.opt (mcet): Removed.
1873 * doc/install.texi: Remove -mcet documentation.
1874 * doc/invoke.texi: Likewise.
1876 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1879 * doc/install.texi: Remove -mcet from bootstrap-cet.
1881 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1884 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1885 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1889 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1890 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1891 containing a CONST_VECTOR.
1893 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1895 * doc/install.texi: Update newlib dependency for nvptx.
1897 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1900 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1901 instead of INTVAL when shifting x left.
1903 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1905 PR tree-optimization/85478
1906 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1907 vect_grouped_store_supported for single element vectors.
1909 2018-04-24 Richard Biener <rguenther@suse.de>
1912 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1913 load cost increase to the case of non-constant step.
1915 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1918 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1919 destination if any_malformed_asm.
1921 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1924 * expr.c (store_field): In the bitfield case, if the value comes from
1925 a function call and is returned in registers by means of a PARALLEL,
1926 do not change the mode of the temporary unless BLKmode and VOIDmode.
1928 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1930 PR rtl-optimization/85423
1931 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1932 dependencies to debug insns when the previous insn is non-debug.
1934 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1936 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1937 enums into a single definition.
1938 (fls): Fix predicates and printing.
1941 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1943 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1944 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1945 and short u6 immediate.
1946 (check_if_valid_sleep_operand): Remove.
1947 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1949 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1951 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1952 flag_always_save_lp condition.
1953 * config/nds32/nds32.opt (malways-save-lp): New option.
1955 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1957 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1958 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1959 * config/nds32/nds32.h
1960 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1961 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1963 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1965 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1966 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1968 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1969 Chung-Ju Wu <jasonwucj@gmail.com>
1971 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1973 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1974 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1976 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1978 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1980 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1982 * config/nds32/nds32-protos.h (nds32_data_alignment,
1983 nds32_local_alignment): Declare.
1984 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1985 nds32_local_alignment): New functions.
1986 (TARGET_CONSTANT_ALIGNMENT): Define.
1987 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1989 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1991 * config/nds32/nds32.c
1992 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1993 (TARGET_MODES_TIEABLE_P): Likewise.
1995 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1997 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2000 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2001 Chung-Ju Wu <jasonwucj@gmail.com>
2003 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2004 for unaligned access.
2005 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2007 * config/nds32/nds32-intrinsic.md: Likewise.
2008 * config/nds32/nds32_intrinsic.h: Likewise.
2009 * config/nds32/nds32.h (nds32_builtins): Likewise.
2010 * config/nds32/nds32.opt (munaligned-access): New option.
2011 * config/nds32/nds32.c (nds32_asm_file_start): Display
2012 flag_unaligned_access status.
2014 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2016 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2017 -mno-relax is present.
2018 * config/riscv/linux.h (LINK_SPEC): Ditto.
2020 2018-04-20 Martin Sebor <msebor@redhat.com>
2023 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2025 (gimple_fold_builtin_stxcpy_chk): Same.
2026 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2028 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2031 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2032 __powikf2 when long double is IEEE 128-bit.
2034 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2036 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2037 step to make sure stack always aligned.
2039 2018-04-20 Carl Love <cel@us.ibm.com>
2042 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2043 size check for arg0.
2045 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2046 Tom de Vries <tom@codesourcery.com>
2049 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2050 Emit insns for calls too.
2051 (nvptx_find_par): Always look for worker-level predecessor insn.
2052 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2054 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2055 (nvptx_process_pars): Propagate frames for calls.
2057 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2060 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2062 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2063 (ix86_handle_option): Don't handle OPT_mibt.
2064 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2066 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2068 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2069 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2070 (ix86_target_macros): Define __CET__ with flag_cf_protection
2071 for -fcf-protection.
2072 * config/i386/i386.c (isa2_opts): Remove -mibt.
2073 * config/i386/i386.h (TARGET_IBT): Removed.
2074 (TARGET_IBT_P): Likewise.
2075 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2076 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2077 * config/i386/i386.opt (mcet): Update help message.
2080 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2081 -mcet as an alias for -mshstk.
2083 2018-04-20 Richard Biener <rguenther@suse.de>
2086 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2087 complexity by forcing a single use of the multiply operand.
2089 2018-04-20 Martin Jambor <mjambor@suse.cz>
2092 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2093 recursion dependency to only apply to non-clones.
2095 2018-04-20 Martin Jambor <mjambor@suse.cz>
2098 * ipa-cp.c (create_specialized_node): Check that clones of
2099 self-recursive edges exist during IPA-CP.
2101 2018-04-19 Toon Moene <toon@moene.org>
2103 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2106 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2108 PR tree-optimization/85467
2109 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2110 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2111 VECTOR_CST element to type.
2113 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2116 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2117 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2118 (builtin_longjmp): Likewise.
2119 (save_stack_nonlocal): New pattern.
2120 (restore_stack_nonlocal): Likewise.
2122 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2125 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2126 Replace ASM_OUTPUT_LABEL with fprintf.
2128 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2131 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2132 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2133 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2134 define __IBT__ and __SHSTK__ for -fcf-protection.
2135 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2137 (ix86_trampoline_init): Likewise.
2138 (x86_output_mi_thunk): Likewise.
2139 (ix86_notrack_prefixed_insn_p): Likewise.
2140 (ix86_option_override_internal): Don't disallow -fcf-protection.
2141 * config/i386/i386.md (rdssp<mode>): Also enable for
2143 (incssp<mode>): Likewise.
2144 (nop_endbr): Likewise.
2145 * config/i386/i386.opt (mcet): Change help message to built-in
2149 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2150 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2151 enable CET built-in functions.
2153 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2155 * common/config/i386/i386-common.c
2156 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2157 OPTION_MASK_ISA_MOVDIRI_UNSET,
2158 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2159 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2160 * config.gcc (movdirintrin.h): New header.
2161 * config/i386/cpuid.h (bit_MOVDIRI,
2162 bit_MOVDIR64B): New bits.
2163 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2165 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2166 (VOID, PVOID, PCVOID)): New function types.
2167 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2168 __builtin_ia32_directstoreu_u64,
2169 __builtin_ia32_movdir64b): New builtins.
2170 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2171 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2173 (ix86_valid_target_attribute_inner_p): Ditto.
2174 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2175 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2176 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2177 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2178 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2179 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2180 (movdiri<mode>, movdir64b_<mode>): New.
2181 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2182 * config/i386/immintrin.h: Include movdirintrin.h.
2183 * config/i386/movdirintrin.h: New file.
2184 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2186 2018-04-19 Richard Biener <rguenther@suse.de>
2189 * cfg.c (clear_bb_flags): When loop state says we have
2190 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2192 2018-04-19 Richard Biener <rguenther@suse.de>
2194 PR tree-optimization/84737
2195 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2196 copying restrict info.
2197 (vect_setup_realignment): Use it.
2198 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2199 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2200 the first DR to all generated stores.
2201 (vectorizable_load): Likewise for loads.
2203 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2205 PR tree-optimization/85446
2206 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2207 the integral and pointer types to have the same precision.
2209 * doc/install.texi: Document --disable-cet being the default and
2212 2018-04-18 Martin Liska <mliska@suse.cz>
2214 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2217 2018-04-18 Martin Liska <mliska@suse.cz>
2220 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2223 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2224 arguments if they are comparable.
2226 2018-04-18 Martin Liska <mliska@suse.cz>
2229 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2232 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2235 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2238 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2239 ENDBR after calling __morestack.
2241 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2244 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2245 by using gcc_base_ver to generate a gcc_driver_version, and use
2246 it when generating GCC_DRIVER_NAME.
2247 * configure: Regenerate.
2249 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2252 * config.gcc: Obsolete powerpc*-*-*spe*.
2254 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2257 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2258 (stabstr_D): Change type of unum from unsigned int to
2259 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2262 2018-04-17 Jim Wilson <jimw@sifive.com>
2265 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2266 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2267 Set arg_pointer_offset after using pretend_args_size.
2269 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2271 PR rtl-optimization/85431
2272 * dse.c (record_store): Ignore zero width stores.
2275 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2276 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2277 __builtin_stack_restore rather than after it.
2278 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2279 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2280 argument instead of virtual_dynamic_stack_rtx.
2282 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2284 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2286 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2287 Add note to error message to explain internal mapping of overloaded
2288 built-in function name to non-overloaded built-in function name.
2289 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2292 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2295 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2296 where the inputs overlap with the output.
2298 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2300 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2301 (=v, v) alternative and explicit "memory" attribute.
2302 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2303 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2305 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2306 "sselog1" type instead of "sselog".
2307 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2308 "sselog". Remove explicit "memory" attribute.
2309 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2310 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2312 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2313 "sselog1" type instead of "sselog". Remove explicit "memory"
2315 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2316 use "sselog1" type instead of "sselog". Remove explicit "memory"
2318 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2319 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2321 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2322 "sselog1" type instead of "sselog". Remove explicit "memory"
2324 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2325 use "sselog1" type instead of "sselog". Remove explicit "memory"
2329 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2332 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2335 2018-04-17 Martin Jambor <mjambor@suse.cz>
2338 * ipa-cp.c (create_specialized_node): Call
2339 expand_all_artificial_thunks if necessary.
2341 2018-04-17 Martin Liska <mliska@suse.cz>
2344 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2345 in message, remote space in between '_G' and '('.
2347 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2350 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2351 avx512f_vmcmp<mode>3<round_saeonly_name>,
2352 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2353 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2354 avx512f_rndscale<mode><round_saeonly_name>,
2355 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2356 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2357 Use %<iptr>2 instead of %2 for -masm=intel.
2358 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2359 avx512f_vcvttss2usi<round_saeonly_name>,
2360 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2362 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2363 avx512f_vcvttsd2usi<round_saeonly_name>,
2364 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2365 Use %q1 instead of %1 for -masm=intel.
2366 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2367 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2368 of %3 for -masm=intel.
2369 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2371 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2373 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2375 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2376 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2378 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2380 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2381 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2382 %g1 and one with %0 and %1.
2383 (avx512er_vmrcp28<mode><round_saeonly_name>,
2384 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2386 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2387 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2388 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2389 of %0 and %{%4%} for -masm=intel.
2390 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2391 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2392 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2393 order of %0 and %{%5%}%{z%} for -masm=intel.
2395 2018-04-17 Jan Hubicka <jh@suse.cz>
2398 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2400 2018-04-17 Martin Liska <mliska@suse.cz>
2403 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2404 for target_clone error message. Make default implementation
2405 clone to be a local declaration.
2406 (separate_attrs): Add new argument and check for an empty
2408 (expand_target_clones): Handle it.
2409 (ipa_target_clone): Make redirection just for target_clones
2412 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2413 Tom de Vries <tom@codesourcery.com>
2416 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2417 tiled basic blocks without omp continue statements.
2419 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2422 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2423 vec_extract expression as having side effects to make sure it gets
2426 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2429 * config/i386/i386.c (get_builtin_code_for_version): Check
2432 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2435 * gcc/config.gcc: Support "skylake".
2436 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2438 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2439 (processor_target_table): Add "skylake".
2440 (ix86_option_override_internal): Add "skylake".
2441 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2442 PROCESSOR_CANNONLAKE.
2443 (get_builtin_code_for_version): Fix priority for
2444 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2445 PROCESSOR_SKYLAKE-AVX512.
2446 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2447 (processor_type): Add PROCESSOR_SKYLAKE.
2449 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2450 Jason Merrill <jason@redhat.com>
2453 * convert.c (convert_to_integer_1): Use direct recursion for
2454 enumeral types and types with a precision less than the number
2455 of bits in their mode.
2457 2018-04-16 Julia Koval <julia.koval@intel.com>
2460 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2461 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2463 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2466 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2467 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2468 and -mno-direct-move.
2470 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2473 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2474 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2475 Ensure negative shifts result in {0}.
2477 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2479 PR rtl-optimization/79916
2480 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2481 regs (if any) to define how to gnerate SD moves when LRA is in
2484 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2486 PR rtl-optimization/85393
2487 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2488 * except.c (expand_dw2_landing_pad_for_region): Make static.
2489 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2490 a label and unconditional jump to old_bb, rather than
2491 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2494 PR rtl-optimization/85376
2495 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2496 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2497 instead of a specific value.
2499 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2500 Bin Cheng <bin.cheng@arm.com>
2502 PR tree-optimization/82965
2503 PR tree-optimization/83991
2504 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2505 by_profile_only parameter.
2506 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2507 information if the loop was predicted to iterate too many times.
2508 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2510 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2513 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2516 2018-04-13 Martin Liska <mliska@suse.cz>
2517 Jakub Jelinek <jakub@redhat.com>
2520 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2521 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2522 * builtins.c (expand_builtin_memory_copy_args): Use
2523 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2524 handle dest_addr == pc_rtx.
2526 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2529 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2530 asked to not generate direct moves.
2531 (fix_trunc<mode>si2_stfiwx): Similar.
2532 (fix_trunc<mode>si2_internal): Similar.
2534 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2537 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2538 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2539 lookup if dest in some wider mode is known to be const0_rtx and
2540 if so, record permanent equivalence for it to be ZERO_EXTEND of
2541 the narrower mode destination.
2543 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2545 * lto-streamer-out.c (output_function): Revert 259346.
2546 * omp-expand.c (expand_oacc_for): Likewise.
2548 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2550 PR rtl-optimization/85354
2551 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2552 * sel-sched.c (sel_global_init): ... here.
2554 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2557 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2558 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2559 mode for PE-COFF targets.
2560 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2561 (i386_pe_asm_lto_end): Likewise.
2562 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2563 (TARGET_ASM_LTO_END): Likewise.
2564 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2565 (i386_pe_asm_lto_start): New function.
2566 (i386_pe_asm_lto_end): Likewise.
2568 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2569 Richard Biener <rguenther@suse.de>
2572 * lto-streamer-out.c (output_function): Fix CFG loop state before
2574 * omp-expand.c (expand_oacc_for): Handle calls to internal
2575 functions like regular functions.
2577 2018-04-12 Richard Biener <rguenther@suse.de>
2580 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2581 for the early LTO debug to properly generate references to it
2582 during DIE emission. Do not re-use that for the skeleton for
2584 (dwarf2out_early_finish): Likewise.
2586 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2589 * config/i386/sse.md
2590 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2591 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2592 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2593 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2594 and output is a reg, avoid creating invalid lowpart subreg, but
2595 instead split into a 512-bit move. Don't split if not AVX512VL,
2596 input is xmm16+ reg and output is a mem.
2597 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2598 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2599 xmm16+ reg and output is a mem.
2601 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2603 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2604 also for flag_dwarf2_cfi_asm.
2606 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2608 PR rtl-optimization/85342
2609 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2610 a bool scalar var inside of the loop instead. Don't try to update
2611 recog_data.operand after failed apply_change_group.
2613 2018-04-12 Tom de Vries <tom@codesourcery.com>
2616 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2617 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2618 array with flexible array member as array without given dimension.
2619 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2620 argument for undefined param to true.
2622 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2625 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2626 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2627 from PowerPC section.
2628 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2629 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2630 help text that is too long.
2631 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2632 help text that is too long.
2633 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2634 help text that is too long.
2636 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2638 * config/alpha/alpha.md (stack_probe_internal): Rename
2639 from "probe_stack". Update all callers.
2641 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2643 PR rtl-optimization/84566
2644 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2645 sched_macro_fuse_insns.
2647 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2650 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2651 (compute_block_dependences): ... from here.
2653 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2655 PR tree-optimization/85331
2656 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2657 from int to HOST_WIDE_INT.
2659 2018-04-11 Martin Jambor <mjambor@suse.cz>
2662 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2663 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2664 not the same as the source val.
2665 (cgraph_edge_brings_value_p): New parameter.
2666 (gather_edges_for_value): Pass destination value to
2667 cgraph_edge_brings_value_p.
2668 (perhaps_add_new_callers): Likewise.
2669 (get_info_about_necessary_edges): Likewise and exclude values brought
2670 only by self-recursive edges.
2671 (create_specialized_node): Redirect only clones of self-calling edges.
2672 (+self_recursive_pass_through_p): New function.
2673 (find_more_scalar_values_for_callers_subset): Use it.
2674 (find_aggregate_values_for_callers_subset): Likewise.
2675 (known_aggs_to_agg_replacement_list): Removed.
2676 (decide_whether_version_node): Re-calculate known constants for all
2677 remaining context clones.
2679 2018-04-11 Richard Biener <rguenther@suse.de>
2682 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2683 from early DWARF output.
2684 (dwarf2out_early_finish): Output line info unconditionally into
2685 early DWARF and add reference to it.
2687 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2690 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2691 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2692 other than V2DFmode using iptr mode attribute.
2693 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2695 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2697 PR rtl-optimization/84659
2698 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2700 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2703 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2705 (output_loc_list): Pass address of a dummy size variable even in the
2706 locview handling loop.
2707 (index_location_lists): Add comment on why skip_loc_list_entry can't
2710 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2713 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2716 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2719 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2720 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2721 and -mstring-compare-inline-limit.
2723 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2726 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2727 for stack clash protection in a register whenever we need it to be in
2730 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2732 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2733 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2735 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2738 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2740 (mlong-double-): Ditto.
2741 * config/rs6000/sysv4.opt (msdata=): Ditto.
2742 (mtls-size=): Ditto.
2744 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2746 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2747 erroneous entries for
2748 "vector int vec_ldl (int, long int *)", and
2749 "vector unsigned int vec_ldl (int, unsigned long int *)".
2750 Add comments and entries for
2751 "vector bool char vec_ldl (int, bool char *)",
2752 "vector bool short vec_ldl (int, bool short *)",
2753 "vector bool int vec_ldl (int, bool int *)",
2754 "vector bool long long vec_ldl (int, bool long long *)",
2755 "vector pixel vec_ldl (int, pixel *)",
2756 "vector long long vec_ldl (int, long long *)",
2757 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2758 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2759 type tree bool_long_long_type_node and correct definition of
2760 bool_V2DI_type_node to make reference to this new type tree.
2761 (rs6000_mangle_type): Replace erroneous reference to
2762 bool_long_type_node with bool_long_long_type_node.
2763 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2764 comments to emphasize sign distinctions for char and int types and
2765 replace RS6000_BTI_bool_long constant with
2766 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2767 use of RS6000_BTI_pixel.
2768 (bool_long_type_node): Remove this macro definition.
2769 (bool_long_long_type_node): New macro definition
2771 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2773 PR rtl-optimization/85300
2774 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2775 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2776 simplify_unary_operation fails.
2778 2018-04-10 Martin Liska <mliska@suse.cz>
2780 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2781 cgraph_edge and ipa_ref.
2783 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2787 * config/i386/sse.md
2788 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2789 computation of the VEC_MERGE selector from mask.
2790 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2791 Fix decoding of the VEC_MERGE selector into mask.
2793 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2795 PR tree-optimization/85286
2796 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2798 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2800 * final.c (final_1): Set insn_last_address as well as
2801 insn_current_address.
2803 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2806 * explow.c (emit_stack_probe): Call validize_mem on memory location
2807 before passing it to gen_probe_stack. Create address operand and
2808 legitimize it for the probe_stack_address case.
2810 2018-04-09 Jan Hubicka <jh@suse.cz>
2813 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2814 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2815 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2817 (free_lang_data): Rebuild type inheritance graph.
2819 2018-04-09 Martin Sebor <msebor@redhat.com>
2821 * invoke.texi (-finline-small-functions): Mention other optimization
2823 (-findirect-inlining, -fpartial-inlining): Same.
2824 (-finline-functions-called-once): Same.
2825 (-freorder-blocks-and-partition): Same.
2827 2018-04-09 Jan Hubicka <jh@suse.cz>
2830 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2831 jumps; choose last target that matches the criteria (i.e.
2832 no partition changes for non-crossing jumps).
2833 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2834 support for redirecting crossing jumps to non-crossing.
2836 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2838 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2839 also for naked functions.
2841 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2843 * config/arc/arc.md (add_shift): New pattern.
2844 (add_shift2): Likewise.
2845 (sub_shift): Likewise.
2846 (sub_shift_cmp0_noout): Likewise.
2847 (compare_si_ashiftsi): Likewise.
2848 (xbfu_cmp0_noout): New combine pattern.
2849 (xbfu_cmp0"): Likewise.
2850 (movsi_set_cc_insn): Place the predicable variant first.
2851 (commutative_binary_cmp0_noout): Remove clobber.
2852 (commutative_binary_cmp0): New pattern.
2853 (noncommutative_binary_cmp0): Likewise.
2854 (noncommutative_binary_cmp0_noout): Likewise.
2855 (noncommutative_binary_comparison_result_used): Removed.
2856 (rsub_cmp0): New pattern.
2857 (rsub_cmp0_noout): Likewise.
2858 (extzvsi): Changed, keep only meaningful variants.
2859 (SQH, SEZ): New iterators.
2860 (SQH_postfix): New mode attribute.
2861 (SEZ_prefix): New code attribute.
2862 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2863 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2864 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2866 (noncommutative_operator): Check the availability of barrel
2869 2018-04-09 Richard Biener <rguenther@suse.de>
2871 PR tree-optimization/85284
2872 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2873 Only use the niter constraining form of simple_iv when the exit
2876 2018-04-09 Tom de Vries <tom@codesourcery.com>
2879 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2880 (define_expand "*memory_barrier"): New define_expand.
2881 (define_insn "memory_barrier"): New insn.
2883 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2885 PR rtl-optimization/80463
2886 PR rtl-optimization/83972
2887 PR rtl-optimization/83480
2889 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2890 correct producer for the insn.
2891 (tidy_control_flow): Fixup seqnos in case of debug insns.
2893 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2895 PR rtl-optimization/83913
2897 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2898 different sched-times when merging exprs.
2900 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2902 PR rtl-optimization/83962
2904 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2905 tidy_fallthru_edge and tidy_control_flow.
2907 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2909 PR rtl-optimization/83530
2911 * sel-sched.c (force_next_insn): New global variable.
2912 (remove_insn_for_debug): When force_next_insn is true, also leave only
2913 next insn in the ready list.
2914 (sel_sched_region): When the region wasn't scheduled, make another pass
2915 over it with force_next_insn set to 1.
2917 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2919 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2921 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2922 for interrupt control.
2923 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2924 functions for interrupt control.
2925 * config/nds32/nds32-intrinsic.md: Likewise.
2926 * config/nds32/nds32_intrinsic.h: Likewise.
2927 * config/nds32/nds32.h (nds32_builtins): Likewise.
2929 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2931 * config/nds32/nds32.c (nds32_init_machine_status,
2932 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2933 strict_aligned_p field.
2934 (nds32_expand_to_rtl_hook): New function.
2935 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2936 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2938 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2939 Chung-Ju Wu <jasonwucj@gmail.com>
2941 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2942 * config/nds32/nds32-n7.md: New file.
2943 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2944 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2946 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2947 * config/nds32/nds32.md (pipeline_model): Add n7.
2948 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2949 * config/nds32/pipelines.md: Include n7 settings.
2951 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2952 Chung-Ju Wu <jasonwucj@gmail.com>
2954 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2955 * config/nds32/nds32-e8.md: New file.
2956 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2957 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2959 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2960 * config/nds32/nds32.md (pipeline_model): Add e8.
2961 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2962 * config/nds32/pipelines.md: Include e8 settings.
2964 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2965 Chung-Ju Wu <jasonwucj@gmail.com>
2967 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2968 * config/nds32/nds32-n8.md: New file.
2969 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2970 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2972 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2973 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2974 * config/nds32/nds32.md (pipeline_model): Add n8.
2975 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2976 * config/nds32/pipelines.md: Include n8 settings.
2978 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2979 Chung-Ju Wu <jasonwucj@gmail.com>
2981 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2982 * config/nds32/nds32-n9-2r1w.md: New file.
2983 * config/nds32/nds32-n9-3r2w.md: New file.
2984 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2985 nds32_register_ports): New or modify for cpu n9.
2986 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2988 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2989 * config/nds32/nds32-utils.c: New file.
2990 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2991 TARGET_MUL_SLOW): Define.
2992 * config/nds32/nds32.md (pipeline_model): New attribute.
2993 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2994 New options that support cpu n9.
2995 * config/nds32/pipelines.md: Include n9 settings.
2996 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2998 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3000 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3001 information if necessary.
3002 (output_cond_branch_compare_zero): Likewise.
3003 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3004 (nds32_target_alignment): Refine for alignment.
3005 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3006 (FUNCTION_BOUNDARY): Modify.
3007 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3009 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3011 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3013 * config/nds32/constants.md (unspec_volatile_element): Add values for
3014 TLB operation and data prefetch.
3015 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3016 functions for TLB operation and data prefetch.
3017 * config/nds32/nds32-intrinsic.md: Likewise.
3018 * config/nds32/nds32_intrinsic.h: Likewise.
3019 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3020 (nds32_print_operand): Likewise.
3021 * config/nds32/nds32.h (nds32_builtins): Likewise.
3023 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3024 Andrew Pinski <pinsika@gcc.gnu.org>
3027 * match.pd: Use constant_boolean_node of correct type instead of
3028 boolean_true_node or boolean_false_node for simplifying
3029 pointer comparisons to zero.
3031 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3033 PR tree-optimization/80021
3034 * tree.c (verify_type_variant): Make error call in verify_variant_match
3035 translatable and remove final full stop.
3037 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3039 * config/nds32/constants.md (unspec_volatile_element): Add
3040 UNSPEC_VOLATILE_EH_RETURN.
3041 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3042 nds32_output_stack_pop): Support dwarf exception handling process.
3043 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3044 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3045 exception handling process.
3046 (nds32_compute_stack_frame): Likewise.
3047 (nds32_return_addr_rtx): Likewise.
3048 (nds32_initial_elimination_offset): Likewise.
3049 (nds32_expand_prologue): Likewise.
3050 (nds32_expand_epilogue): Likewise.
3051 (nds32_dynamic_chain_address): New function.
3052 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3054 (DYNAMIC_CHAIN_ADDRESS): Define.
3055 (EH_RETURN_DATA_REGNO): Define.
3056 (EH_RETURN_STACKADJ_RTX): Define.
3057 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3058 patterns for dwarf exception handling.
3060 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3062 * config/nds32/nds32.h: Clean up obsolete macros.
3064 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3066 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3067 Add enum values for particular instructions.
3068 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3069 particular intrinsic functions.
3070 * config/nds32/nds32-intrinsic.md: Likewise.
3071 * config/nds32/nds32_intrinsic.h: Likewise.
3072 * config/nds32/nds32.h (nds32_builtins): Likewise.
3073 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3074 (btst, ave): New patterns for particular instructions.
3076 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3078 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3079 Add enum values for atomic load/store and memory sync.
3080 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3082 * config/nds32/nds32-intrinsic.md: Likewise.
3083 * config/nds32/nds32_intrinsic.h: Likewise.
3084 * config/nds32/nds32.h (nds32_builtins): Likewise.
3086 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3088 PR tree-optimization/85257
3089 * fold-const.c (native_encode_vector): If not all elts could fit
3090 and off is -1, return 0 rather than offset.
3091 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3092 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3093 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3094 adjust buffer in native_interpret_expr call.
3096 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3098 * config/nds32/constants.md (unspec_volatile_element): Add cache
3099 control enum values.
3100 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3101 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3102 * config/nds32/nds32.c (nds32_cctl_names): New.
3103 (nds32_print_operand): Handle cache control register names.
3104 * config/nds32/nds32.h (nds32_builtins): New enum values.
3105 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3107 * config/nds32/nds32.md (type): Add mmu.
3108 * config/nds32/pipelines.md (simple_insn): Add mmu.
3110 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3112 * config/nds32/nds32.md (type): Remove call.
3113 * config/nds32/pipelines.md (simple_insn): Likewise.
3115 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3117 * config/nds32/constants.md (unspec_volatile_element): Add
3118 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3119 UNSPEC_VOLATILE_FMFCFG.
3120 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3121 description for fmfcfg and fmfcsr.
3122 (bdesc_1arg): Add fmtcsr.
3123 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3124 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3125 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3126 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3127 unspec_fmfcfg): New patterns.
3128 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3129 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3130 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3131 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3132 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3133 __nds32__fmfcfg): Define.
3135 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3137 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3138 intrinsic register names.
3139 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3140 intrinsic register enum values and macros.
3142 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3144 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3145 for load/store addressing form.
3146 (nds32_print_operand_address): Likewise.
3148 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3151 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3152 based on LABEL_REF. Remove useless assertion.
3153 (pic_address_needs_scratch): Fix formatting.
3154 (sparc_legitimize_pic_address): Minor tweaks.
3155 (sparc_delegitimize_address): Adjust assertion accordingly.
3156 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3157 into symbolic_operand.
3158 (movsi_high_pic_label_ref): Likewise.
3159 (movsi_lo_sum_pic_label_ref): Likewise.
3160 (movdi_pic_label_ref): Likewise.
3161 (movdi_high_pic_label_ref): Likewise.
3162 (movdi_lo_sum_pic_label_ref): Likewise.
3164 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3166 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3167 custom LIB_SPEC setup.
3169 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3170 Kito Cheng <kito.cheng@gmail.com>
3172 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3173 * config/riscv/freebsd.h: New.
3175 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3177 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3178 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3181 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3182 Kito Cheng <kito.cheng@gmail.com>
3184 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3185 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3186 * config/nds32/nds32-protos.h (nds32_output_call,
3187 nds32_output_return): Declare.
3188 * config/nds32/nds32.md: Refine all the call and return patterns.
3190 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3193 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3194 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3196 PR rtl-optimization/84872
3197 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3198 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3201 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3203 * expr.c (copy_blkmode_to_reg): Revert 254862.
3204 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3206 2018-04-06 Richard Biener <rguenther@suse.de>
3209 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3210 after seeing a component reference with an adjacent field. Treat
3211 refs to arrays at struct end of external decls similar to
3212 refs to unconstrained commons.
3214 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3217 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3218 look through SAVE_EXPRs with non-side-effects argument. Adjust
3220 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3223 2018-04-06 Richard Biener <rguenther@suse.de>
3226 * alias.c (find_base_term): New wrapper around find_base_term
3227 unwinding CSELIB_VAL_PTR changes.
3228 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3231 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3233 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3235 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3236 constant definitions.
3237 ("nop"): lr 0,0 -> nopr r0
3238 ("nop_lr0", "nop_lr1"): New insn definitions.
3240 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3242 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3243 NDS32_V3PUSH_AVAILABLE_P macro.
3245 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3246 Chung-Ju Wu <jasonwucj@gmail.com>
3248 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3249 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3250 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3251 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3252 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3253 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3254 * config/nds32/constraints.md: New constraints and checking for hard
3255 float configuration.
3256 * config/nds32/iterators.md: New mode iterator and attribute for hard
3257 float configuration.
3258 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3260 * config/nds32/nds32-fpu.md: New file.
3261 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3262 deal with hard float code generation.
3263 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3265 (abi_type, float_reg_number): New enum type.
3266 * config/nds32/nds32-predicates.c: New predicates for hard float.
3267 * config/nds32/nds32-protos.h: Declare functions for hard float.
3268 * config/nds32/nds32.c: Implementation for hard float configuration.
3269 * config/nds32/nds32.h: Definitions for hard float configuration.
3270 * config/nds32/nds32.md: Include hard float machine description and
3271 modify patterns for hard float configuration.
3272 * config/nds32/nds32.opt: New options for hard float configuration.
3273 * config/nds32/predicates.md: New predicates for hard float
3276 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3278 * common/config/nds32/nds32-common.c
3279 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3281 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3284 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3285 CONSTRUCTOR_ELT (ctor, ...)->value.
3287 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3290 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3292 2018-04-05 Tom de Vries <tom@codesourcery.com>
3295 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3298 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3299 Kito Cheng <kito.cheng@gmail.com>
3301 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3302 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3303 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3304 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3306 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3307 Kito Cheng <kito.cheng@gmail.com>
3309 * config/nds32/constraints.md (Ufe): New memory constraint.
3310 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3311 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3312 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3314 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3315 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3317 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3319 * config/nds32/nds32.md: Use optimize_size in the condition for
3320 alu-shift instructions.
3322 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3324 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3326 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3328 * config/nds32/nds32.md (negsi2): Refine pattern.
3330 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3331 Chung-Ju Wu <jasonwucj@gmail.com>
3333 * config/nds32/iterators.md (shift_rotate): New code iterator.
3334 (shift): New code attribute.
3335 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3336 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3337 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3338 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3339 bit-wise operations.
3340 (andsi3, *andsi3): Ditto.
3341 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3342 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3343 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3344 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3345 nds32_ior_operand, nds32_xor_operand): New predicates.
3347 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3349 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3350 (addsi3, subsi3): ... this.
3352 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3354 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3356 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3358 * config/nds32/nds32.md: Adjust indention.
3360 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3362 * config/nds32/nds32.md (feature): New attribute.
3364 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3366 * config/nds32/nds32.md (subtype): New attribute.
3368 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3371 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3372 expansion to perform a bitwise AND of the argument followed by a
3373 boolean negation of the result.
3375 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3377 PR rtl-optimization/84878
3378 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3379 the basic block. Assert the use reference is not artificial and that
3380 it has an associated insn.
3382 2018-04-04 Michael Matz <matz@suse.de>
3384 * builtins.c (compute_objsize): Pass correct operand
3385 to array_at_struct_end_p.
3387 2018-04-04 Richard Biener <rguenther@suse.de>
3390 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3391 from contexts for DINFO_LEVEL_TERSE and below.
3393 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3395 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3396 resiter_operand condition.
3397 * config/nds32/nds32.md (*move<mode>): Ditto.
3399 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3400 Monk Chiang <sh.chiang04@gmail.com>
3402 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3404 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3406 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3408 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3409 Kito Cheng <kito.cheng@gmail.com>
3411 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3412 nds32_cond_code_str, output_cond_branch,
3413 output_cond_branch_compare_zero, nds32_expand_cbranch,
3414 nds32_expand_cstore, nds32_expand_movcc,
3415 nds32_output_cbranchsi4_equality_zero,
3416 nds32_output_cbranchsi4_equality_reg,
3417 nds32_output_cbranchsi4_equality_reg_or_const_int,
3418 nds32_output_cbranchsi4_greater_less_zero: New functions.
3419 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3420 nds32_expand_cstore, nds32_expand_movcc,
3421 nds32_output_cbranchsi4_equality_zero,
3422 nds32_output_cbranchsi4_equality_reg,
3423 nds32_output_cbranchsi4_equality_reg_or_const_int,
3424 nds32_output_cbranchsi4_greater_less_zero): Declare.
3425 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3426 nds32_rimm11s_operand): New predicates.
3427 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3428 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3431 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3433 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3434 * config/nds32/nds32.md: Ditto.
3435 * config/nds32/pipelines.md: Ditto.
3437 2018-04-04 Richard Biener <rguenther@suse.de>
3439 PR tree-optimization/85168
3440 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3441 propagating abnormals.
3443 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3445 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3447 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3448 Kito Cheng <kito.cheng@gmail.com>
3450 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3451 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3452 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3453 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3454 * config/nds32/nds32.md (sibcall_internal): New.
3455 (sibcall_register): Remove.
3456 (sibcall_immediate): Remove.
3457 (sibcall_value_internal): New.
3458 (sibcall_value_register): Remove.
3459 (sibcall_value_immediate): Remove.
3460 * config/nds32/predicates.md (nds32_general_register_operand): New.
3461 (nds32_call_address_operand): New.
3463 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3465 PR rtl-optimization/85167
3466 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3467 bb_defs if *split_p, instead preinitialize it to NULL.
3469 PR tree-optimization/85156
3470 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3471 evaluating the argument multiple times.
3473 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3475 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3477 (_mm_cvtpd_ps): Likewise.
3478 (_mm_cvttpd_epi32): Likewise.
3479 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3480 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3481 vector, pixel, and bool following altivec.h include.
3483 2018-04-03 Martin Sebor <msebor@redhat.com>
3485 * doc/extend.texi (Common Function Attributes): Clarify.
3486 (const attribute): Likewise.
3487 (pure attribute): Likewise.
3489 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3492 * config/i386/i386.c (ix86_expand_vector_set): Use
3493 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3495 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3497 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3498 instructions when changing rounding bits to preserve precision bits
3499 in the x87 control word.
3501 2018-04-03 Martin Liska <mliska@suse.cz>
3503 PR tree-optimization/82491
3504 * rtl.h (strip_offset_and_add): Replace += suboffset with
3505 poly_uint64 () + suboffset.
3507 2018-03-29 Martin Liska <mliska@suse.cz>
3508 Martin Jambor <mjambor@suse.cz>
3511 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3512 param_type is not an integral or pointer type.
3514 2018-04-03 Richard Biener <rguenther@suse.de>
3516 * sese.h (recompute_all_dominators): Remove.
3518 2018-04-02 Martin Sebor <msebor@redhat.com>
3520 * doc/invoke.texi (-Wrestrict): Fix typos.
3522 2018-04-02 Jim Wilson <jimw@sifive.com>
3524 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3525 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3526 (<optab>di3, <optab>si3_extend): Likewise.
3527 (<optab>si3_mask, <optab>si3_mask_1): New.
3528 (<optab>di3_mask, <optab>di3_mask_1): New.
3529 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3530 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3531 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3533 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3535 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3538 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3540 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3541 (nds32_canonicalize_comparison): New function.
3543 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3544 Kito Cheng <kito.cheng@gmail.com>
3545 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3547 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3548 * config/nds32/constants.md (unspec_volatile_element): Add
3549 UNSPEC_VOLATILE_RELAX_GROUP.
3550 * config/nds32/nds32-relax-opt.c: New file.
3551 * config/nds32/nds32-predicates.c
3552 (nds32_symbol_load_store_p): New function.
3553 * config/nds32/nds32-protos.h
3554 (nds32_symbol_load_store_p): Declare function.
3555 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3556 * config/nds32/nds32.c
3557 (nds32_register_pass): New function to register pass.
3558 (nds32_register_passes): New function to register passes.
3559 * config/nds32/nds32.md (relax_group): New pattern.
3560 * config/nds32/nds32.opt (mrelax-hint): New option.
3561 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3563 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3565 * config/nds32/t-nds32: Modify files dependency.
3567 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3569 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3570 (PROFILE_HOOK): Define its implementation.
3572 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3574 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3575 type and 32-bit size.
3577 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3580 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3581 (V_128_256): New mode iterator.
3582 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3583 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3584 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3586 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3587 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3589 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3592 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3593 NaN inputs correctly.
3595 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3598 * config/rs6000/vsx.md (??r): New mode attribute.
3599 (*vsx_mov<mode>_64bit): Use it.
3600 (*vsx_mov<mode>_32bit): Likewise.
3602 2018-03-30 Martin Sebor <msebor@redhat.com>
3604 PR tree-optimization/84818
3605 * builtins.c (check_access): Use warning_n.
3607 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3610 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3612 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3615 2018-03-30 Julia Koval <julia.koval@intel.com>
3618 * x86-tune.def (movx, partial_reg_dependency): Enable for
3621 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3624 * lra-constraints.c (process_alt_operands): Move setting
3625 this_alternative_matches below.
3627 2018-03-29 Martin Liska <mliska@suse.cz>
3630 * doc/invoke.texi: Document how LTO works with debug info.
3631 Describe auto-load support of binutils. Mention 'x86-64'
3632 as valid option value of -march option.
3634 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3636 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3639 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3640 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3641 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3644 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3647 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3648 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3649 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3650 for RS6000_BTM_POWERPC64.
3651 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3652 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3653 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3658 2018-03-28 Carl Love <cel@us.ibm.com>
3661 2017-09-27 Carl Love <cel@us.ibm.com>
3663 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3664 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3665 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3668 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3670 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3671 instead of __vector bool.
3672 (_mm_max_pu8): Likewise.
3673 (_mm_min_pi16): Likewise.
3675 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3678 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3679 (DIVWEUO): Likewise.
3681 (DIVDEUO): Likewise.
3682 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3683 DIVWEUO and DIVDEUO.
3684 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3685 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3686 (div_extend): Likewise.
3687 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3689 (__builtin_divweuo): Likewise.
3690 (__builtin_divdeo): Likewise.
3691 (__builtin_divdeuo): Likewise.
3693 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3696 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3697 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3699 PR tree-optimization/82004
3700 * gimple-match-head.c (optimize_pow_to_exp): New function.
3701 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3702 Don't fold to exp if optimize_pow_to_exp is false.
3704 2018-03-28 Martin Liska <mliska@suse.cz>
3707 * calls.c (initialize_argument_information): Fix trailing space.
3708 * common.opt: Fix typo and provide better explanation for
3709 -fsanitize-coverage option.
3710 * config/i386/i386.opt: Fix typo.
3712 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3713 Martin Liska <mliska@suse.cz>
3716 * gimplify.c (asan_poison_variable): Don't do the check for
3717 gimplify_omp_ctxp here.
3718 (gimplify_decl_expr): Do it here.
3719 (gimplify_target_expr): Likewise.
3721 2018-03-28 Martin Liska <mliska@suse.cz>
3724 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3725 chkp_type_bounds_count if MPX is not enabled.
3727 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3729 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3731 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3734 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3735 function to create the function decl for complex long double
3736 multiply and divide for -mabi=ieeelongdouble.
3737 (init_float128_ieee): Call it.
3739 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3742 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3743 -fcf-protection=branch -mibt.
3744 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3746 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3749 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3751 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3754 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3755 extern array declarations.
3757 2018-03-27 Richard Biener <rguenther@suse.de>
3760 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3761 explicit single_use checks.
3763 2018-03-27 Richard Biener <rguenther@suse.de>
3765 PR tree-optimization/85082
3766 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3769 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3771 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3772 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3773 Turn on fasynchronous-unwind-tables and funwind-tables.
3775 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3778 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3779 (*bmi_blsr_<mode>_ccz): Ditto.
3781 2018-03-26 Tom de Vries <tom@codesourcery.com>
3783 PR tree-optimization/85063
3784 * omp-general.c (offloading_function_p): New function. Factor out
3786 * omp-offload.c (pass_omp_target_link::gate): ... here.
3787 * omp-general.h (offloading_function_p): Declare.
3788 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3789 with attribute omp declare target for offloading functions.
3791 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3793 PR tree-optimization/84005
3794 * tree-data-ref.h (get_base_for_alignment): Declare.
3795 * tree-data-ref.c (get_base_for_alignment_1): New function.
3796 (get_base_for_alignment): Likewise.
3797 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3798 get_base_for_alignment to find a suitable base object, instead
3799 of always using drb->base_address.
3801 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3804 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3805 known size by default.
3807 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3810 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3811 and non BLKmode operands.
3813 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3816 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3817 Clean up attributes.
3819 2018-03-23 Richard Biener <rguenther@suse.de>
3822 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3823 we are going to emit early debug for LTO.
3825 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3828 * function.c (match_asm_constraints_1): Don't optimize if input
3829 doesn't satisfy general_operand predicate for output's mode.
3832 * alias.c (write_dependence_p): Don't require for x_canonicalized
3833 non-VOIDmode if x has VOIDmode.
3836 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3837 just don't try to optimize it rather than assert it never happens.
3839 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3841 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3842 macro expansions for definition of ST_INTERNAL_<mode> and
3843 LD_INTERNAL_<mode> builtins.
3844 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3846 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3848 (altivec_expand_st_builtin): Likewise.
3849 (altivec_expand_builtin): Remove calls to deleted functions.
3850 (rs6000_address_for_altivec): Delete this function.
3851 * config/rs6000/vector.md: Remove expands for
3852 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3854 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3857 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3858 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3859 re-computing once computed.
3860 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3861 (arm_init_machine_status): Initialize
3862 machine->static_chain_stack_bytes.
3864 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3867 * doc/extend.texi: Add four new prototypes for vec_ld.
3868 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3869 definitions for more logical presentation.
3870 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3871 entries for V1TI variants of __builtin_altivec_ld builtin.
3872 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3873 handling of V1TI variant of LVX icode pattern.
3874 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3875 (rs6000_gimple_fold_builtin): Likewise.
3876 (altivec_init_builtins): Add code to define
3877 __builtin_altivec_lvx_v1ti function.
3879 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3882 * function.c (match_asm_constraints_1): Don't do the optimization
3883 if input isn't a REG, SUBREG, MEM or constant.
3885 2018-03-22 Tom de Vries <tom@codesourcery.com>
3887 PR tree-optimization/84956
3888 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3889 bb_has_abnormal_pred.
3891 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3894 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3895 DECL_INITIAL (decl) to decl at the end.
3896 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3899 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3901 * doc/extend.texi (__builtin_tgmath): Document when complex
3902 integer types are treated as _Complex _Float64.
3904 2018-03-21 Tom de Vries <tom@codesourcery.com>
3906 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3908 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3910 PR tree-optimization/84960
3911 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3912 if it is ENTRY block, move them into single succ of ENTRY in that case.
3914 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3916 PR tree-optimization/84811
3917 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3918 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3919 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3920 (known_subrange_p): Update accordingly. Cast each value involved
3921 in the size comparison, rather than casting the result of the
3924 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3926 PR tree-optimization/84982
3927 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3928 by flipping the least significant bit rather than all bits from
3929 bitpos to bitpos + bitsize - 1.
3931 2018-03-21 Nathan Sidwell <nathan@acm.org>
3933 * doc/extend.texi (Deprecated Features): Remove mention of
3934 long-deleted deprecations.
3936 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3939 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3940 * configure: Regenerate.
3942 2018-03-21 Tom de Vries <tom@codesourcery.com>
3944 PR tree-optimization/83126
3945 * tree-parloops.c (num_phis): New function.
3946 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3948 2018-03-21 Nathan Sidwell <nathan@acm.org>
3950 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3951 mention anon-struct/union members and trailing attributes.
3953 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3955 PR tree-optimization/84969
3956 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3957 builtin memset partitions if they set different rhs values.
3959 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3961 PR rtl-optimization/84989
3962 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3963 VEC_DUPLICATE with scalar result mode.
3965 2018-03-21 Martin Liska <mliska@suse.cz>
3968 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3969 not intended return statement.
3971 2018-03-21 Martin Liska <mliska@suse.cz>
3974 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3975 (chkp_find_bound_slots_1): Limit number of iterations.
3977 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3980 * Minor grammar fixes for x86 options.
3982 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3985 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3986 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3988 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3991 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3992 (altivec_lvx_<mode>_1op): Likewise.
3993 (altivec_stvx_<mode>_2op): Likewise.
3994 (altivec_stvx_<mode>_1op): Likewise.
3995 (altivec_lvx_<VM2:mode>): New define_expand.
3996 (altivec_stvx_<VM2:mode>): Likewise.
3997 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3998 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3999 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4000 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4001 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4002 (rs6000_gen_lvx): Likewise.
4003 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4004 (altivec_expand_stv_builtin): Likewise.
4005 (altivec_expand_builtin): Likewise.
4006 * config/rs6000/vector.md: Likewise.
4008 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4011 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4014 2018-03-20 Richard Biener <rguenther@suse.de>
4017 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4018 sign-conversions as zero, fall back to standard scalar_stmt
4021 2018-03-20 Martin Liska <mliska@suse.cz>
4024 * predict.c (rebuild_frequencies): Handle case when we have
4025 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4027 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4030 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4031 flag_section_anchors.
4032 * varasm.c (use_blocks_for_decl_p): Remove hack for
4033 dw2_force_const_mem.
4036 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4038 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4039 be created, use lowpart_subreg of operands[0] rather than operands[0]
4041 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4042 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4043 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4044 and n constraint instead of aarch64_shift_imm_di and Usd.
4045 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4046 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4048 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4051 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4052 to favor GPR over NEON registers.
4053 (<shift>di3_neon): Likewise.
4055 2018-03-20 Tom de Vries <tom@codesourcery.com>
4058 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4059 (nvptx_process_pars): Emit bar.sync asap and alap.
4061 2018-03-20 Tom de Vries <tom@codesourcery.com>
4064 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4065 seen_label if seen_label is already set.
4067 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4070 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4071 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4072 Use 1U instead of 1. Formatting fixes.
4075 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4076 instead of TREE_TYPE (s1) for the return value.
4078 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4080 PR tree-optimization/84946
4081 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4082 bitsize + bitsize in poly_uint64 rather than poly_int64.
4085 * dwarf2asm.c: Include fold-const.c.
4086 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4087 of decl rather than decl itself.
4089 PR rtl-optimization/84643
4090 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4092 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4095 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4096 calling assemble_variable.
4098 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4101 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4102 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4104 2018-03-19 Jim Wilson <jimw@sifive.com>
4107 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4108 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4109 (riscv_first_stack_step): Likewise.
4110 (riscv_option_override): Use STACK_BOUNDARY instead of
4112 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4114 (BIGGEST_ALIGNMENT): Set to 128.
4115 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4116 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4119 2018-03-19 Richard Biener <rguenther@suse.de>
4121 PR tree-optimization/84933
4122 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4123 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4125 2018-03-19 Richard Biener <rguenther@suse.de>
4127 PR tree-optimization/84859
4128 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4129 (cond_if_else_store_replacement): Perform sinking operation on
4130 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4131 Generalize what a BB with a single eligible store is.
4133 2018-03-19 Richard Biener <rguenther@suse.de>
4135 PR tree-optimization/84929
4136 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4137 chrec_is_positive against non-chrec arg.
4139 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4142 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4144 2018-03-18 Martin Liska <mliska@suse.cz>
4146 PR rtl-optimization/84635
4147 * regrename.c (build_def_use): Use matches_mode only when
4150 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4152 PR tree-optimization/84913
4153 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4154 vectorize chains of COND_EXPRs.
4156 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4158 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4160 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4162 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4164 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4166 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4168 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4169 Kito Cheng <kito.cheng@gmail.com>
4171 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4172 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4173 (nds32_adjust_reg_alloc_order): New function.
4174 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4176 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4178 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4179 nds32_print_operand, nds32_print_operand_address): Use
4180 HOST_WIDE_INT_PRINT_DEC instead.
4182 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4184 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4186 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4189 * config/i386/i386.c (initial_ix86_tune_features,
4190 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4192 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4193 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4194 rather than 1u << ix86_tune. Formatting fix.
4195 (ix86_option_override_internal): Change ix86_arch_mask from
4196 unsigned int to unsigned HOST_WIDE_INT, initialize to
4197 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4198 (ix86_function_specific_restore): Likewise.
4200 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4203 * postreload.c (reload_combine_recognize_pattern): Perform
4204 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4205 truncate_int_for_mode the result for the destination's mode.
4208 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4209 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4211 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4214 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4217 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4218 regno_allocno_class_array and sorted_pseudos.
4219 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4220 insns where regno is used.
4222 2018-03-16 Martin Liska <mliska@suse.cz>
4225 * multiple_target.c (create_dispatcher_calls): Redirect
4226 reference in the symbol table.
4228 2018-03-16 Martin Liska <mliska@suse.cz>
4231 * multiple_target.c (create_dispatcher_calls): Redirect also
4234 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4238 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4239 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4240 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4242 2018-03-16 Julia Koval <julia.koval@intel.com>
4244 * doc/invoke.texi (Skylake Server): Add CLWB.
4245 Cannonlake): Remove CLWB.
4247 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4249 PR tree-optimization/84841
4250 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4252 (FLOAT_ONE_CONST_TYPE): Define.
4253 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4254 (sort_by_operand_rank): Put entries with higher constant_type last
4255 rather than first to match comments.
4257 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4259 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4262 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4265 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4266 check and use address of target always.
4268 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4271 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4272 (indirect_thunk_bnd_needed): Likewise.
4273 (indirect_thunks_used): Likewise.
4274 (indirect_thunks_bnd_used): Likewise.
4275 (indirect_return_needed): New.
4276 (indirect_return_bnd_needed): Likewise.
4277 (output_indirect_thunk_function): Add a bool argument for
4279 (output_indirect_thunk_function): Don't generate alias for
4280 function return thunk.
4281 (ix86_code_end): Call output_indirect_thunk_function to generate
4282 function return thunks.
4283 (ix86_output_function_return): Set indirect_return_bnd_needed
4284 and indirect_return_needed instead of indirect_thunk_bnd_needed
4285 and indirect_thunk_needed.
4287 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4289 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4290 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4291 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4293 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4294 Paul Hua <paul.hua.gm@gmail.com>
4297 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4299 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4301 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4302 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4305 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4308 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4309 instead of GET_MODE_SIZE when comparing Units.
4311 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4314 * varasm.c (hash_section): Return an unchangeble hash value
4315 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4316 Return !aarch64_can_use_per_function_literal_pools_p ().
4318 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4321 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4322 rather than address of cmode as last argument to prepare_cmp_insn.
4324 2018-03-15 Julia Koval <julia.koval@intel.com>
4326 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4327 F_AVX512VNNI, F_AVX512BITALG): New.
4329 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4332 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4333 insn for floating-point loads and stores.
4335 2018-03-14 Carl Love <cel@us.ibm.com>
4337 * config/rs6000/rs6000-c.c: Add macro definitions for
4338 ALTIVEC_BUILTIN_VEC_PERMXOR.
4339 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4340 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4341 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4342 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4344 * config/doc/extend.texi: Add prototypes for vec_permxor.
4346 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4349 * diagnostic-show-locus.c (class layout_point): Convert m_line
4350 from int to linenum_type.
4351 (line_span::comparator): Use linenum "compare" function when
4352 comparing line numbers.
4353 (test_line_span): New function.
4354 (layout_range::contains_point): Convert param "row" from int to
4356 (layout_range::intersects_line_p): Likewise.
4357 (layout::will_show_line_p): Likewise.
4358 (layout::print_source_line): Likewise.
4359 (layout::should_print_annotation_line_p): Likewise.
4360 (layout::print_annotation_line): Likewise.
4361 (layout::print_leading_fixits): Likewise.
4362 (layout::annotation_line_showed_range_p): Likewise.
4363 (struct line_corrections): Likewise for field m_row.
4364 (line_corrections::line_corrections): Likewise for param "row".
4365 (layout::print_trailing_fixits): Likewise.
4366 (layout::get_state_at_point): Likewise.
4367 (layout::get_x_bound_for_row): Likewise.
4368 (layout::print_line): Likewise.
4369 (diagnostic_show_locus): Likewise for locals "last_line" and
4371 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4372 * input.c (selftest::test_linenum_comparisons): New function.
4373 (selftest::input_c_tests): Call it.
4374 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4375 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4376 * selftest.h (ASSERT_GT): New macro.
4377 (ASSERT_GT_AT): New macro.
4378 (ASSERT_LT): New macro.
4379 (ASSERT_LT_AT): New macro.
4381 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4383 PR rtl-optimization/84780
4384 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4386 2018-03-14 Martin Liska <mliska@suse.cz>
4388 * tree.c (record_node_allocation_statistics): Use
4389 get_stats_node_kind.
4390 (get_stats_node_kind): New function extracted from
4391 record_node_allocation_statistics.
4392 (free_node): Use get_stats_node_kind.
4394 2018-03-14 Richard Biener <rguenther@suse.de>
4396 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4397 that the value-set of ANTIC_IN doesn't grow.
4400 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4402 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4403 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4405 2018-03-14 Julia Koval <julia.koval@intel.com>
4407 * config.gcc (icelake-client, icelake-server): New.
4409 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4410 (initial_ix86_arch_features): Ditto.
4411 (PTA_SKYLAKE): Add SGX.
4412 (PTA_ICELAKE): Remove.
4413 (PTA_ICELAKE_CLIENT): New.
4414 (PTA_ICELAKE_SERVER): New.
4415 (ix86_option_override_internal): Split up icelake on icelake client and
4417 (get_builtin_code_for_version): Ditto.
4418 (fold_builtin_cpu): Ditto.
4419 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4420 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4421 * config/i386/i386.h (processor_type): Ditto.
4422 * doc/invoke.texi: Ditto.
4424 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4427 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4428 INTEGER_CST offset, add it together with bitpos / 8 and
4429 sign extend based on POINTER_SIZE.
4433 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4436 * config/i386/constraints.md (Yc): New register constraint.
4437 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4438 Use Yc constraint for alternative 2 of operand 0. Remove
4439 preferred_for_speed attribute.
4441 2018-03-14 Richard Biener <rguenther@suse.de>
4443 PR tree-optimization/84830
4444 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4445 with the old one to avoid oscillations.
4447 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4450 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4452 (assign_by_spills): Return a flag of reload assignment failure.
4453 Do not process the reload assignment failures. Do not spill other
4454 reload pseudos if they has the same reg class. Update n if
4456 (lra_assign): Add a return arg. Set up from the result of
4457 assign_by_spills call.
4458 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4459 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4460 usage_insns if it is not NULL.
4461 (spill_hard_reg_in_range): New function.
4462 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4463 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4464 function prototypes.
4465 (lra_assign): Change prototype.
4466 * lra.c (lra): Add code to deal with fails by splitting hard reg
4469 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4471 * config/riscv/riscv.opt (mrelax): New option.
4472 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4473 "norelax" when riscv_mrelax is disabled.
4474 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4476 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4479 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4480 reassociation for int modes.
4482 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4484 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4485 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4487 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4488 * config/aarch64/aarch64-sve.md
4489 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4490 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4491 (*extend<mode><Vwide>2): Rename to...
4492 (aarch64_sve_extend<mode><Vwide>2): ...this.
4493 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4494 renaming the old pattern to...
4495 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4497 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4498 define_expand, renaming the old pattern to...
4499 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4500 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4501 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4502 account when deciding which SVE instruction the optab should use.
4503 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4505 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4507 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4508 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4509 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4510 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4511 (tlsdesc_small_<mode>): Turn a define_expand and use
4512 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4513 (tlsdesc_small_advsimd_<mode>): ...this.
4514 (tlsdesc_small_sve_<mode>): New pattern.
4516 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4518 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4519 (UNSPEC_UMUL_HIGHPART): New constants.
4520 (MUL_HIGHPART): New int iteraor.
4521 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4522 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4524 (*<su>mul<mode>3_highpart): New define_insn.
4526 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4529 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4532 2018-03-13 Martin Liska <mliska@suse.cz>
4535 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4537 (sem_item_optimizer::~sem_item_optimizer): Release it.
4538 (sem_item_optimizer::merge_classes): Register variable aliases.
4539 (sem_item_optimizer::fixup_pt_set): New function.
4540 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4541 * ipa-icf.h: Declare new variables and functions.
4543 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4546 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4547 integer_pow2p@2 and test integer_pow2p in condition.
4548 (A < 0 ? C : 0): Similarly for @1.
4551 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4552 characters starting at p contain '\0' character, don't look beyond
4556 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4557 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4560 * reg-stack.c (change_stack): Change update_end var from int to
4561 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4562 also call set_block_for_insn on the newly added insns and rescan.
4565 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4566 on the last operand.
4569 * tree.c (stabilize_reference_1): Return save_expr (e) for
4570 STATEMENT_LIST even if it doesn't have side-effects.
4572 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4574 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4576 2018-03-12 Renlin Li <renlin.li@arm.com>
4578 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4579 aarch64_output_scalar_simd_mov_immediate.
4581 2018-03-12 Martin Sebor <msebor@redhat.com>
4583 PR tree-optimization/83456
4584 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4585 for perfectly overlapping calls to memcpy.
4586 (gimple_fold_builtin_memory_chk): Same.
4587 (gimple_fold_builtin_strcpy): Handle no-warning.
4588 (gimple_fold_builtin_stxcpy_chk): Same.
4589 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4591 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4593 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4594 parameter. Use it for SFmode.
4595 (rs6000_function_arg_advance_1): Adjust.
4596 (rs6000_function_arg): Adjust.
4597 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4599 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4601 PR rtl-optimization/84169
4602 PR rtl-optimization/84780
4603 * combine.c (can_combine_p): Check for a 2-insn combination whether
4604 the destination register is used between the two insns, too.
4606 2018-03-12 Richard Biener <rguenther@suse.de>
4608 PR tree-optimization/84803
4609 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4610 for refs DR analysis didn't process.
4612 2018-03-12 Richard Biener <rguenther@suse.de>
4614 PR tree-optimization/84777
4615 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4616 force-vectorize loops ignore whether we are optimizing for size.
4618 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4620 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4621 (TARGET_MD_ASM_ADJUST): Define.
4623 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4624 Kito Cheng <kito.cheng@gmail.com>
4625 Chung-Ju Wu <jasonwucj@gmail.com>
4627 * config/nds32/nds32.c (nds32_compute_stack_frame,
4628 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4629 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4630 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4631 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4632 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4633 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4634 * config/nds32/nds32.md (prologue, epilogue): Use macro
4635 NDS32_V3PUSH_AVAILABLE_P to do checking.
4637 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4640 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4641 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4642 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4643 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4644 addition of most attributes on !orig_type_die or the attribute not
4645 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4647 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4648 Chung-Ju Wu <jasonwucj@gmail.com>
4650 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4652 * config/nds32/nds32.opt (mvh): New option.
4654 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4655 Chung-Ju Wu <jasonwucj@gmail.com>
4657 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4659 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4660 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4663 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4664 Chung-Ju Wu <jasonwucj@gmail.com>
4666 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4668 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4669 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4671 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4672 Kito Cheng <kito.cheng@gmail.com>
4673 Chung-Ju Wu <jasonwucj@gmail.com>
4675 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4676 UNSPEC_FFMISM and UNSPEC_FLMISM.
4677 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4678 for ffb, ffmism and flmism.
4679 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4680 (unspec_ffmism): Ditto.
4681 (unspec_flmism): Ditto.
4682 (nds32_expand_builtin_impl): Check if string extension is available.
4683 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4684 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4686 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4689 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4692 * lra-assigns.c (assign_by_spills): Return a flag of reload
4693 assignment failure. Do not process the reload assignment
4694 failures. Do not spill other reload pseudos if they has the same
4696 (lra_assign): Add a return arg. Set up from the result of
4697 assign_by_spills call.
4698 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4699 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4700 usage_insns if it is not NULL.
4701 (spill_hard_reg_in_range): New function.
4702 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4703 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4704 function prototypes.
4705 (lra_assign): Change prototype.
4706 * lra.c (lra): Add code to deal with fails by splitting hard reg
4709 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4712 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4714 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4717 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4718 (dw_val_node): Add val_symbolic_view.
4719 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4720 (symview_upper_bound): New.
4721 (new_line_info_table): Initialize symviews_since_reset.
4722 (dwarf2out_source_line): Count symviews_since_reset and set
4723 symview_upper_bound.
4724 (dw_val_equal_p): Handle symview.
4725 (add_AT_symview): New.
4726 (print_dw_val): Handle symview.
4727 (attr_checksum, attr_checksum_ordered): Likewise.
4728 (same_dw_val_p, size_of_die): Likewise.
4729 (value_format, output_die): Likewise.
4730 (add_high_low_attributes): Use add_AT_symview for entry_view.
4731 (dwarf2out_finish): Reset symview_upper_bound, clear
4734 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4737 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4738 Add strict argument and use it.
4739 (rs6000_split_multireg_move): Update for new strict argument.
4740 (mem_operand_gpr): Disallow all non-offsettable addresses.
4741 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4743 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4746 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4747 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4748 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4751 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4752 decl, use remap_type if we want to use the type.
4754 2018-03-09 Martin Sebor <msebor@redhat.com>
4756 PR tree-optimization/84526
4757 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4759 (builtin_access::generic_overlap): Be prepared to handle non-array
4762 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4764 PR rtl-optimization/84682
4765 * lra-constraints.c (process_address_1): Check is_address flag
4766 for address constraints.
4767 (process_alt_operands): Likewise.
4768 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4769 preprocess_constraints.
4770 * recog.h (preprocess_constraints): Add oploc parameter.
4772 * recog.c (preprocess_constraints): Test address_operand for
4773 CT_ADDRESS constraints.
4775 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4778 * lra-assigns.c (assign_by_spills): Return a flag of reload
4779 assignment failure. Do not process the reload assignment
4780 failures. Do not spill other reload pseudos if they has the same
4782 (lra_assign): Add a return arg. Set up from the result of
4783 assign_by_spills call.
4784 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4785 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4786 usage_insns if it is not NULL.
4787 (spill_hard_reg_in_range): New function.
4788 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4789 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4790 function prototypes.
4791 (lra_assign): Change prototype.
4792 * lra.c (lra): Add code to deal with fails by splitting hard reg
4795 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4798 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4799 Accept complain bool parameter. Only emit errors if it is true.
4800 (arm_parse_cpu_option_name): Likewise.
4801 (arm_target_thumb_only): Adjust callers of the above.
4802 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4803 prototype to take a default true bool parameter.
4804 (arm_parse_arch_option_name): Likewise.
4806 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4807 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4811 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4812 * configure: Regenerate.
4813 * configure.ac ("linker --version-script option"): New.
4814 ("linker soname option"): New.
4816 2018-03-09 Richard Biener <rguenther@suse.de>
4818 PR tree-optimization/84775
4819 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4820 immediate uses of predicate stmts and mark them modified.
4823 PR tree-optimization/84178
4824 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4826 (version_loop_for_if_conversion): Delay update_ssa call.
4827 (tree_if_conversion): Delay update_ssa until after predicate
4830 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4833 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4834 when the function accesses prior frames.
4836 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4839 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4840 gen_llsym, otherwise call maybe_gen_llsym.
4843 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4844 has ',' character inside of it.
4846 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4849 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4850 as clobbering CC_REGNUM.
4852 2018-03-08 Richard Biener <rguenther@suse.de>
4855 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4856 (follow_copies_to_constant): Do not follow SSA names registered
4859 2018-03-08 Richard Biener <rguenther@suse.de>
4861 PR tree-optimization/84178
4862 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4864 (version_loop_for_if_conversion): Delay update_ssa call.
4865 (tree_if_conversion): Delay update_ssa until after predicate
4868 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4870 PR tree-optimization/84178
4871 * tree-if-conv.c (release_bb_predicate): Remove the
4872 the assertion that the stmts have NULL use_ops.
4873 Discard the statements, asserting that they haven't
4874 yet been added to a BB.
4876 2018-03-08 Richard Biener <rguenther@suse.de>
4878 PR tree-optimization/84746
4879 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4880 (phi_translate): Pass in destination ANTIC_OUT set.
4881 (phi_translate_1): Likewise. For a simplified result lookup
4882 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4883 (phi_translate_set): Adjust.
4884 (do_pre_regular_insertion): Likewise.
4885 (do_pre_partial_partial_insertion): Likewise.
4887 2018-03-08 Martin Liska <mliska@suse.cz>
4889 PR gcov-profile/84735
4890 * doc/gcov.texi: Document usage of profile files.
4891 * gcov-io.h: Document changes in the format.
4893 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4897 * dwarf2out.c (struct dw_line_info_table): Update comments for
4899 (FORCE_RESET_NEXT_VIEW): New.
4900 (FORCE_RESETTING_VIEW_P): New.
4901 (RESETTING_VIEW_P): Check for -1 too.
4902 (ZERO_VIEW_P): Likewise.
4903 (new_line_info_table): Force-reset next view.
4904 (dwarf2out_begin_function): Likewise.
4905 (dwarf2out_source_line): Simplify zero_view_p initialization.
4906 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4907 view directly. Omit view when omitting .loc at line 0.
4909 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4911 PR tree-optimization/84740
4912 * tree-switch-conversion.c (process_switch): Call build_constructors
4913 only if info.phi_count is non-zero.
4915 PR tree-optimization/84739
4916 * tree-tailcall.c (find_tail_calls): Check call arguments against
4917 DECL_ARGUMENTS (current_function_decl) rather than
4918 DECL_ARGUMENTS (func) when checking for tail recursion.
4920 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4922 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4923 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4924 Volker Reichelt's entry and add entries for people that perform
4925 GCC fuzzy testing and report numerous bugs.
4927 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4930 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4931 readonly data in sdata, if that is disabled.
4932 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4933 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4934 -mreadonly-in-sdata option.
4936 2018-03-07 Martin Sebor <msebor@redhat.com>
4938 PR tree-optimization/84468
4939 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4940 basic block when looking for nul assignment.
4942 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4945 * except.h (output_function_exception_table): Adjust prototype.
4946 * except.c (output_function_exception_table): Remove FNNAME parameter
4947 and add SECTION parameter. Ouput one part of the table at a time.
4948 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4949 the first part of the exception table and emit unwind directives.
4950 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4951 (i386_pe_seh_cold_init): Likewise.
4952 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4953 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4954 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4955 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4956 * config/i386/winnt.c: Include except.h.
4957 (struct seh_frame_state): Add reg_offset, after_prologue and
4958 in_cold_section fields.
4959 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4960 (i386_pe_seh_cold_init): New function.
4961 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4962 to seh->in_cold_section.
4963 (seh_emit_push): Record the offset of the push.
4964 (seh_emit_save): Record the offet of the save.
4965 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4966 Test seh->after_prologue to disregard the epilogue.
4967 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4968 (i386_pe_end_cold_function): New function.
4970 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4973 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4974 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4977 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4979 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4980 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4983 * multiple_target.c: Include tree-inline.h and intl.h.
4984 (expand_target_clones): Diagnose and fail if node->definition and
4985 !tree_versionable_function_p (node->decl).
4987 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4989 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4991 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4992 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4993 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4995 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4998 * combine.c (try_combine): Use reg_or_subregno instead of handling
4999 just paradoxical SUBREGs and REGs.
5001 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5003 * config/arc/arc.c (arc_finalize_pic): Remove function.
5004 (arc_must_save_register): We use single base PIC register, remove
5005 checks to save/restore the PIC register.
5006 (arc_expand_prologue): Likewise.
5007 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5009 (arc_verify_short): Likewise.
5010 (arc_attr_type): Likewise.
5011 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5012 (walk_stores): Likewise.
5013 (arc_address_cost): Make it static.
5014 (arc_verify_short): Likewise.
5015 (branch_dest): Likewise.
5016 (arc_attr_type): Likewise.
5017 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5018 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5019 (arc_final_prescan_insn): Remove inserting the nops due to
5020 hardware hazards. It is done in reorg step.
5021 (insn_length_variant_t): Remove.
5022 (insn_length_parameters_t): Likewise.
5023 (arc_insn_length_parameters): Likewise.
5024 (arc_get_insn_variants): Likewise.
5025 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5027 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5030 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5033 PR tree-optimization/84687
5034 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5036 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5038 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5040 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5041 Rename to ppc_speculation_barrier.
5042 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5043 __builtin_ppc_speculation_barrier.
5045 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5048 * combine.c (combine_simplify_rtx): Don't try to simplify if
5049 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5052 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5054 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5055 to 32 bytes when compiling for POWER9.
5057 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5060 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5061 regparm >= 3 with no arg reg available also for calls with
5062 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5065 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5067 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5069 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5072 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5074 2018-03-05 Richard Biener <rguenther@suse.de>
5076 PR tree-optimization/84486
5077 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5078 When inserting a __builtin_assume_aligned call set the LHS
5079 SSA name alignment info accordingly.
5081 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5083 PR tree-optimization/84114
5084 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5085 Avoid reassociation of FLOAT_MODE addition.
5087 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5089 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5090 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5091 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5092 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5093 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5094 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5095 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5097 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5098 __builtin_ia32_wbinvd): New builtins.
5099 (SPECIAL_ARGS2): New.
5100 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5101 (SPECIAL_ARGS2): New.
5102 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5103 (ix86_valid_target_attribute_inner_p): Ditto.
5104 (ix86_init_mmx_sse_builtins): Add special_args2.
5105 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5106 TARGET_WBNOINVD_P): New.
5107 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5108 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5109 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5110 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5111 * config/i386/pconfigintrin.h: New file.
5112 * config/i386/wbnoinvdintrin.h: Ditto.
5113 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5114 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5116 2018-03-05 Richard Biener <rguenther@suse.de>
5118 PR tree-optimization/84670
5119 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5121 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5122 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5123 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5124 doesn't grow if all successors (recursively) were visited at least
5127 2018-03-05 Richard Biener <rguenther@suse.de>
5129 PR tree-optimization/84650
5130 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5131 if executed in the loop pipeline.
5133 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5135 * doc/configfiles.texi (Configuration Files): Move info about
5136 conditionalizing $target-protos.h to...
5137 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5138 differs from $target-protos.h.
5140 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5141 Chung-Ju Wu <jasonwucj@gmail.com>
5143 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5144 * config/nds32/nds32-multiple.md (setmemsi): Define.
5145 * config/nds32/nds32-memory-manipulation.c
5146 (nds32_gen_dup_4_byte_to_word_value): New.
5147 (emit_setmem_word_loop): New.
5148 (emit_setmem_byte_loop): New.
5149 (nds32_expand_setmem_loop): New.
5150 (nds32_expand_setmem_loop_v3m): New.
5151 (nds32_expand_setmem_unroll): New.
5152 (nds32_expand_setmem): New.
5154 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5155 Chung-Ju Wu <jasonwucj@gmail.com>
5157 * config/nds32/nds32-memory-manipulation.c
5158 (nds32_emit_load_store): New.
5159 (nds32_emit_post_inc_load_store): New.
5160 (nds32_emit_mem_move): New.
5161 (nds32_emit_mem_move_block): New.
5162 (nds32_expand_movmemsi_loop_unknown_size): New.
5163 (nds32_expand_movmemsi_loop_known_size): New.
5164 (nds32_expand_movmemsi_loop): New.
5165 (nds32_expand_movmemsi_unroll): New.
5166 (nds32_expand_movmemqi): Rename ...
5167 (nds32_expand_movmemsi): ... to this.
5168 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5169 (movmemsi): ... to this.
5170 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5171 (nds32_expand_movmemsi): ... to this.
5173 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5174 Monk Chiang <sh.chiang04@gmail.com>
5175 Chung-Ju Wu <jasonwucj@gmail.com>
5177 * config/nds32/nds32-protos.h
5178 (nds32_expand_load_multiple): New arguments.
5179 (nds32_expand_store_multiple): Ditto.
5180 (nds32_valid_multiple_load_store): Rename ...
5181 (nds32_valid_multiple_load_store_p): ... to this.
5182 * config/nds32/nds32-memory-manipulation.c
5183 (nds32_expand_load_multiple): Refine implementation.
5184 (nds32_expand_store_multiple): Ditto.
5185 * config/nds32/nds32-multiple.md
5186 (load_multiple): Update nds32_expand_load_multiple interface.
5187 (store_multiple): Update nds32_expand_store_multiple interface.
5188 * config/nds32/nds32-predicates.c
5189 (nds32_valid_multiple_load_store): Rename ...
5190 (nds32_valid_multiple_load_store_p): ... to this and refine
5192 * config/nds32/predicates.md
5193 (nds32_load_multiple_and_update_address_operation): New predicate.
5194 (nds32_store_multiple_and_update_address_operation): New predicate.
5196 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5197 Chung-Ju Wu <jasonwucj@gmail.com>
5199 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5200 (combo): New attribute.
5201 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5203 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5205 * config/nds32/nds32.opt: Change -mcmodel= default value.
5207 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5208 Monk Chiang <sh.chiang04@gmail.com>
5209 Chung-Ju Wu <jasonwucj@gmail.com>
5211 * config/nds32/constants.md (unspec_element): New enum.
5212 * config/nds32/constraints.md (Umw): New constraint.
5213 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5214 * config/nds32/nds32-intrinsic.md: Likewise.
5215 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5216 (nds32_valid_smw_lwm_base_p): New.
5217 (nds32_output_smw_single_word): New.
5218 (nds32_output_lmw_single_word): New.
5219 (nds32_expand_unaligned_load): New.
5220 (nds32_expand_unaligned_store): New.
5221 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5222 (nds32_output_smw_single_word): Declare.
5223 (nds32_output_lmw_single_word): Declare.
5224 (nds32_expand_unaligned_load): Declare.
5225 (nds32_expand_unaligned_store): Declare.
5226 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5227 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5228 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5229 NDS32_BUILTIN_UASTORE_DW.
5230 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5233 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5234 Kito Cheng <kito.cheng@gmail.com>
5235 Chung-Ju Wu <jasonwucj@gmail.com>
5237 * config/nds32/nds32-intrinsic.c
5238 (nds32_expand_builtin_null_ftype_reg): Delete.
5239 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5240 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5241 (nds32_read_argument): New.
5242 (nds32_legitimize_target): Ditto.
5243 (nds32_legitimize_argument): Ditto.
5244 (nds32_check_constant_argument): Ditto.
5245 (nds32_expand_unop_builtin): Ditto.
5246 (nds32_expand_unopimm_builtin): Ditto.
5247 (nds32_expand_binop_builtin): Ditto.
5248 (nds32_builtin_decl_impl): Ditto.
5249 (builtin_description): Ditto.
5250 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5251 (nds32_init_builtins_impl): Ditto.
5252 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5253 (nds32_builtin_decl): New.
5254 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5255 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5257 2018-03-02 Jeff Law <law@redhat.com>
5259 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5260 (redundant_insn, fill_simple_delay_slots): Likewise.
5261 (fill_slots_from_thread): Likewise.
5262 * resource.c (mark_referenced_resources): Likewise.
5263 (mark_set_resources, find_dead_or_set_registers): Likewise.
5265 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5267 * substring-locations.h (format_warning_va): Formatting fix for
5269 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5271 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5272 * substring-locations.c: Include intl.h.
5273 (format_warning_va): Turned into small wrapper around
5274 format_warning_n_va, renamed to ...
5275 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5276 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5278 (format_warning_at_substring_n): New function.
5279 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5280 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5281 format_warning_at_substring with just a shorter name instead of
5282 const function pointer.
5283 (fmtwarn_n): New function.
5284 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5285 appropriate, get rid of all the fmtstr temporaries, move conditionals
5286 with G_() wrapped string literals directly into fmtwarn arguments,
5287 cast dir.len to (int), formatting fixes.
5289 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5291 * doc/invoke.texi: Remove "Cilk Plus" references.
5293 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5294 Richard Biener <rguenther@suse.de>
5297 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5298 for error or warning attributes if CALL_FROM_THUNK_P is set.
5301 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5304 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5305 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5308 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5309 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5311 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5312 instead of -1U in last predictors element's probability member.
5314 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5317 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5318 arguments if they are comparable.
5320 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5322 PR tree-optimization/84634
5323 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5324 masks and masked_loop_p with a single loop_masks, making sure it's
5325 null for bb vectorization.
5327 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5329 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5330 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5331 loop->force_vectorize to check whether there is no alias.
5333 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5336 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5338 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5340 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5341 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5342 instead of a loop around prev_real_insn.
5343 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5347 * config/i386/i386.c (ix86_print_operand): Use conditional
5348 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5351 2018-03-02 Richard Biener <rguenther@suse.de>
5353 PR tree-optimization/84427
5354 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5355 (bitmap_set_subtract_values): Rewrite to handle multiple
5358 (prune_clobbered_mems): Likewise.
5359 (phi_translate): Take edge instead of pred/phiblock.
5360 (phi_translate_1): Likewise.
5361 (phi_translate_set): Likewise. Insert all translated
5362 exprs for a value into the set, keeping possibly multiple
5363 expressions per value.
5364 (compute_antic_aux): Adjust for phi_translate changes.
5365 When intersecting union the expressions and prune those
5366 not in the final value set, keeping possibly multiple
5367 expressions per value. Do not use value-insertion
5368 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5369 all expressions. Add verification that the value-sets
5370 only shrink during iteration.
5371 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5372 (do_pre_regular_insertion): Likewise.
5373 (do_pre_partial_partial_insertion): Likewise.
5375 2018-03-02 Richard Biener <rguenther@suse.de>
5378 * config/darwin.c (saved_debug_info_level): New static global.
5379 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5380 (darwin_asm_lto_end): Restore debug info generation settings.
5382 2018-03-01 Martin Liska <mliska@suse.cz>
5385 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5388 2018-03-01 Richard Biener <rguenther@suse.de>
5391 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5392 type annotation with LTO when debug was disabled at compile-time.
5394 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5396 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5398 (mips_final_postscan_insn): Likewise.
5400 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5402 PR rtl-optimization/84528
5403 * alias.c (init_alias_target): Add commentary.
5404 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5405 a unique base value if the frame pointer is not eliminated
5406 to the stack pointer.
5408 2018-03-01 Tom de Vries <tom@codesourcery.com>
5410 PR rtl-optimization/83327
5411 * lra-int.h (hard_regs_spilled_into): Declare.
5412 * lra.c (hard_regs_spilled_into): Define.
5413 (init_reg_info): Init hard_regs_spilled_into.
5414 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5415 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5416 (process_bb_lives): Handle hard_regs_spilled_into.
5417 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5418 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5420 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5422 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5423 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5424 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5425 * config/rs6000/aix72.h: New file.
5427 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5429 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5430 instead of warning_at with conditional singular and plural messages
5434 * stor-layout.c (update_alignment_for_field): For
5435 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5436 && !DECL_PACKED (field), do the alignment update, just use
5437 only desired_align instead of MAX (type_align, desired_align)
5439 (place_field): Don't do known_align < desired_align handling
5440 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5441 is non-NULL, instead do it after rli->prev_field handling and
5442 only if not within a bitfield word. For DECL_PACKED (field)
5443 use type_align of BITS_PER_UNIT.
5445 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5447 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5448 superfluous parentheses and trailing spaces.
5450 2018-02-28 Richard Biener <rguenther@suse.de>
5452 PR tree-optimization/84584
5453 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5454 SCoPs with fake exit edge.
5456 2018-02-28 Martin Liska <mliska@suse.cz>
5459 * timevar.c (timer::print): Fix format to properly print 100%
5462 2018-02-28 Richard Biener <rguenther@suse.de>
5465 * genmatch.c (capture_info::walk_match): Do not mark
5466 captured expressions without operands as expr_p given
5467 they act more like predicates and should be subject to
5468 "lost tail" side-effect preserving.
5470 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5472 PR rtl-optimization/81611
5473 * auto-inc-dec.c (attempt_change): Move dead note from
5474 mem_insn if it's the next use of regno
5475 (find_address): Take address use of reg holding
5476 non-incremented value. Add parm to limit search to the named
5478 (merge_in_block): Attempt to use a mem insn that is the next
5479 use of the original regno.
5481 2018-02-27 Martin Sebor <msebor@redhat.com>
5484 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5485 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5487 2018-02-27 Martin Sebor <msebor@redhat.com>
5489 PR translation/84207
5490 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5491 n argument to unsigned HOST_WIDE_INT.
5492 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5493 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5494 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5495 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5497 2018-02-27 Richard Biener <rguenther@suse.de>
5499 PR tree-optimization/84512
5500 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5501 Do not use the estimate returned from record_stmt_cost for
5502 the scalar iteration cost but sum properly using add_stmt_cost.
5504 2018-02-27 Richard Biener <rguenther@suse.de>
5506 PR tree-optimization/84466
5507 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5508 Adjust last change to less strictly validate use operands.
5510 2018-02-27 Martin Liska <mliska@suse.cz>
5512 PR gcov-profile/84548
5513 * gcov.c (process_file): Allow partial overlap and consider it
5514 also as group functions.
5515 (output_lines): Properly calculate range of lines for a group.
5517 2018-02-27 Martin Liska <mliska@suse.cz>
5519 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5520 'ggc' suffixes. Change first column width.
5521 (timer::print): Fix formatting of the column.
5523 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5525 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5526 preserve inline entry blocks for the sake of debug inline
5527 entry point markers alone.
5528 (remove_unused_locals): Suggest in comments a better place to
5529 force the preservation of inline entry blocks that are
5530 otherwise unused, but do not preserve them.
5532 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5534 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5536 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5539 * config/i386/constraints.md (Bs): Replace
5540 ix86_indirect_branch_register with
5541 TARGET_INDIRECT_BRANCH_REGISTER.
5543 * config/i386/i386.md (indirect_jump): Likewise.
5544 (tablejump): Likewise.
5545 (*sibcall_memory): Likewise.
5546 (*sibcall_value_memory): Likewise.
5547 Peepholes of indirect call and jump via memory: Likewise.
5548 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5549 (*sibcall_value_GOT_32): Likewise.
5550 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5551 (GOT_memory_operand): Likewise.
5552 (call_insn_operand): Likewise.
5553 (sibcall_insn_operand): Likewise.
5554 (GOT32_symbol_operand): Likewise.
5555 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5557 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5559 PR rtl-optimization/83496
5560 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5561 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5562 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5563 redundant insn, if any.
5564 (relax_delay_slots): Likewise.
5565 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5567 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5569 PR tree-optimization/83965
5570 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5571 that grouped statements are part of a reduction chain. Return
5572 true if the statement is not marked as a reduction itself but
5574 (vect_recog_dot_prod_pattern): Don't check whether the statement
5575 is part of a group here.
5576 (vect_recog_sad_pattern): Likewise.
5577 (vect_recog_widen_sum_pattern): Likewise.
5579 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5582 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5584 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5587 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5589 (ix86_output_indirect_function_return): New prototype.
5590 (ix86_split_simple_return_pop_internal): Likewise.
5591 * config/i386/i386.c (indirect_return_via_cx): New.
5592 (indirect_return_via_cx_bnd): Likewise.
5593 (indirect_thunk_name): Handle return va CX_REG.
5594 (output_indirect_thunk_function): Create alias for
5595 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5596 (ix86_output_indirect_jmp): Remove the bool argument.
5597 (ix86_output_indirect_function_return): New function.
5598 (ix86_split_simple_return_pop_internal): Likewise.
5599 * config/i386/i386.md (*indirect_jump): Don't pass false
5600 to ix86_output_indirect_jmp.
5601 (*tablejump_1): Likewise.
5602 (simple_return_pop_internal): Change it to define_insn_and_split.
5603 Call ix86_split_simple_return_pop_internal to split it for
5605 (simple_return_indirect_internal): Call
5606 ix86_output_indirect_function_return instead of
5607 ix86_output_indirect_jmp.
5609 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5612 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5613 memset and value initialization afterwards.
5615 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5617 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5619 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5622 * common/config/aarch64/aarch64-common.c
5623 (aarch_option_optimization_table[]): Switch
5624 off fomit-frame-pointer
5626 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5627 Chung-Ju Wu <jasonwucj@gmail.com>
5629 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5631 (store_multiple): Ditto.
5633 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5635 * config.gcc: Add --with-cpu support for nds32 target.
5636 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5637 * config/nds32/nds32.opt: Add -mcpu= option.
5639 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5641 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5642 isel=yes): Warn for these deprecated options.
5644 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5646 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5647 ISA_2_5_MASKS_EMBEDDED.
5649 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5651 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5652 p->max as pointers rather than using iterative_hash_expr.
5654 2018-02-23 Carl Love <cel@us.ibm.com>
5656 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5657 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5659 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5660 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5661 P8V_BUILTIN_VEC_VUNSIGNED2.
5663 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5666 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5667 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5669 * lra-constraints.c (curr_insn_transform): Set up
5670 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5672 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5673 * lra-eliminations.c (spill_pseudos): Ditto.
5674 (process_insn_for_elimination): Ditto.
5675 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5676 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5677 LRA_NON_CLOBBERED_ALT.
5679 2018-02-22 Martin Sebor <msebor@redhat.com>
5681 PR tree-optimization/84480
5682 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5683 to maybe_diag_stxncpy_trunc. Call it.
5684 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5685 from gimple_fold_builtin_strcpy. Print inlining stack.
5686 (handle_builtin_stxncpy): Print inlining stack.
5687 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5689 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5692 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5693 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5694 and -fcheck-pointer-bounds are used together.
5695 (indirect_thunk_prefix): New enum.
5696 (indirect_thunk_need_prefix): New function.
5697 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5698 "_nt" instead of "_bnd" for NOTRACK prefix.
5699 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5700 (output_indirect_thunk_function): Likewise.
5702 (ix86_code_end): Update output_indirect_thunk_function calls.
5703 (ix86_output_indirect_branch_via_reg): Replace
5704 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5705 (ix86_output_indirect_branch_via_push): Likewise.
5706 (ix86_output_function_return): Likewise.
5707 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5708 incompatible with -fcf-protection=branch and
5709 -fcheck-pointer-bounds.
5711 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5714 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5715 Change gcc_assert call to output_operand_lossage.
5717 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5719 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5721 2018-02-22 DJ Delorie <dj@redhat.com>
5722 Sebastian Perta <sebastian.perta@renesas.com>
5723 Oleg Endo <olegendo@gcc.gnu.org>
5725 * config/rx/rx.c (rx_rtx_costs): New function.
5726 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5728 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5730 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5732 2018-02-22 Martin Liska <mliska@suse.cz>
5735 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5736 Add "native" as a possible value.
5738 2018-02-22 Martin Liska <mliska@suse.cz>
5741 * config/i386/i386.c (ix86_option_override_internal):
5742 Add "native" as a possible value for -march and -mtune.
5744 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5747 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5748 to all type variants.
5750 PR tree-optimization/84503
5751 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5752 width as info->bitpos + info->bitsize - start.
5753 (merged_store_group::merge_overlapping): Simplify width computation.
5754 (check_no_overlap): New function.
5755 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5756 start + width and last_order of the group, fail if check_no_overlap
5758 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5759 to group if check_no_overlap fails.
5761 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5763 * config/rs6000/altivec.md: Delete contraint arguments to
5764 define_expand, define_split, and define_peephole2, and in
5765 define_insn_and_split if always unused.
5766 * config/rs6000/darwin.md: Ditto.
5767 * config/rs6000/dfp.md: Ditto.
5768 * config/rs6000/rs6000.md: Ditto.
5769 * config/rs6000/sync.md: Ditto.
5770 * config/rs6000/vector.md: Ditto.
5771 * config/rs6000/vsx.md: Ditto.
5773 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5775 * config/rs6000/altivec.md: Write output control strings as braced
5776 blocks instead of double-quoted strings.
5777 * config/rs6000/darwin.md: Ditto.
5778 * config/rs6000/rs6000.md: Ditto.
5779 * config/rs6000/vector.md: Ditto.
5780 * config/rs6000/vsx.md: Ditto.
5782 2018-02-21 Jason Merrill <jason@redhat.com>
5784 PR c++/84314 - ICE with templates and fastcall attribute.
5785 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5787 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5789 * ipa-cp.c (determine_versionability): Fix comment typos.
5791 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5794 * ipa-cp.c (determine_versionability): Do not version functions caling
5797 2018-02-21 Martin Liska <mliska@suse.cz>
5800 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5801 Add "native" as a possible value.
5802 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5803 the macro when native cpu detection is available.
5805 2018-02-21 Martin Liska <mliska@suse.cz>
5808 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5809 Add "native" as a possible value.
5810 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5811 when native cpu detection is available.
5813 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5814 Martin Sebor <msebor@redhat.com>
5816 PR tree-optimization/84478
5817 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5819 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5820 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5821 support which is conservatively correct, for 2 only stay conservative
5822 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5823 argument to the 2 argument get_range_strlen, adjust 6 arg
5824 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5826 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5827 (gimple_fold_builtin_strlen): Pass true as last argument to
5830 2018-02-20 Martin Sebor <msebor@redhat.com>
5833 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5834 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5835 (builtin_memref::builtin_memref): Factor out parts into
5836 set_base_and_offset and call it.
5838 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5841 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5842 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5843 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5844 search at the associated MODE_INT.
5846 2018-02-20 Jeff Law <law@redhat.com>
5849 PR tree-optimization/81592
5851 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5852 for range data rather than using global data.
5853 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5854 range data rather than using global data.
5855 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5856 pass it to children as needed.
5857 (struct directive::fmtresult): Similarly.
5858 (struct directive::set_width): Similarly.
5859 (struct directive::set_precision): Similarly.
5860 (format_integer, format_directive, parse_directive): Similarly.
5861 (format_none): Accept unnamed vr_values parameter.
5862 (format_percent, format_floating, format_character): Similarly.
5863 (format_string, format_plain): Similarly.
5864 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5865 the EVRP range analyzer for range data rather than using global data.
5866 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5867 gimple-ssa-evrp-analyze.h
5868 (class sprintf_dom_walker): Add after_dom_children member function.
5869 Add evrp_range_analyzer member.
5870 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5871 range analyzer as needed.
5872 (sprintf_dom_walker::after_dom_children): New member function.
5873 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5875 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5876 (evrp_range_analyzer::pop_to_marker): Likewise.
5878 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5880 PR tree-optimization/84419
5881 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5882 with the required type if its current type is compatible but
5885 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5888 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5889 after vectorization.
5891 2018-02-20 Martin Liska <mliska@suse.cz>
5894 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5895 possible values if we don't have a hint.
5897 2018-02-20 Martin Liska <mliska@suse.cz>
5901 * final.c (shorten_branches): Build align_tab array with one
5903 * opts.c (finish_options): Add alignment option limit check.
5904 (MAX_CODE_ALIGN): Likewise.
5905 (MAX_CODE_ALIGN_VALUE): Likewise.
5906 * doc/invoke.texi: Document maximum allowed option value for
5907 all -falign-* options.
5909 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5912 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5913 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5914 * var-tracking.c (emit_note_insn_var_location): Remove all references
5915 to NOTE_INSN_CALL_ARG_LOCATION.
5916 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5917 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5918 Use copy_rtx_if_shared.
5919 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5920 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5921 (dwarf2out_var_location): Remove handling of
5922 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5924 * final.c (final_scan_insn): Remove all references to
5925 NOTE_INSN_CALL_ARG_LOCATION.
5926 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5927 before dumping final insns.
5928 * except.c (emit_note_eh_region_end): Remove all references to
5929 NOTE_INSN_CALL_ARG_LOCATION.
5930 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5931 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5932 * config/arc/arc.c (hwloop_optimize): Likewise.
5933 * config/arm/arm.c (create_fix_barrier): Likewise.
5934 * config/s390/s390.c (s390_chunkify_start): Likewise.
5935 * config/sh/sh.c (find_barrier): Likewise.
5936 * config/i386/i386.c (rest_of_insert_endbranch,
5937 ix86_seh_fixup_eh_fallthru): Likewise.
5938 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5939 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5940 * config/frv/frv.c (frv_function_prologue): Likewise.
5941 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5943 (note_outside_basic_block_p): Remove all references to
5944 NOTE_INSN_CALL_ARG_LOCATION.
5945 * gengtype.c (adjust_field_rtx_def): Likewise.
5946 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5948 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5949 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5952 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5955 PR tree-optimization/84452
5956 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5957 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5960 2018-02-19 Martin Liska <mliska@suse.cz>
5963 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5965 2018-02-19 Martin Liska <mliska@suse.cz>
5966 Richard Sandiford <richard.sandiford@linaro.org>
5968 PR tree-optimization/82491
5969 * gimple-fold.c (get_base_constructor): Make earlier bail out
5972 2018-02-19 Carl Love <cel@us.ibm.com>
5974 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5975 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5977 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5978 P8V_BUILTIN_VEC_NEG.
5980 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5982 * config/rl78/rl78.md (movdf): New define expand.
5984 2018-02-19 Martin Liska <mliska@suse.cz>
5987 * doc/invoke.texi: Fix typo.
5988 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5990 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5992 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5993 handle rs6000_single_float and rs6000_double_float specially for
5996 2018-02-16 Jeff Law <law@redhat.com>
5998 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6001 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6004 * ipa-inline.c (inline_small_functions): Fix a typo.
6006 2018-02-16 Nathan Sidwell <nathan@acm.org>
6008 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6010 2018-02-16 Carl Love <cel@us.ibm.com>
6012 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6013 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6014 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6015 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6016 expansion to P8V_BUILTIN_VEC_FLOAT2.
6018 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6020 PR rtl-optimization/70023
6021 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6022 src_regno into account.
6024 2018-02-16 Carl Love <cel@us.ibm.com>
6026 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6027 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6028 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6029 * config/rs6000/rs6000.c: Remove case statements for
6030 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6031 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6032 and P9V_BUILTIN_VEC_VINSERT4B.
6033 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6034 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6035 * config/rs6000/vsx.md:
6036 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6039 2018-02-16 Carl Love <cel@us.ibm.com>
6041 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6043 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6045 * config/rs6000/rs6000-c.c: Add the definitions for
6046 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6047 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6048 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6049 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6050 definition for insert4b and define insn *insert3b_internal.
6051 * doc/extend.texi: Add documentation for vec_extract4b.
6053 2018-02-16 Nathan Sidwell <nathan@acm.org>
6055 * doc/extend.texi (Backwards Compatibility): Mention friend
6056 injection. Note for-scope is deprecated.
6057 * doc/invoke.texi (-ffriend-injection): Deprecate.
6059 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6061 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6062 that moved to I2, also allow destinations that are a paradoxical
6063 subreg (instead of a normal reg).
6065 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6068 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6071 2018-02-16 Richard Biener <rguenther@suse.de>
6073 PR tree-optimization/84037
6074 PR tree-optimization/84016
6076 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6077 Adjust vec_construct for the fact we need additional higher latency
6078 128bit inserts for AVX256 and AVX512 vector builds.
6079 (ix86_add_stmt_cost): Scale vector construction cost for
6082 2018-02-16 Richard Biener <rguenther@suse.de>
6084 PR tree-optimization/84417
6085 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6086 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6087 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6089 2018-02-16 Martin Liska <mliska@suse.cz>
6092 * internal-fn.def (ASAN_CHECK): Set proper flags.
6093 (ASAN_MARK): Likewise.
6095 2018-02-16 Julia Koval <julia.koval@intel.com>
6097 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6098 from PTA_CANNONLAKE.
6100 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6103 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6104 Use ++iter rather than iter++ for std::list iterators.
6105 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6106 defer deleting them until all nodes in the forest are processed. Do
6107 free even leaf nodes. Change to_process into auto_vec.
6110 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6111 * vec.h (vec_default_construct): Use memset instead of placement new
6112 if BROKEN_VALUE_INITIALIZATION is defined.
6113 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6114 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6117 PR rtl-optimization/83723
6118 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6119 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6120 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6122 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6124 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6126 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6128 PR rtl-optimization/81443
6129 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6130 from inner REGs to paradoxical SUBREGs.
6132 2018-02-16 Richard Biener <rguenther@suse.de>
6134 PR tree-optimization/84399
6135 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6136 For operands we can analyze at their definition make sure we can
6137 analyze them at each use as well.
6139 2018-02-16 Richard Biener <rguenther@suse.de>
6141 PR tree-optimization/84190
6142 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6143 volatile accesses if the decl isn't volatile.
6145 2018-02-15 Jason Merrill <jason@redhat.com>
6147 PR c++/84314 - ICE with templates and fastcall attribute.
6148 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6149 TYPE_CANONICAL on an existing type.
6151 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6153 PR tree-optimization/84383
6154 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6155 dstoff nor call operand_equal_p if dstbase is NULL.
6157 PR tree-optimization/84334
6158 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6159 also a CONSTANT_CLASS_P, punt.
6161 2018-02-14 Jim Wilson <jimw@sifive.com>
6163 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6164 first SMALL_OPERAND check. New local min_second_step. Move assert
6165 to where locals are set. Add TARGET_RVC support.
6166 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6168 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6170 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6172 2018-02-14 Martin Sebor <msebor@redhat.com>
6174 PR tree-optimization/83698
6175 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6176 arrays constrain the offset range to their bounds.
6177 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6178 (builtin_access::overlap): Avoid setting the size of overlap if it's
6180 (maybe_diag_overlap): Also consider arrays when deciding what values
6181 of offsets to include in diagnostics.
6183 2018-02-14 Martin Sebor <msebor@redhat.com>
6186 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6187 that correspond to the kind of a declaration.
6189 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6192 * config/pa/pa.md: Load address of PIC label using the linkage table
6193 if the label is nonlocal.
6195 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6197 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6198 warning message if user requests -maltivec=be.
6199 * doc/invoke.texi: Document deprecation of -maltivec=be.
6201 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6204 * config/rs6000/rs6000-c.c: Update definitions for
6205 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6206 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6208 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6211 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6212 add _get_ssp intrinsics. Remove argument from
6213 __builtin_ia32_rdssp[d|q].
6214 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6215 * config/i386/i386-builtin.def: Remove argument from
6216 __builtin_ia32_rdssp[d|q].
6217 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6218 ix86_expand_special_args_builtin for _rdssp[d|q].
6219 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6220 Clear register before usage.
6221 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6222 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6224 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6226 PR tree-optimization/84357
6227 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6228 operand 1 of an ARRAY_REF too.
6230 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6233 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6234 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6236 (set_of_reg): New struct.
6237 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6238 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6239 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6241 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6242 Split into bitclr, bitset, bitinvert patterns if appropriate.
6243 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6244 use rx_fuse_in_memory_bitop.
6245 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6246 to named insn, correct maximum insn length.
6248 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6251 * machmode.def: Define a complex mode for PARTIAL_INT.
6252 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6254 * doc/rtl.texi: Document CSPImode.
6255 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6257 (msp430_hard_regno_nregs_with_padding): Likewise.
6259 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6262 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6264 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6266 PR rtl-optimization/84169
6267 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6268 we generated a parallel as new i3 and we split that to new i2 and i3
6269 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6270 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6271 those to i2, not i1. Partially rewrite this scan code.
6273 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6276 * stor-layout.c (place_field): For variable length fields, adjust
6277 offset_align afterwards not just based on the field's alignment,
6278 but also on the size.
6281 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6282 of exps and logs in the use_exp2 case.
6284 2018-02-13 Jeff Law <law@redhat.com>
6286 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6289 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6292 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6296 * common.opt (gas-loc-support, gas-locview-support): New.
6297 (ginline-points, ginternal-reset-location-views): New.
6298 * doc/invoke.texi: Document them. Use @itemx where intended.
6299 (gvariable-location-views): Adjust.
6300 * target.def (reset_location_view): New.
6301 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6302 (TARGET_RESET_LOCATION_VIEW): New.
6303 * doc/tm.texi: Rebuilt.
6304 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6305 (dwarf2out_default_as_locview_support): New.
6306 (output_asm_line_debug_info): Use option variables.
6307 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6308 (output_loc_list): Likewise.
6309 (add_high_low_attributes): Check option variables.
6310 Don't output entry view attribute in strict mode.
6311 (gen_inlined_subroutine_die): Check option variables.
6312 (dwarf2out_inline_entry): Likewise.
6313 (init_sections_and_labels): Likewise.
6314 (dwarf2out_early_finish): Likewise.
6315 (maybe_reset_location_view): New, from...
6316 (dwarf2out_var_location): ... here. Call it.
6317 * debug.h (dwarf2out_default_as_loc_support): Declare.
6318 (dwarf2out_default_as_locview_support): Declare.
6319 * hooks.c (hook_int_rtx_insn_0): New.
6320 * hooks.h (hook_int_rtx_insn_0): Declare.
6321 * toplev.c (process_options): Take -gas-loc-support and
6322 -gas-locview-support from dwarf2out. Enable
6323 -gvariable-location-views by default only with locview
6324 assembler support. Enable -ginternal-reset-location-views by
6325 default only if the target defines the corresponding hook.
6326 Enable -ginline-points by default if location views are
6327 enabled; force it disabled if statement frontiers are
6329 * tree-inline.c (expand_call_inline): Check option variables.
6330 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6332 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6334 PR tree-optimization/84321
6335 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6336 handling. Also check whether the anti-range contains any values
6337 that satisfy the mask; switch to a VR_RANGE if not.
6339 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6342 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6344 2018-02-13 Martin Jambor <mjambor@suse.cz>
6347 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6348 of call statements, also set location of a load to a temporary.
6350 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6352 * config/rl78/rl78.c (add_vector_labels): New function.
6353 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6354 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6355 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6356 which checks that no arguments are passed.
6357 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6358 * doc/extend.texi: Documentation for the new attribute.
6360 2018-02-13 Andreas Schwab <schwab@suse.de>
6362 * config/riscv/linux.h (CPP_SPEC): Define.
6364 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6367 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6368 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6369 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6370 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6371 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6372 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6373 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6374 temporarily for AES and PCLMUL builtins.
6376 PR tree-optimization/84339
6377 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6378 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6382 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6383 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6384 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6386 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6388 * omp-simd-clone.h: New file.
6389 * omp-simd-clone.c: Include omp-simd-clone.h.
6390 (expand_simd_clones): No longer static.
6391 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6392 cgraph.h and omp-simd-clone.h.
6393 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6394 (vect_recog_widen_shift_pattern): Formatting fix.
6395 (vect_pattern_recog_1): Don't check optab for calls.
6398 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6399 operands[2] into a REG before using gen_lowpart on it.
6401 2018-02-12 Jeff Law <law@redhat.com>
6404 * config/sh/sh.c (find_barrier): Consider a sibling call
6407 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6408 successfully back substituting a reg.
6410 2018-02-12 Richard Biener <rguenther@suse.de>
6412 PR tree-optimization/84037
6413 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6414 parameter, move visited init to caller.
6415 (vect_slp_analyze_operations): Separate cost from validity
6416 check, initialize visited once for all instances.
6417 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6419 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6421 (vect_model_promotion_demotion_cost): Likewise.
6422 (vectorizable_bswap): Guard cost modeling with !slp_node
6423 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6425 (vectorizable_call): Likewise.
6426 (vectorizable_conversion): Likewise.
6427 (vectorizable_assignment): Likewise.
6428 (vectorizable_shift): Likewise.
6429 (vectorizable_operation): Likewise.
6430 (vectorizable_store): Likewise.
6431 (vectorizable_load): Likewise.
6432 (vectorizable_condition): Likewise.
6433 (vectorizable_comparison): Likewise.
6435 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6438 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6439 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6440 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6442 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6445 * params.def (inline-min-speedup): Increase from 8 to 15.
6446 (max-inline-insns-auto): Decrease from 40 to 30.
6447 * ipa-split.c (consider_split): Add some buffer for function to
6448 be considered inlining candidate.
6449 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6452 2018-02-12 Richard Biener <rguenther@suse.de>
6454 PR tree-optimization/84037
6455 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6456 matched stmts if we cannot swap the non-matched ones.
6458 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6460 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6461 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6462 _mm_maskz_scalef_round_ss): New intrinsics.
6463 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6464 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6465 __builtin_ia32_scalefss_round): Remove.
6466 (__builtin_ia32_scalefsd_mask_round,
6467 __builtin_ia32_scalefss_mask_round): New intrinsics.
6468 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6469 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6470 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6471 "<round_constraint>")): Changed to ...
6472 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6473 "<round_scalar_constraint>")): ... this.
6474 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6475 %0, %1, %2<round_op3>}"): Changed to ...
6476 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6477 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6478 %2<round_scalar_mask_op3>}"): ... this.
6479 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6481 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6483 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6484 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6485 (_mm_maskz_sqrt_round_ss): New intrinsics.
6486 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6487 (__builtin_ia32_sqrtsd_mask_round)
6488 (__builtin_ia32_sqrtss_mask_round): New builtins.
6489 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6490 (__builtin_ia32_sqrtss_round): Remove.
6491 (__builtin_ia32_sqrtsd_mask_round)
6492 (__builtin_ia32_sqrtss_mask_round): New builtins.
6493 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6494 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6495 ((match_operand:VF_128 1 "vector_operand"
6496 "xBm,<round_constraint>")): Changed to ...
6497 ((match_operand:VF_128 1 "vector_operand"
6498 "xBm,<round_scalar_constraint>")): ... this.
6499 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6500 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6501 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6502 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6503 %<iptr>1<round_scalar_mask_op3>}): ... this.
6504 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6505 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6507 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6510 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6511 Cast vec_cmpeq result to correct type.
6512 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6513 Cast vec_cmpgt result to correct type.
6515 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6517 * final.c (final_scan_insn_1): Renamed from...
6518 (final_scan_insn): ... this. New wrapper, to recover
6519 seen from the outermost call in recursive ones.
6520 * config/sparc/sparc.c (output_return): Drop seen from call.
6521 (output_sibcall): Likewise.
6522 * config/visium/visium.c (output_branch): Likewise.
6524 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6526 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6529 2018-02-10 Alan Modra <amodra@gmail.com>
6532 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6533 Specify LR as an input.
6535 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6538 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6539 remove_member_access_dummy_vars): New functions.
6540 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6541 lower_omp_1, execute_lower_omp): Use them.
6543 PR rtl-optimization/84308
6544 * shrink-wrap.c (spread_components): Release todo vector.
6546 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6548 PR rtl-optimization/57193
6549 * ira-color.c (struct allocno_color_data): Add member
6550 conflict_allocno_hard_prefs.
6551 (update_conflict_allocno_hard_prefs): New.
6552 (bucket_allocno_compare_func): Add a preference based on
6553 conflict_allocno_hard_prefs.
6554 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6555 (color_allocnos): Remove a dead code. Initiate
6556 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6558 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6561 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6562 constraint from =wa to wa. Avoid a subreg on the output operand,
6563 instead use a pseudo and subreg it in a move.
6564 (p9_xxbrd_<mode>): Changed to ...
6565 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6566 (p9_xxbrd_v2df): New expander.
6567 (p9_xxbrw_<mode>): Changed to ...
6568 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6569 (p9_xxbrw_v4sf): New expander.
6571 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6573 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6575 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6578 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6579 multiply in 32-bit mode.
6580 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6581 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6584 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6586 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6587 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6588 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6589 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6591 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6594 * dwarf2out.c (is_trivial_indirect_ref): New function.
6595 (dwarf2out_late_global_decl): Do not generate a location
6596 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6597 and that are not defined in the current unit.
6599 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6601 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6602 instead of a libcall for UNORDERED.
6604 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6607 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6608 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6610 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6613 * config/s390/s390.c (s390_set_current_function): Invoke
6614 s390_indirect_branch_settings also if fndecl didn't change.
6616 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6618 * config/rs6000/rs6000.md (blockage): Set length to zero.
6620 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6622 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6624 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6627 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6628 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6632 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6633 PARALLEL incoming that failed vt_get_decl_and_offset check.
6636 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6637 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6639 (get_variable_section): For decls in named .bss* sections pass true as
6640 second argument to bss_initializer_p.
6642 2018-02-09 Marek Polacek <polacek@redhat.com>
6643 Jakub Jelinek <jakub@redhat.com>
6646 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6647 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6648 Sync some changes from cxx_fold_indirect_ref.
6650 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6652 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6654 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6655 (BLOCK_INLINE_ENTRY_LABEL): New.
6656 (dwarf2out_var_location): Disregard inline entry markers.
6657 (inline_entry_data): New struct.
6658 (inline_entry_data_hasher): New hashtable type.
6659 (inline_entry_data_hasher::hash): New.
6660 (inline_entry_data_hasher::equal): New.
6661 (inline_entry_data_table): New variable.
6662 (add_high_low_attributes): Add DW_AT_entry_pc and
6663 DW_AT_GNU_entry_view attributes if a pending entry is found
6664 in inline_entry_data_table. Add old entry_pc attribute only
6665 if debug nonbinding markers are disabled.
6666 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6667 markers are enabled.
6668 (block_within_block_p, dwarf2out_inline_entry): New.
6669 (dwarf2out_finish): Check that no entries remained in
6670 inline_entry_data_table.
6671 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6672 (final_scan_insn, notice_source_line): Likewise.
6673 (rest_of_clean_state): Skip inline entry markers.
6674 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6676 * gimple.c (gimple_build_debug_inline_entry): New.
6677 * gimple.h (enum gimple_debug_subcode): Add
6678 GIMPLE_DEBUG_INLINE_ENTRY.
6679 (gimple_build_debug_inline_entry): Declare.
6680 (gimple_debug_inline_entry_p): New.
6681 (gimple_debug_nonbind_marker_p): Adjust.
6682 * insn-notes.def (INLINE_ENTRY): New.
6683 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6684 inline entry marker notes.
6685 (print_insn): Likewise.
6686 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6687 (INSN_DEBUG_MARKER_KIND): Likewise.
6688 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6689 * tree-inline.c (expand_call_inline): Build and insert
6690 debug_inline_entry stmt.
6691 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6692 inline entry blocks early, if nonbind markers are enabled.
6693 (dump_scope_block): Dump fragment info.
6694 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6695 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6696 (gimple_build_debug_inline_entry): New.
6697 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6698 Enable/disable inline entry points too.
6699 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6700 (DEBUG_INSN): Describe inline entry markers.
6702 * common.opt (gvariable-location-views): New.
6703 (gvariable-location-views=incompat5): New.
6704 * config.in: Rebuilt.
6705 * configure: Rebuilt.
6706 * configure.ac: Test assembler for view support.
6707 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6708 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6709 * dwarf2out.c (var_loc_view): New typedef.
6710 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6711 (dwarf2out_locviews_in_attribute): New.
6712 (dwarf2out_locviews_in_loclist): New.
6713 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6714 (enum dw_line_info_opcode): Add LI_adv_address.
6715 (struct dw_line_info_table): Add view.
6716 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6717 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6718 (zero_view_p): New variable.
6719 (ZERO_VIEW_P): New macro.
6720 (output_asm_line_debug_info): New.
6721 (struct var_loc_node): Add view.
6722 (add_AT_view_list, AT_loc_list): New.
6723 (add_var_loc_to_decl): Add view param. Test it against last.
6724 (new_loc_list): Add view params. Record them.
6725 (AT_loc_list_ptr): Handle loc and view lists.
6726 (view_list_to_loc_list_val_node): New.
6727 (print_dw_val): Handle dw_val_class_view_list.
6728 (size_of_die): Likewise.
6729 (value_format): Likewise.
6730 (loc_list_has_views): New.
6731 (gen_llsym): Set vl_symbol too.
6732 (maybe_gen_llsym, skip_loc_list_entry): New.
6733 (dwarf2out_maybe_output_loclist_view_pair): New.
6734 (output_loc_list): Output view list or entries too.
6735 (output_view_list_offset): New.
6736 (output_die): Handle dw_val_class_view_list.
6737 (output_dwarf_version): New.
6738 (output_compilation_unit_header): Use it.
6739 (output_skeleton_debug_sections): Likewise.
6740 (output_rnglists, output_line_info): Likewise.
6741 (output_pubnames, output_aranges): Update version comments.
6742 (output_one_line_info_table): Output view numbers in asm comments.
6743 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6744 Call maybe_gen_llsym.
6745 (loc_list_from_tree_1): Adjust.
6746 (add_AT_location_description): Create view list attribute if
6747 needed, check it's absent otherwise.
6748 (convert_cfa_to_fb_loc_list): Adjust.
6749 (maybe_emit_file): Call output_asm_line_debug_info for test.
6750 (dwarf2out_var_location): Reset views as needed. Precompute
6751 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6752 attribute. Set view.
6753 (new_line_info_table): Reset next view.
6754 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6755 (dwarf2out_source_line): Likewise. Output view resets and labels to
6756 the assembler, or select appropriate line info opcodes.
6757 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6758 (optimize_string_length): Catch it. Adjust.
6759 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6760 dw_val_class_view_list, and remove it if no longer needed.
6761 (hash_loc_list): Hash view numbers.
6762 (loc_list_hasher::equal): Compare them.
6763 (optimize_location_lists): Check whether a view list symbol is
6764 needed, and whether the locview attribute is present, and
6765 whether they match. Remove the locview attribute if no longer
6767 (index_location_lists): Call skip_loc_list_entry for test.
6768 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6769 Use output_dwarf_version.
6770 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6771 (struct dw_val_node): Add val_view_list.
6772 * final.c (SEEN_NEXT_VIEW): New.
6773 (set_next_view_needed): New.
6774 (clear_next_view_needed): New.
6775 (maybe_output_next_view): New.
6776 (final_start_function): Rename to...
6777 (final_start_function_1): ... this. Take pointer to FIRST,
6778 add SEEN parameter. Emit param bindings in the initial view.
6779 (final_start_function): Reintroduce SEEN-less interface.
6780 (final): Rename to...
6781 (final_1): ... this. Take SEEN parameter. Output final pending
6782 next view at the end.
6783 (final): Reintroduce seen-less interface.
6784 (final_scan_insn): Output pending next view before switching
6785 sections or ending a block. Mark the next view as needed when
6786 outputting variable locations. Notify debug backend of section
6787 changes, and of location view changes.
6788 (rest_of_handle_final): Adjust.
6789 * toplev.c (process_options): Autodetect value for debug variable
6790 location views option. Warn on incompat5 without -gdwarf-5.
6791 * doc/invoke.texi (gvariable-location-views): New.
6792 (gvariable-location-views=incompat5): New.
6793 (gno-variable-location-views): New.
6795 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6797 PR tree-optimization/84136
6798 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6799 that the result of find_edge is non-NULL.
6801 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6804 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6805 storing integer register in SImode. Fix cost of 256 and 512
6806 byte aligned SSE register store.
6808 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6810 * config/i386/i386.c (ix86_multiplication_cost): Fix
6811 multiplication cost for TARGET_AVX512DQ.
6813 2018-02-08 Marek Polacek <polacek@redhat.com>
6815 PR tree-optimization/84238
6816 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6819 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6821 PR tree-optimization/84265
6822 * tree-vect-stmts.c (vectorizable_store): Don't treat
6823 VMAT_CONTIGUOUS accesses as grouped.
6824 (vectorizable_load): Likewise.
6826 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6828 PR tree-optimization/81635
6829 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6830 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6831 (test_round_for_mask): New functions.
6832 (wide_int_cc_tests): Call test_round_for_mask.
6833 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6834 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6835 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6836 range returned by get_range_info.
6838 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6841 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6842 * symtab.c: Include builtins.h
6843 (symtab_node::output_to_lto_symbol_table_p): Move here
6844 from lto-streamer-out.c:output_symbol_p.
6845 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6846 (output_symbol_p): Move all logic to symtab.c
6847 (produce_symtab): Update.
6849 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6851 * config/s390/s390-opts.h (enum indirect_branch): Define.
6852 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6853 (s390_indirect_branch_via_thunk)
6854 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6855 (enum s390_indirect_branch_type): Define.
6856 * config/s390/s390.c (struct s390_frame_layout, struct
6857 machine_function): Remove.
6858 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6859 (indirect_branch_table_label_no, indirect_branch_table_name):
6861 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6862 (enum s390_indirect_branch_option): Define.
6863 (s390_return_addr_from_memory): New function.
6864 (s390_handle_string_attribute): New function.
6865 (s390_attribute_table): Add new attribute handler.
6866 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6867 (s390_indirect_branch_via_thunk): New function.
6868 (s390_indirect_branch_via_inline_thunk): New function.
6869 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6870 sibling call optimization for non z10 compiles.
6871 (s390_emit_call): Force indirect branch target to be a single
6872 register. Add r1 clobber for non-z10 compiles.
6873 (s390_emit_epilogue): Emit return jump via return_use expander.
6874 (s390_reorg): Handle JUMP_INSNs as execute targets.
6875 (s390_option_override_internal): Perform validity checks for the
6876 new command line options.
6877 (s390_indirect_branch_attrvalue): New function.
6878 (s390_indirect_branch_settings): New function.
6879 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6880 (s390_output_indirect_thunk_function): New function.
6881 (s390_code_end): Implement target hook.
6882 (s390_case_values_threshold): Implement target hook.
6883 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6885 * config/s390/s390.h (struct s390_frame_layout)
6886 (struct machine_function): Move here from s390.c.
6887 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6888 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6889 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6890 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6891 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6892 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6893 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6894 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6895 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6896 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6897 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6898 (mnemonic attribute): Add values which aren't recognized
6900 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6901 pattern for branch conversion. Fix mnemonic attribute.
6902 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6903 indirect branch via thunk if requested.
6904 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6905 ("*indirect_jump"): Disable for branch conversion using out of
6907 ("indirect_jump_via_thunk<mode>_z10")
6908 ("indirect_jump_via_thunk<mode>")
6909 ("indirect_jump_via_inlinethunk<mode>_z10")
6910 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6911 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6912 ("casesi_jump_via_inlinethunk<mode>_z10")
6913 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6914 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6915 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6916 ("*indirect2_jump"): Disable for branch conversion.
6917 ("casesi_jump"): Turn into expander and expand patterns for branch
6919 ("return_use"): New expander.
6920 ("*return"): Emit return via thunk and rename it to ...
6921 ("*return<mode>"): ... this one.
6922 * config/s390/s390.opt: Add new options and and enum for the
6925 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6927 * lra-constraints.c (match_reload): Unconditionally use
6928 gen_lowpart_SUBREG, rather than selecting between that
6929 and equivalent gen_rtx_SUBREG code.
6931 2018-02-08 Richard Biener <rguenther@suse.de>
6933 PR tree-optimization/84233
6934 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6935 changed flag instead of boguously re-using phi_inserted.
6937 2018-02-08 Martin Jambor <mjambor@suse.cz>
6939 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6940 static local variables.
6942 2018-02-08 Richard Biener <rguenther@suse.de>
6944 PR tree-optimization/84278
6945 * tree-vect-stmts.c (vectorizable_store): When looking for
6946 smaller vector types to perform grouped strided loads/stores
6947 make sure the mode is supported by the target.
6948 (vectorizable_load): Likewise.
6950 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6952 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6953 Increase LDP/STP opportunities by adding adjacent callee-saves.
6955 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6957 PR rtl-optimization/84068
6958 PR rtl-optimization/83459
6959 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6961 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6963 PR tree-optimization/84224
6964 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6965 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6968 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6971 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6972 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6973 count, remove one USE.
6975 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6977 * doc/install.texi (Configuration): Document the
6978 --with-long-double-format={ibm,ieee} PowerPC configuration
6982 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6983 Convert from define_expand to be define_insn_and_split. Rework
6984 float/double/_Float128 conversions to QI/HI/SImode to work with
6985 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6986 conversions to QI/HImode types did a store and then a load to
6987 truncate the value. For conversions to VSX registers, don't split
6988 the insn, instead emit the code directly. Use the code iterator
6989 any_fix to combine signed and unsigned conversions.
6990 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6991 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6992 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6993 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6994 (fix_<mode>di2_hw): Likewise.
6995 (fixuns_<mode>di2_hw): Likewise.
6996 (fix_<mode>si2_hw): Likewise.
6997 (fixuns_<mode>si2_hw): Likewise.
6998 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6999 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7000 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7001 fix<uns>_trunc<SFDF:mode>si2_p8.
7002 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7004 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7005 (fix<uns>_<mode>_mem): Likewise.
7006 (fctiw<u>z_<mode>_mem): Likewise.
7007 (fix<uns>_<mode>_mem): Likewise.
7008 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7009 the register allocator from doing a direct move to the GPRs to do
7010 a store, and instead use the ISA 3.0 store byte/half-word from
7011 vector register instruction. For IEEE 128-bit floating point,
7012 also optimize stores of 32-bit ints.
7013 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7015 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7017 * genextract.c (push_pathstr_operand): New function to support
7019 (walk_rtx): Call push_pathstr_operand.
7020 (print_path): Support [a-zA-Z].
7022 2018-02-07 Richard Biener <rguenther@suse.de>
7024 PR tree-optimization/84037
7025 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7026 (cse_and_gimplify_to_preheader): Declare.
7027 (vect_get_place_in_interleaving_chain): Likewise.
7028 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7030 (_loop_vec_info::~_loop_vec_info): Delete it.
7031 (cse_and_gimplify_to_preheader): New function.
7032 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7033 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7034 (vectorizable_load): Likewise. For grouped stores always base
7035 the IV on the first element.
7036 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7037 condition before gimplifying.
7039 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7041 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7042 *DIV_EXPR and *MOD_EXPR.
7044 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7047 * config/i386/i386.c (ix86_option_override_internal): Mask out
7048 the CF_SET bit when checking -fcf-protection.
7050 2018-02-07 Tom de Vries <tom@codesourcery.com>
7053 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7056 2018-02-07 Richard Biener <rguenther@suse.de>
7058 PR tree-optimization/84204
7059 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7062 PR tree-optimization/84205
7063 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7064 special-case isl_ast_op_zdiv_r.
7066 PR tree-optimization/84223
7067 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7068 Only add conditions from within the region.
7069 (gather_bbs::after_dom_children): Adjust.
7071 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7074 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7075 * config/avr/avr.md: Only post-reload split REG-REG moves if
7076 either register is GENERAL_REG_P.
7078 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7080 PR tree-optimization/84235
7081 * tree-ssa-scopedtables.c
7082 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7083 if the subtraction is performed in floating point type where NaNs are
7084 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7085 build 1. Formatting fix.
7087 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7090 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7091 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7092 and skip it regardless of bb boundaries. Use CALL_P macro,
7093 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7094 unnecessarily, formatting fix.
7096 2018-02-06 Michael Collison <michael.collison@arm.com>
7098 * config/arm/thumb2.md:
7099 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7100 (*thumb_mov_notscc): Ditto.
7102 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7105 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7106 unsigned_fix, not "s".
7108 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7110 * configure.ac (gcc_fn_eh_frame_ro): New function.
7111 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7112 correct .eh_frame permissions.
7113 * configure: Regenerate.
7115 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7117 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7120 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7122 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7123 Display warning message for -mno-speculate-indirect-jumps.
7125 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7127 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7129 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7131 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7133 PR tree-optimization/84225
7134 * tree-eh.c (find_trapping_overflow): Only call
7135 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7137 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7140 * config/i386/i386.c: Reimplement the check of possible options
7141 -mibt/-mshstk conbination. Change error messages.
7142 * doc/invoke.texi: Fix a typo: remove extra '='.
7144 2018-02-06 Marek Polacek <polacek@redhat.com>
7146 PR tree-optimization/84228
7147 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7149 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7152 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7153 emitted arch directives.
7154 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7155 __ARM_FEATURE_COPROC before changing architectures.
7157 2018-02-06 Richard Biener <rguenther@suse.de>
7159 * config/i386/i386.c (print_reg): Fix typo.
7160 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7162 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7164 * configure: Regenerate.
7166 2018-02-05 Martin Sebor <msebor@redhat.com>
7168 PR tree-optimization/83369
7169 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7172 2018-02-05 Martin Liska <mliska@suse.cz>
7174 * doc/invoke.texi: Cherry-pick upstream r323995.
7176 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7178 * ira.c (ira_init_register_move_cost): Adjust comment.
7180 2018-02-05 Martin Liska <mliska@suse.cz>
7182 PR gcov-profile/84137
7183 * doc/gcov.texi: Fix typo in documentation.
7185 2018-02-05 Martin Liska <mliska@suse.cz>
7187 PR gcov-profile/83879
7188 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7189 using dlopen functionality.
7191 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7193 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7194 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7195 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7196 _mm_maskz_range_round_ss): New intrinsics.
7197 (__builtin_ia32_rangesd128_round)
7198 (__builtin_ia32_rangess128_round): Remove.
7199 (__builtin_ia32_rangesd128_mask_round,
7200 __builtin_ia32_rangess128_mask_round): New builtins.
7201 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7202 __builtin_ia32_rangess128_round): Remove.
7203 (__builtin_ia32_rangesd128_mask_round,
7204 __builtin_ia32_rangess128_mask_round): New builtins.
7205 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7206 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7207 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7208 "<round_saeonly_constraint>")): Changed to ...
7209 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7210 "<round_saeonly_scalar_constraint>")): ... this.
7211 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7212 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7213 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7214 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7215 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7217 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7219 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7221 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7222 Remove all values except native, 8540 and 8548.
7224 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7226 * config/i386/i386.c (ix86_output_function_return): Pass
7227 INVALID_REGNUM, instead of -1, as invalid register number to
7228 indirect_thunk_name and output_indirect_thunk.
7230 2018-02-02 Julia Koval <julia.koval@intel.com>
7232 * config.gcc: Add -march=icelake.
7233 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7234 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7235 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7236 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7237 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7238 (processor_target_table): Add icelake.
7239 (ix86_option_override_internal): Handle new PTAs.
7240 (get_builtin_code_for_version): Handle icelake.
7241 (M_INTEL_COREI7_ICELAKE): New.
7242 (fold_builtin_cpu): Handle icelake.
7243 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7244 * doc/invoke.texi: Add -march=icelake.
7246 2018-02-02 Julia Koval <julia.koval@intel.com>
7248 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7249 to wide_int_bitmask.
7250 * wide-int-bitmask.h: New.
7252 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7255 * config/i386/i386.md: Replace Pmode with word_mode in
7256 builtin_setjmp_setup and builtin_longjmp to support x32.
7258 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7262 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7264 (rs6000_supported_cpu_names): New static variable.
7265 (linux_cpu_translation_table): Likewise.
7266 (elf_platform) <cpu>: Define new static variable and use it.
7267 Translate kernel AT_PLATFORM name to canonical name if needed.
7268 Error if platform name is unknown.
7270 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7273 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7275 2018-02-01 Jeff Law <law@redhat.com>
7278 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7279 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7280 the scratch if RELEASE_VIA_POP is false.
7281 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7282 If we have to save a temporary register, decrement SIZE appropriately.
7283 Pass new arguments to release_scratch_register_on_entry.
7284 (ix86_adjust_stack_and_probe): Likewise.
7285 (ix86_emit_probe_stack_range): Pass new arguments to
7286 release_scratch_register_on_entry.
7288 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7290 PR rtl-optimization/84157
7291 * combine.c (change_zero_ext): Use REG_P predicate in
7292 front of HARD_REGISTER_P predicate.
7294 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7296 * config/avr/avr.c (avr_option_override): Move disabling of
7297 -fdelete-null-pointer-checks to...
7298 * common/config/avr/avr-common.c (avr_option_optimization_table):
7301 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7303 PR tree-optimization/81635
7304 * tree-data-ref.c (split_constant_offset_1): For types that
7305 wrap on overflow, try to use range info to prove that wrapping
7308 2018-02-01 Renlin Li <renlin.li@arm.com>
7311 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7313 (aarch64_register_move_cost): Likewise.
7314 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7316 (REG_CLASS_NAMES): Likewise.
7317 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7318 TAILCALL_ADDR_REGS. Remove IP registers.
7319 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7321 2018-02-01 Richard Biener <rguenther@suse.de>
7323 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7324 for specifying RPO order and allow NULL for that.
7325 * domwalk.c (dom_walker::dom_walker): Likewise.
7326 (dom_walker::walk): Handle NULL RPO order.
7327 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7329 (rewrite_update_dom_walker): Likewise.
7330 (mark_def_dom_walker): Likewise.
7332 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7334 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7335 (aarch64_maybe_expand_sve_subreg_move): Declare.
7336 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7337 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7339 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7340 that are semantically a reverse operation.
7341 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7342 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7343 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7345 (aarch64_can_change_mode_class): For big-endian, forbid changes
7346 between two SVE modes if they have different element sizes.
7348 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7350 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7351 the TImode handling for big-endian targets.
7353 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7355 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7356 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7358 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7359 Remove BSWAP handing for big-endian targets and use the form of
7360 LD1RQ appropariate for the mode.
7362 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7364 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7365 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7368 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7371 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7372 check for operands that need to go through aarch64_sve_reload_be.
7374 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7376 PR tree-optimization/81661
7377 PR tree-optimization/84117
7378 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7379 * tree-eh.c: Include gimplify.h.
7380 (find_trapping_overflow, replace_trapping_overflow,
7381 rewrite_to_non_trapping_overflow): New functions.
7382 * tree-vect-loop.c: Include tree-eh.h.
7383 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7384 * tree-data-ref.c: Include tree-eh.h.
7385 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7387 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7389 PR rtl-optimization/84123
7390 * combine.c (change_zero_ext): Check if hard register satisfies
7391 can_change_dest_mode before calling gen_lowpart_SUBREG.
7393 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7396 * ira.c (ira_init_register_move_cost): Remove assert.
7398 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7400 PR rtl-optimization/84071
7401 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7402 * doc/tm.texi: Regenerate.
7404 2018-01-31 Richard Biener <rguenther@suse.de>
7406 PR tree-optimization/84132
7407 * tree-data-ref.c (analyze_miv_subscript): Properly
7408 check whether evolution_function_is_affine_multivariate_p
7409 before calling gcd_of_steps_may_divide_p.
7411 2018-01-31 Julia Koval <julia.koval@intel.com>
7414 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7415 * config/i386/i386.md (rdpid_rex64) New.
7416 (rdpid): Make 32bit only.
7418 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7421 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7422 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7424 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7427 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7429 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7431 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7433 PR rtl-optimization/84071
7434 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7435 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7437 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7439 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7440 (arc_attribute_table): Add 'aux' attribute.
7441 (arc_in_small_data_p): Consider aux like variables.
7442 (arc_is_aux_reg_p): New function.
7443 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7444 (arc_get_aux_arg): New function.
7445 (prepare_move_operands): Handle aux-register access.
7446 (arc_handle_aux_attribute): New function.
7447 * doc/extend.texi (ARC Variable attributes): Add subsection.
7449 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7451 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7452 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7453 (arc_attribute_table): Add 'uncached' attribute.
7454 (arc_print_operand): Print '.di' flag for uncached memory
7456 (arc_in_small_data_p): Do not consider for small data the uncached
7458 (arc_is_uncached_mem_p): New function.
7459 * config/arc/predicates.md (compact_store_memory_operand): Check
7460 for uncached memory accesses.
7461 (nonvol_nonimm_operand): Likewise.
7462 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7464 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7467 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7468 falign-loops=): Add Optimization flag.
7470 2018-01-30 Jeff Law <law@redhat.com>
7473 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7474 INT_REGISTERS_SAVED. Check it prior to calling
7475 get_scratch_register_on_entry.
7476 (ix86_adjust_stack_and_probe): Similarly.
7477 (ix86_emit_probe_stack_range): Similarly.
7478 (ix86_expand_prologue): Corresponding changes.
7480 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7483 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7484 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7486 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7489 * lra-constraints.c (curr_insn_transform): Process AND in the
7492 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7494 PR rtl-optimization/83986
7495 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7496 dependence against last_pending_memory_flush in addition to
7499 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7501 PR tree-optimization/81611
7502 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7505 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7508 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7511 2018-01-30 Richard Biener <rguenther@suse.de>
7512 Jakub Jelinek <jakub@redhat.com>
7514 PR tree-optimization/84111
7515 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7516 inner loops added during recursion, as they don't have up-to-date
7519 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7522 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7523 (can_inline_edge_by_limits_p): ... here.
7524 (can_early_inline_edge_p, check_callers,
7525 update_caller_keys, update_callee_keys, recursive_inlining,
7526 add_new_edges_to_heap, speculation_useful_p,
7527 inline_small_functions,
7528 inline_small_functions, flatten_function,
7529 inline_to_all_callers_1): Update.
7531 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7533 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7536 2018-01-30 Richard Biener <rguenther@suse.de>
7538 PR tree-optimization/83008
7539 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7540 invariant and constant vector uses in stmts when they need
7543 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7546 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7547 * configure: Regenerate.
7549 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7551 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7553 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7554 Use gen_rtx_REG rather than gen_lowpart.
7556 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7558 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7559 rather than 0 when creating partial subregs.
7561 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7563 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7566 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7569 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7570 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7571 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7572 flags. This restores the settings used before the 2017-07-24.
7573 Turning off pre increment/decrement/modify allows IVOPTS to
7574 optimize DF/SF loops where the index is an int.
7576 2018-01-29 Richard Biener <rguenther@suse.de>
7577 Kelvin Nilsen <kelvin@gcc.gnu.org>
7580 * tree-vect-stmts.c (vectorizable_call): Don't call
7581 targetm.vectorize_builtin_md_vectorized_function if callee is
7584 2018-01-22 Carl Love <cel@us.ibm.com>
7586 * doc/extend.tex: Fix typo in second arg in
7587 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7589 2018-01-29 Richard Biener <rguenther@suse.de>
7591 PR tree-optimization/84086
7592 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7593 (flush_ssaname_freelist): When SSA names were released reset
7594 the SCEV hash table.
7596 2018-01-29 Richard Biener <rguenther@suse.de>
7598 PR tree-optimization/84057
7599 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7600 removed paths when removing edges.
7602 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7604 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7605 -mfunction-return=@var{choice}.
7607 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7610 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7612 (layout::print_source_line): Likewise.
7613 (test_get_line_width_without_trailing_whitespace): Add test cases.
7615 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7618 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7621 2018-01-26 Jim Wilson <jimw@sifive.com>
7623 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7625 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7628 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7630 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7631 and CMP + SUB-immediate -> SUBS.
7633 2018-01-26 Martin Sebor <msebor@redhat.com>
7635 PR tree-optimization/83896
7636 * tree-ssa-strlen.c (get_string_len): Rename...
7637 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7638 Avoid assuming length is constant.
7639 (handle_char_store): Use HOST_WIDE_INT for string length.
7641 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7644 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7645 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7647 2018-01-26 Richard Biener <rguenther@suse.de>
7649 PR rtl-optimization/84003
7650 * dse.c (record_store): Only record redundant stores when
7651 the earlier store aliases at least all accesses the later one does.
7653 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7655 PR rtl-optimization/83985
7656 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7657 REG_CFA_RESTORE insns.
7658 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7659 REG_CFA_RESTORE insns here.
7662 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7663 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7665 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7667 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7668 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7669 (arc_init): Likewise.
7670 (arc_override_options): Likewise.
7671 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7673 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7675 * config/arc/arc.h (TARGET_DBNZ): Define.
7676 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7677 properly set the tune attribute.
7678 (dbnz): Use TARGET_DBNZ guard.
7679 * config/arc/arc.opt (mtune): Add core3 option.
7681 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7683 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7684 recognize new pic like addresses.
7685 (arc_delegitimize_address): Clean up.
7687 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7689 * config/arc/arc-arches.def: Option mrf16 valid for all
7691 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7692 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7693 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7694 * config/arc/arc-tables.opt: Regenerate.
7695 * config/arc/arc.c (arc_conditional_register_usage): Handle
7696 reduced register file case.
7697 (arc_file_start): Set must have build attributes.
7698 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7700 * config/arc/arc.opt (mrf16): Add new option.
7701 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7702 * config/arc/genmultilib.awk: Handle new mrf16 option.
7703 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7704 * config/arc/t-multilib: Regenerate.
7705 * doc/invoke.texi (ARC Options): Document mrf16 option.
7707 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7709 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7710 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7711 (arc_attribute_table): Add 'secure_call' attribute.
7712 (arc_print_operand): Print secure call operand.
7713 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7715 (arc_is_secure_call_p): New function. * config/arc/arc.md
7716 (call_i): Add support for sjli instruction.
7717 (call_value_i): Likewise.
7718 * config/arc/constraints.md (Csc): New constraint.
7720 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7721 John Eric Martin <John.Martin@emmicro-us.com>
7723 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7724 * config/arc/arc.c (_arc_jli_section): New struct.
7725 (arc_jli_section): New type.
7726 (rc_jli_sections): New static variable.
7727 (arc_handle_jli_attribute): New function.
7728 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7729 (arc_file_end): New function.
7730 (TARGET_ASM_FILE_END): Define.
7731 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7732 (arc_add_jli_section): New function.
7733 (jli_call_scan): Likewise.
7734 (arc_reorg): Call jli_call_scan.
7735 (arc_output_addsi): Remove 'S' from printing asm operand.
7736 (arc_is_jli_call_p): New function.
7737 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7739 (movhi_insn): Likewise.
7740 (movsi_insn): Likewise.
7741 (movsi_set_cc_insn): Likewise.
7742 (loadqi_update): Likewise.
7743 (load_zeroextendqisi_update): Likewise.
7744 (load_signextendqisi_update): Likewise.
7745 (loadhi_update): Likewise.
7746 (load_zeroextendhisi_update): Likewise.
7747 (load_signextendhisi_update): Likewise.
7748 (loadsi_update): Likewise.
7749 (loadsf_update): Likewise.
7750 (movsicc_insn): Likewise.
7751 (bset_insn): Likewise.
7752 (bxor_insn): Likewise.
7753 (bclr_insn): Likewise.
7754 (bmsk_insn): Likewise.
7755 (bicsi3_insn): Likewise.
7756 (cmpsi_cc_c_insn): Likewise.
7757 (movsi_ne): Likewise.
7758 (movsi_cond_exec): Likewise.
7759 (clrsbsi2): Likewise.
7770 (call_i): Remove 'S' asm letter, add jli instruction.
7771 (call_value_i): Likewise.
7772 * config/arc/arc.op (mjli-always): New option.
7773 * config/arc/constraints.md (Cji): New constraint.
7774 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7776 (subsf3_fpx): Likewise.
7777 (mulsf3_fpx): Likewise.
7778 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7780 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7782 * doc/invoke.texi (ARC): Document mjli-always option.
7784 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7786 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7787 avoid addition with 0 and use incw and decw where possible.
7789 2018-01-26 Richard Biener <rguenther@suse.de>
7791 PR tree-optimization/81082
7792 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7793 association if it requires casting to unsigned.
7794 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7795 from fold_plusminus_mult_expr to catch important cases late when
7796 range info is available.
7798 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7800 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7801 * configure.ac (hidden_linkonce): New test.
7802 * configure: Regenerate.
7803 * config.in: Regenerate.
7805 2018-01-26 Julia Koval <julia.koval@intel.com>
7807 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7808 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7809 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7810 _mm_mask_bitshuffle_epi64_mask): Fix type.
7811 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7812 USI_FTYPE_V4DI_V4DI_USI): Remove.
7813 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7814 __builtin_ia32_vpshufbitqmb256_mask,
7815 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7816 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7817 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7819 2018-01-26 Alan Modra <amodra@gmail.com>
7822 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7823 UNSPEC_VBPERMQ. Sort other unspecs.
7825 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7827 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7829 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7832 * predict.c (drop_profile): Do not push/pop cfun; update also
7834 (handle_missing_profiles): Fix logic looking for zero profiles.
7836 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7839 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7840 on functions with #pragma omp declare simd or functions with simd
7842 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7843 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7844 Remove trailing \n from warning_at calls.
7846 2018-01-25 Tom de Vries <tom@codesourcery.com>
7849 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7850 for neutered workers.
7852 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7855 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7856 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7858 2018-01-24 Jeff Law <law@redhat.com>
7861 * i386.c (get_probe_interval): Move to earlier point.
7862 (ix86_compute_frame_layout): If -fstack-clash-protection and
7863 the frame is larger than the probe interval, then use pushes
7864 to save registers rather than reg->mem moves.
7865 (ix86_expand_prologue): Remove conditional for int_registers_saved
7868 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7871 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7872 min/max for never referenced object.
7874 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7877 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7879 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7880 attributes from DECL_ATTRIBUTES (decl) without affecting
7881 DECL_ATTRIBUTES (current_function_decl).
7882 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7883 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7885 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7887 PR tree-optimization/83979
7888 * fold-const.c (fold_comparison): Use constant_boolean_node
7889 instead of boolean_{true,false}_node.
7891 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7893 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7896 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7898 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7899 Simplify the clause that sets the length attribute.
7900 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7901 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7902 clause that sets the length attribute.
7903 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7905 2018-01-24 Tom de Vries <tom@codesourcery.com>
7908 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7909 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7910 Add strict parameter.
7911 (prevent_branch_around_nothing): Insert dummy insn between branch to
7912 label and label with no ptx insn inbetween.
7913 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7915 2018-01-24 Tom de Vries <tom@codesourcery.com>
7918 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7919 for neutered threads in warp.
7920 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7922 2018-01-24 Richard Biener <rguenther@suse.de>
7924 PR tree-optimization/83176
7925 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7928 2018-01-24 Richard Biener <rguenther@suse.de>
7930 PR tree-optimization/82819
7931 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7932 code generating pluses that are no-ops in the target precision.
7934 2018-01-24 Richard Biener <rguenther@suse.de>
7937 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7939 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7941 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7942 to merge probabilities.
7943 * predict.c (probably_never_executed): Also mark as cold functions
7944 with global 0 profile and guessed local profile.
7945 * profile-count.c (profile_probability::combine_with_count): New
7947 * profile-count.h (profile_probability::operator*,
7948 profile_probability::operator*=, profile_probability::operator/,
7949 profile_probability::operator/=): Reduce precision to adjusted
7950 and set value to guessed on contradictory divisions.
7951 (profile_probability::combine_with_freq): Remove.
7952 (profile_probability::combine_wiht_count): Declare.
7953 (profile_count::force_nonzero):: Set to adjusted.
7954 (profile_count::probability_in):: Set quality to adjusted.
7955 * tree-ssa-tail-merge.c (replace_block_by): Use
7958 2018-01-23 Andrew Waterman <andrew@sifive.com>
7959 Jim Wilson <jimw@sifive.com>
7961 * config/riscv/riscv.c (riscv_stack_boundary): New.
7962 (riscv_option_override): Set riscv_stack_boundary. Handle
7963 riscv_preferred_stack_boundary_arg.
7964 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7965 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7966 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7967 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7968 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7969 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7971 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7974 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7975 of struct ix86_frame.
7976 (ix86_expand_epilogue): Likewise. Add a local variable for
7977 the reg_save_offset field in struct ix86_frame.
7979 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7981 PR tree-optimization/82604
7982 * tree-loop-distribution.c (enum partition_kind): New enum item
7983 PKIND_PARTIAL_MEMSET.
7984 (partition_builtin_p): Support above new enum item.
7985 (generate_code_for_partition): Ditto.
7986 (compute_access_range): Differentiate cases that equality can be
7987 proven at all loops, the innermost loops or no loops.
7988 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7989 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7990 (finalize_partitions, distribute_loop): Don't fuse partition of
7991 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7992 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7995 2018-01-23 Martin Liska <mliska@suse.cz>
7997 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7998 order to ignore the predictor.
7999 (PRED_POLYMORPHIC_CALL): Likewise.
8000 (PRED_RECURSIVE_CALL): Likewise.
8002 2018-01-23 Martin Liska <mliska@suse.cz>
8004 * tree-profile.c (tree_profiling): Print function header to
8005 aware reader which function we are working on.
8006 * value-prof.c (gimple_find_values_to_profile): Do not print
8007 not interesting value histograms.
8009 2018-01-23 Martin Liska <mliska@suse.cz>
8011 * profile-count.h (enum profile_quality): Add
8012 profile_uninitialized as the first value. Do not number values
8013 as they are zero based.
8014 (profile_count::verify): Update sanity check.
8015 (profile_probability::verify): Likewise.
8017 2018-01-23 Nathan Sidwell <nathan@acm.org>
8019 * doc/invoke.texi (ffor-scope): Deprecate.
8021 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8023 PR tree-optimization/83510
8024 * domwalk.c (set_all_edges_as_executable): New function.
8025 (dom_walker::dom_walker): Convert bool param
8026 "skip_unreachable_blocks" to enum reachability. Move setup of
8027 edge flags to set_all_edges_as_executable and only do it when
8028 reachability is REACHABLE_BLOCKS.
8029 * domwalk.h (enum dom_walker::reachability): New enum.
8030 (dom_walker::dom_walker): Convert bool param
8031 "skip_unreachable_blocks" to enum reachability.
8032 (set_all_edges_as_executable): New decl.
8033 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8034 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8036 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8037 but converting true to REACHABLE_BLOCKS.
8038 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8040 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8041 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8042 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8044 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8045 if check_all_array_refs will be called.
8047 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8049 * tree.c (selftest::test_location_wrappers): Add more test
8052 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8054 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8055 (selftest::test_bit_in_range): Likewise.
8057 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8060 * doc/sourcebuild.texi (vect_float): Say that the selector
8061 only describes the situation when -funsafe-math-optimizations is on.
8062 (vect_float_strict): Document.
8064 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8066 PR tree-optimization/83965
8067 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8068 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8069 instead of checking only for a reduction.
8070 (vect_recog_widen_sum_pattern): Likewise.
8072 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8074 * predict.c (probably_never_executed): Only use precise profile info.
8075 (compute_function_frequency): Skip after inlining hack since we now
8076 have quality checking.
8078 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8080 * profile-count.h (profile_probability::very_unlikely,
8081 profile_probability::unlikely, profile_probability::even): Set
8082 precision to guessed.
8084 2018-01-23 Richard Biener <rguenther@suse.de>
8086 PR tree-optimization/83963
8087 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8088 Properly terminate dominator walk when crossing the exit edge not
8089 when visiting its source block.
8091 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8094 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8095 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8097 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8099 PR tree-optimization/83957
8100 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8101 semicolon after for body surrounded by braces.
8103 PR tree-optimization/83081
8104 * profile-count.h (profile_probability::split): New method.
8105 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8106 Use profile_probability::split.
8107 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8108 when splitting a single conditional jump into 2.
8110 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8112 PR tree-optimization/69452
8113 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8116 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8118 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8119 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8120 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8122 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8124 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8125 * config/rl78/rl78.md (movdi): New define_expand.
8126 * config/rl78/rl78.c (rl78_split_movdi): New function.
8128 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8131 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8133 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8134 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8135 128-bit to produce an UNSPEC move to get the double word with the
8136 signbit and then a shift directly to do signbit.
8137 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8138 implementation with a new version that just does either a direct
8139 move or a regular move. Move memory interface to separate insns.
8140 Move insns so they are next to the expander.
8141 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8142 with signbit move. Split big and little endian case.
8143 (signbit<mode>2_dm_mem_le): Likewise.
8144 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8145 (signbit<mode>2_dm2): Likewise.
8147 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8149 * config/rl78/rl78.md (anddi3): New define_expand.
8151 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8153 * config/rl78/rl78.md (umindi3): New define_expand.
8155 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8157 * config/rl78/rl78.md (smindi3): New define_expand.
8159 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8161 * config/rl78/rl78.md (smaxdi3): New define_expand.
8163 2018-01-22 Carl Love <cel@us.ibm.com>
8165 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8166 LVX_V1TI): Add macro expansion.
8167 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8168 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8169 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8170 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8171 Change check to determine if the instruction is a byte reversing
8172 entry. Fix typo in comment.
8173 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8174 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8175 Add def_builtin calls for new builtins.
8176 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8177 Add define_insn expansion.
8179 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8181 * config/rl78/rl78.md (umaxdi3): New define_expand.
8183 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8185 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8186 for non-QImode registers.
8188 2018-01-22 Richard Biener <rguenther@suse.de>
8190 PR tree-optimization/83963
8191 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8192 including the loop exit block.
8193 (scop_detection::merge_sese): Likewise.
8194 (scop_detection::add_scop): Do it here instead.
8196 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8198 * doc/sourcebuild.texi (arm_softfloat): Document.
8200 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8203 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8204 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8205 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8207 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8208 David Edelsohn <dje.gcc@gmail.com>
8211 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8212 Change "crset eq" to "crset 2".
8213 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8214 (*call_indirect_aix<mode>_nospec): Likewise.
8215 (*call_value_indirect_aix<mode>_nospec): Likewise.
8216 (*call_indirect_elfv2<mode>_nospec): Likewise.
8217 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8218 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8219 change assembly output from . to $.
8220 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8221 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8222 (*tablejump<mode>_internal1_nospec): Likewise.
8224 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8227 * config/sh/sh_optimize_sett_clrt.cc:
8228 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8230 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8232 PR tree-optimization/83940
8233 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8234 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8235 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8236 use it to pass back the definition type.
8237 (vect_check_store_rhs): Likewise rhs_dt_out.
8238 (vect_build_gather_load_calls): Add a mask_dt argument and use
8239 it instead of a call to vect_is_simple_use.
8240 (vectorizable_store): Update calls to vect_check_load_store_mask
8241 and vect_check_store_rhs. Use the dt returned by the latter instead
8242 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8243 instead of calls to vect_is_simple_use. Pass the scalar rather
8244 than the vector operand to vect_is_simple_use when handling
8245 second and subsequent copies of an rhs value.
8246 (vectorizable_load): Update calls to vect_check_load_store_mask
8247 and vect_build_gather_load_calls. Use the cached mask_dt and
8248 gs_info.offset_dt instead of calls to vect_is_simple_use.
8250 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8253 * tree-emutls.c: Include gimplify.h.
8254 (lower_emutls_2): New function.
8255 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8256 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8257 it before further processing.
8260 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8261 UINTVAL (trueop1) instead of INTVAL (op1).
8263 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8267 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8268 INCOMING_FRAME_SP_OFFSET if not defined.
8269 (scan_trace): Add ENTRY argument. If true and
8270 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8271 emit a note to adjust the CFA offset.
8272 (create_cfi_notes): Adjust scan_trace callers.
8273 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8274 INCOMING_FRAME_SP_OFFSET in the CIE.
8275 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8276 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8278 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8279 * doc/tm.texi: Regenerated.
8281 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8283 PR rtl-optimization/83147
8284 * lra-constraints.c (remove_inheritance_pseudos): Use
8285 lra_substitute_pseudo_within_insn.
8287 2018-01-19 Tom de Vries <tom@codesourcery.com>
8288 Cesar Philippidis <cesar@codesourcery.com>
8291 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8293 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8296 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8297 spaces for function labels.
8299 2018-01-19 Martin Liska <mliska@suse.cz>
8301 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8302 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8303 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8304 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8305 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8306 (PRED_CONST_RETURN): Change from 69 to 65.
8307 (PRED_NULL_RETURN): Change from 91 to 71.
8308 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8309 (PRED_LOOP_GUARD): Change from 66 to 73.
8311 2018-01-19 Martin Liska <mliska@suse.cz>
8313 * predict.c (predict_insn_def): Add new assert.
8314 (struct branch_predictor): Change type to signed integer.
8315 (test_prediction_value_range): Amend test to cover
8317 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8318 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8319 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8320 (PRED_LOOP_IV_COMPARE): Likewise.
8321 * predict.h (PROB_UNINITIALIZED): Define new constant.
8323 2018-01-19 Martin Liska <mliska@suse.cz>
8325 * predict.c (dump_prediction): Add new format for
8326 analyze_brprob.py script which is enabled with -details
8328 * profile-count.h (precise_p): New function.
8330 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8332 PR tree-optimization/83922
8333 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8334 there are no statements that need masking.
8335 (vect_active_double_reduction_p): New function.
8336 (vect_analyze_loop_operations): Use it when handling phis that
8337 are not in the loop header.
8339 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8341 PR tree-optimization/83914
8342 * tree-vect-loop.c (vectorizable_induction): Don't convert
8343 init_expr or apply the peeling adjustment for inductions
8344 that are nested within the vectorized loop.
8346 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8348 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8351 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8355 * function.h (gimplify_parameters): Add gimple_seq * argument.
8356 * function.c: Include gimple.h and options.h.
8357 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8358 for the added local temporaries if needed.
8359 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8360 if there are any parameter cleanups, wrap whole body into a
8361 try/finally with the cleanups.
8363 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8366 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8367 Use GET_MODE_CLASS for scalar floating point.
8369 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8373 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8374 Fix call of call_cgraph_insertion_hooks.
8376 2018-01-18 Martin Sebor <msebor@redhat.com>
8378 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8380 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8383 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8386 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8389 * common.opt: (-ffile-prefix-map): New option.
8390 * opts.c (common_handle_option): Defer it.
8391 * opts-global.c (handle_common_deferred_options): Handle it.
8392 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8393 * file-prefix-map.h: New file.
8394 (remap_debug_filename, add_debug_prefix_map): ...here.
8395 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8396 * final.c (debug_prefix_map, add_debug_prefix_map
8397 remap_debug_filename): Move to...
8398 * file-prefix-map.c: New file.
8399 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8400 generalize, get rid of alloca(), use strrchr() instead of strchr().
8401 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8402 Implement in terms of add_prefix_map().
8403 (remap_macro_filename, remap_debug_filename): Implement in term of
8405 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8406 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8407 * dbxout.c: Include file-prefix-map.h.
8408 * varasm.c: Likewise.
8409 * vmsdbgout.c: Likewise.
8410 * xcoffout.c: Likewise.
8411 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8412 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8413 * doc/invoke.texi (-ffile-prefix-map): Document.
8414 (-fdebug-prefix-map): Update description.
8416 2018-01-18 Martin Liska <mliska@suse.cz>
8418 * config/i386/i386.c (indirect_thunk_name): Document that also
8420 (output_indirect_thunk): Document why both instructions
8421 (pause and lfence) are generated.
8423 2018-01-18 Richard Biener <rguenther@suse.de>
8425 PR tree-optimization/83887
8426 * graphite-scop-detection.c
8427 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8428 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8429 (scop_detection::merge_sese): Re-implement with a flood-fill
8430 algorithm that properly finds a SESE region if it exists.
8432 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8435 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8436 pointer_diff optimizations use view_convert instead of convert.
8438 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8440 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8441 Generate different code for -mno-speculate-indirect-jumps.
8442 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8443 (*call_indirect_aix<mode>): Disable for
8444 -mno-speculate-indirect-jumps.
8445 (*call_indirect_aix<mode>_nospec): New define_insn.
8446 (*call_value_indirect_aix<mode>): Disable for
8447 -mno-speculate-indirect-jumps.
8448 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8449 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8450 -mno-speculate-indirect-jumps.
8451 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8453 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8455 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8456 long double type, set the flags for noting the default long double
8457 type, even if we don't pass or return a long double type.
8459 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8462 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8465 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8467 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8468 support for merge[hl].
8469 (fold_mergehl_helper): New helper function.
8470 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8471 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8472 (altivec_vmrglw_direct): Add xxmrglw insn.
8474 2018-01-17 Andrew Waterman <andrew@sifive.com>
8476 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8477 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8479 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8482 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8483 call the lto_location_cache before reading the
8484 DECL_SOURCE_LOCATION of the types.
8486 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8487 Richard Sandiford <richard.sandiford@linaro.org>
8489 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8490 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8491 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8492 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8493 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8495 * config/aarch64/constraints.md (aarch64_movti_operand):
8497 * config/aarch64/predicates.md (Uti): Add new constraint.
8499 2018-01-17 Carl Love <cel@us.ibm.com>
8501 * config/rs6000/vsx.md (define_expand xl_len_r,
8502 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8503 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8505 (define_expand, define_insn): Move the shift left from the
8506 define_insn to the define_expand for lxvl and stxvl instructions.
8507 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8508 and XL_LEN_R definitions to PURE.
8510 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8512 * config/i386/i386.c (indirect_thunk_name): Declare regno
8513 as unsigned int. Compare regno with INVALID_REGNUM.
8514 (output_indirect_thunk): Ditto.
8515 (output_indirect_thunk_function): Ditto.
8516 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8517 in the call to output_indirect_thunk_function.
8519 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8522 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8523 rather than the size of inner_type to determine the stack slot size
8524 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8526 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8529 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8532 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8534 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8535 endian Linux systems to optionally enable multilibs for selecting
8536 the long double type if the user configured an explicit type.
8537 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8538 have no long double multilibs if not defined.
8539 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8540 warn if the user used -mabi={ieee,ibm}longdouble and we built
8541 multilibs for long double.
8542 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8543 appropriate multilib option.
8544 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8546 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8547 for building long double multilibs.
8548 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8550 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8552 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8555 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8557 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8560 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8563 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8566 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8568 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8569 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8571 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8573 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8574 different rtl trees depending on TARGET_64BIT.
8575 (rs6000_gen_lvx): Likewise.
8577 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8579 * config/visium/visium.md (nop): Tweak comment.
8580 (hazard_nop): Likewise.
8582 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8584 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8585 -mspeculate-indirect-jumps.
8586 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8587 for -mno-speculate-indirect-jumps.
8588 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8589 (*call_value_indirect_elfv2<mode>): Disable for
8590 -mno-speculate-indirect-jumps.
8591 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8592 (indirect_jump): Emit different RTL for
8593 -mno-speculate-indirect-jumps.
8594 (*indirect_jump<mode>): Disable for
8595 -mno-speculate-indirect-jumps.
8596 (*indirect_jump<mode>_nospec): New define_insn.
8597 (tablejump): Emit different RTL for
8598 -mno-speculate-indirect-jumps.
8599 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8600 (tablejumpsi_nospec): New define_expand.
8601 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8602 (tablejumpdi_nospec): New define_expand.
8603 (*tablejump<mode>_internal1): Disable for
8604 -mno-speculate-indirect-jumps.
8605 (*tablejump<mode>_internal1_nospec): New define_insn.
8606 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8609 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8611 * caller-save.c (insert_save): Drop unnecessary parameter. All
8614 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8615 Richard Biener <rguenth@suse.de>
8618 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8619 return early, inline manually is_gimple_sizepos. Make sure if we
8620 call gimplify_expr we don't end up with a gimple constant.
8621 * tree.c (variably_modified_type_p): Don't return true for
8622 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8623 * gimplify.h (is_gimple_sizepos): Remove.
8625 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8627 PR tree-optimization/83857
8628 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8629 vectorizable_live_operation for pure SLP statements.
8630 (vectorizable_live_operation): Handle PHIs.
8632 2018-01-16 Richard Biener <rguenther@suse.de>
8634 PR tree-optimization/83867
8635 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8636 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8638 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8641 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8642 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8643 If off is not INTEGER_CST, issue a may not be aligned warning
8644 rather than isn't aligned. Use isn%'t rather than isn't.
8645 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8647 <case MULT_EXPR>: Improve the case when bottom and one of the
8648 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8649 operand, in that case check if the other operand is multiple of
8650 bottom divided by the INTEGER_CST operand.
8652 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8655 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8656 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8657 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8658 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8659 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8660 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8661 (pa_function_arg_size): New function.
8663 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8665 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8666 in a separate statement.
8668 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8670 PR tree-optimization/83847
8671 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8672 group gathers and scatters.
8674 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8676 PR rtl-optimization/86620
8677 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8679 PR rtl-optimization/83213
8680 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8681 to last if both are JUMP_INSNs.
8683 PR tree-optimization/83843
8684 * gimple-ssa-store-merging.c
8685 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8686 store_immediate_info for bswap/nop orig_stores.
8688 2018-01-15 Andrew Waterman <andrew@sifive.com>
8690 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8692 <UDIV>: Increase cost if !TARGET_DIV.
8694 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8696 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8697 (define_attr "cr_logical_3op"): New.
8698 (cceq_ior_compare): Adjust.
8699 (cceq_ior_compare_complement): Adjust.
8700 (*cceq_rev_compare): Adjust.
8701 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8702 (is_cracked_insn): Adjust.
8703 (insn_must_be_first_in_group): Adjust.
8704 * config/rs6000/40x.md: Adjust.
8705 * config/rs6000/440.md: Adjust.
8706 * config/rs6000/476.md: Adjust.
8707 * config/rs6000/601.md: Adjust.
8708 * config/rs6000/603.md: Adjust.
8709 * config/rs6000/6xx.md: Adjust.
8710 * config/rs6000/7450.md: Adjust.
8711 * config/rs6000/7xx.md: Adjust.
8712 * config/rs6000/8540.md: Adjust.
8713 * config/rs6000/cell.md: Adjust.
8714 * config/rs6000/e300c2c3.md: Adjust.
8715 * config/rs6000/e500mc.md: Adjust.
8716 * config/rs6000/e500mc64.md: Adjust.
8717 * config/rs6000/e5500.md: Adjust.
8718 * config/rs6000/e6500.md: Adjust.
8719 * config/rs6000/mpc.md: Adjust.
8720 * config/rs6000/power4.md: Adjust.
8721 * config/rs6000/power5.md: Adjust.
8722 * config/rs6000/power6.md: Adjust.
8723 * config/rs6000/power7.md: Adjust.
8724 * config/rs6000/power8.md: Adjust.
8725 * config/rs6000/power9.md: Adjust.
8726 * config/rs6000/rs64.md: Adjust.
8727 * config/rs6000/titan.md: Adjust.
8729 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8731 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8732 ix86_indirect_branch_register logic.
8734 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8736 * config/i386/constraints.md (Bs): Update
8737 ix86_indirect_branch_register check. Don't check
8738 ix86_indirect_branch_register with GOT_memory_operand.
8740 * config/i386/predicates.md (GOT_memory_operand): Don't check
8741 ix86_indirect_branch_register here.
8742 (GOT32_symbol_operand): Likewise.
8744 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8746 * config/i386/predicates.md (constant_call_address_operand):
8747 Rewrite ix86_indirect_branch_register logic.
8748 (sibcall_insn_operand): Likewise.
8750 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8752 * config/i386/constraints.md (Bs): Replace
8753 ix86_indirect_branch_thunk_register with
8754 ix86_indirect_branch_register.
8756 * config/i386/i386.md (indirect_jump): Likewise.
8757 (tablejump): Likewise.
8758 (*sibcall_memory): Likewise.
8759 (*sibcall_value_memory): Likewise.
8760 Peepholes of indirect call and jump via memory: Likewise.
8761 * config/i386/i386.opt: Likewise.
8762 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8763 (GOT_memory_operand): Likewise.
8764 (call_insn_operand): Likewise.
8765 (sibcall_insn_operand): Likewise.
8766 (GOT32_symbol_operand): Likewise.
8768 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8771 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8772 type rather than type addr's type points to.
8773 (expand_omp_atomic_mutex): Likewise.
8774 (expand_omp_atomic): Likewise.
8776 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8779 * config/i386/i386.c (output_indirect_thunk_function): Use
8780 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8781 for __x86_return_thunk.
8783 2018-01-15 Richard Biener <rguenther@suse.de>
8786 * expmed.c (extract_bit_field_1): Fix typo.
8788 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8791 * config/arm/iterators.md (VF): New mode iterator.
8792 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8793 Remove integer-related logic from pattern.
8794 (neon_vabd<mode>_3): Likewise.
8796 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8799 * common.opt (fstrict-overflow): No longer an alias.
8800 (fwrapv-pointer): New option.
8801 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8802 also for pointer types based on flag_wrapv_pointer.
8803 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8804 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8805 opts->x_flag_wrapv got set.
8806 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8807 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8808 POINTER_TYPE_OVERFLOW_UNDEFINED.
8809 * match.pd: Likewise in address comparison pattern.
8810 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8812 2018-01-15 Richard Biener <rguenther@suse.de>
8815 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8816 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8817 Reset type names to their identifier if their TYPE_DECL doesn't
8818 have linkage (and thus is used for ODR and devirt).
8819 (save_debug_info_for_decl): Remove.
8820 (save_debug_info_for_type): Likewise.
8821 (add_tree_to_fld_list): Adjust.
8822 * tree-pretty-print.c (dump_generic_node): Make dumping of
8823 type names more robust.
8825 2018-01-15 Richard Biener <rguenther@suse.de>
8827 * BASE-VER: Bump to 8.0.1.
8829 2018-01-14 Martin Sebor <msebor@redhat.com>
8832 * builtins.c (check_access): Avoid warning when the no-warning bit
8835 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8837 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8838 * ira-color (allocno_hard_regs_compare): Likewise.
8840 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8843 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8844 Use .pushsection/.popsection.
8846 2018-01-14 Martin Sebor <msebor@redhat.com>
8849 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8851 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8853 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8854 entry from extra_headers.
8855 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8856 extra_headers, make the list bitwise identical to the i?86-*-* one.
8858 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8860 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8861 -mcmodel=large with -mindirect-branch=thunk,
8862 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8863 -mfunction-return=thunk-extern.
8864 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8865 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8866 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8868 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8870 * config/i386/i386.c (print_reg): Print the name of the full
8871 integer register without '%'.
8872 (ix86_print_operand): Handle 'V'.
8873 * doc/extend.texi: Document 'V' modifier.
8875 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8877 * config/i386/constraints.md (Bs): Disallow memory operand for
8878 -mindirect-branch-register.
8880 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8881 (GOT_memory_operand): Likewise.
8882 (call_insn_operand): Likewise.
8883 (sibcall_insn_operand): Likewise.
8884 (GOT32_symbol_operand): Likewise.
8885 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8886 for -mindirect-branch-register.
8887 (tablejump): Likewise.
8888 (*sibcall_memory): Likewise.
8889 (*sibcall_value_memory): Likewise.
8890 Disallow peepholes of indirect call and jump via memory for
8891 -mindirect-branch-register.
8892 (*call_pop): Replace m with Bw.
8893 (*call_value_pop): Likewise.
8894 (*sibcall_pop_memory): Replace m with Bs.
8895 * config/i386/i386.opt (mindirect-branch-register): New option.
8896 * doc/invoke.texi: Document -mindirect-branch-register option.
8898 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8900 * config/i386/i386-protos.h (ix86_output_function_return): New.
8901 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8902 set function_return_type.
8903 (indirect_thunk_name): Add ret_p to indicate thunk for function
8905 (output_indirect_thunk_function): Pass false to
8906 indirect_thunk_name.
8907 (ix86_output_indirect_branch_via_reg): Likewise.
8908 (ix86_output_indirect_branch_via_push): Likewise.
8909 (output_indirect_thunk_function): Create alias for function
8910 return thunk if regno < 0.
8911 (ix86_output_function_return): New function.
8912 (ix86_handle_fndecl_attribute): Handle function_return.
8913 (ix86_attribute_table): Add function_return.
8914 * config/i386/i386.h (machine_function): Add
8915 function_return_type.
8916 * config/i386/i386.md (simple_return_internal): Use
8917 ix86_output_function_return.
8918 (simple_return_internal_long): Likewise.
8919 * config/i386/i386.opt (mfunction-return=): New option.
8920 (indirect_branch): Mention -mfunction-return=.
8921 * doc/extend.texi: Document function_return function attribute.
8922 * doc/invoke.texi: Document -mfunction-return= option.
8924 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8926 * config/i386/i386-opts.h (indirect_branch): New.
8927 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8928 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8929 with local indirect jump when converting indirect call and jump.
8930 (ix86_set_indirect_branch_type): New.
8931 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8932 (indirectlabelno): New.
8933 (indirect_thunk_needed): Likewise.
8934 (indirect_thunk_bnd_needed): Likewise.
8935 (indirect_thunks_used): Likewise.
8936 (indirect_thunks_bnd_used): Likewise.
8937 (INDIRECT_LABEL): Likewise.
8938 (indirect_thunk_name): Likewise.
8939 (output_indirect_thunk): Likewise.
8940 (output_indirect_thunk_function): Likewise.
8941 (ix86_output_indirect_branch_via_reg): Likewise.
8942 (ix86_output_indirect_branch_via_push): Likewise.
8943 (ix86_output_indirect_branch): Likewise.
8944 (ix86_output_indirect_jmp): Likewise.
8945 (ix86_code_end): Call output_indirect_thunk_function if needed.
8946 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8948 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8949 (ix86_attribute_table): Add indirect_branch.
8950 * config/i386/i386.h (machine_function): Add indirect_branch_type
8951 and has_local_indirect_jump.
8952 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8954 (tablejump): Likewise.
8955 (*indirect_jump): Use ix86_output_indirect_jmp.
8956 (*tablejump_1): Likewise.
8957 (simple_return_indirect_internal): Likewise.
8958 * config/i386/i386.opt (mindirect-branch=): New option.
8959 (indirect_branch): New.
8962 (thunk-inline): Likewise.
8963 (thunk-extern): Likewise.
8964 * doc/extend.texi: Document indirect_branch function attribute.
8965 * doc/invoke.texi: Document -mindirect-branch= option.
8967 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8970 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8972 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8974 * ipa-inline.c (want_inline_small_function_p): Return false if
8975 inlining has already failed with CIF_FINAL_ERROR.
8976 (update_caller_keys): Call want_inline_small_function_p before
8978 (update_callee_keys): Likewise.
8980 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8982 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8984 (rs6000_quadword_masked_address_p): Likewise.
8985 (quad_aligned_load_p): Likewise.
8986 (quad_aligned_store_p): Likewise.
8987 (const_load_sequence_p): Add comment to describe the outer-most loop.
8988 (mimic_memory_attributes_and_flags): New function.
8989 (rs6000_gen_stvx): Likewise.
8990 (replace_swapped_aligned_store): Likewise.
8991 (rs6000_gen_lvx): Likewise.
8992 (replace_swapped_aligned_load): Likewise.
8993 (replace_swapped_load_constant): Capitalize argument name in
8994 comment describing this function.
8995 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8996 and stores that access quad-word aligned addresses and replace
8997 with stvx or lvx instructions when appropriate.
8998 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8999 New function prototype.
9000 (rs6000_quadword_masked_address_p): Likewise.
9001 (rs6000_gen_lvx): Likewise.
9002 (rs6000_gen_stvx): Likewise.
9003 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9004 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9005 when memory address is aligned.
9006 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9007 this split to select lvx instruction when memory address is aligned.
9008 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9009 instruction when memory address is aligned.
9010 (*vsx_le_perm_load_v16qi): Likewise.
9011 (four unnamed splitters): Modify to select the stvx instruction
9012 when memory is aligned.
9014 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9016 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9017 which appears in the queue multiple times.
9019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9020 Alan Hayward <alan.hayward@arm.com>
9021 David Sherwood <david.sherwood@arm.com>
9023 * tree-vectorizer.h (vec_lower_bound): New structure.
9024 (_loop_vec_info): Add check_nonzero and lower_bounds.
9025 (LOOP_VINFO_CHECK_NONZERO): New macro.
9026 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9027 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9028 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9029 fields. Make seg_len the distance travelled, not including the
9031 (dr_direction_indicator): Declare.
9032 (dr_zero_step_indicator): Likewise.
9033 (dr_known_forward_stride_p): Likewise.
9034 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9036 (runtime_alias_check_p): Allow runtime alias checks with
9038 (operator ==): Compare access_size and align.
9039 (prune_runtime_alias_test_list): Rework for new distinction between
9040 the access_size and seg_len.
9041 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9043 (get_segment_min_max): New function.
9044 (create_intersect_range_checks): Use it.
9045 (dr_step_indicator): New function.
9046 (dr_direction_indicator): Likewise.
9047 (dr_zero_step_indicator): Likewise.
9048 (dr_known_forward_stride_p): Likewise.
9049 * tree-loop-distribution.c (data_ref_segment_size): Return
9050 DR_STEP * (niters - 1).
9051 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9053 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9054 (vect_preserves_scalar_order_p): New function, split out from...
9055 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9056 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9057 (vect_vfa_access_size): New function.
9058 (vect_vfa_align): Likewise.
9059 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9060 (dump_lower_bound): New function.
9061 (vect_check_lower_bound): Likewise.
9062 (vect_small_gap_p): Likewise.
9063 (vectorizable_with_step_bound_p): Likewise.
9064 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9065 depencies if the vectorization factor is 1. Convert the checks
9066 for nonzero steps into checks on the bounds of DR_STEP. Try using
9067 a bunds check for variable steps if the minimum required step is
9068 relatively small. Update calls to the dr_with_seg_len
9069 constructor and to vect_compile_time_alias.
9070 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9072 (vect_loop_versioning): Call it.
9073 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9075 (vect_estimate_min_profitable_iters): Account for any bounds checks.
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 * doc/sourcebuild.texi (vect_scatter_store): Document.
9082 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9084 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9086 * genopinit.c (main): Add supports_vec_scatter_store and
9087 supports_vec_scatter_store_cached to target_optabs.
9088 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9089 IFN_MASK_SCATTER_STORE.
9090 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9092 * internal-fn.h (internal_store_fn_p): Declare.
9093 (internal_fn_stored_value_index): Likewise.
9094 * internal-fn.c (scatter_store_direct): New macro.
9095 (expand_scatter_store_optab_fn): New function.
9096 (direct_scatter_store_optab_supported_p): New macro.
9097 (internal_store_fn_p): New function.
9098 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9099 IFN_MASK_SCATTER_STORE.
9100 (internal_fn_mask_index): Likewise.
9101 (internal_fn_stored_value_index): New function.
9102 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9104 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9105 * optabs-query.c (supports_vec_scatter_store_p): New function.
9106 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9107 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9108 true for scatter stores.
9109 (vect_gather_scatter_fn_p): Handle scatter stores too.
9110 (vect_check_gather_scatter): Consider using scatter stores if
9111 supports_vec_scatter_store_p.
9112 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9114 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9115 internal_fn_stored_value_index.
9116 (check_load_store_masking): Handle scatter stores too.
9117 (vect_get_store_rhs): Make public.
9118 (vectorizable_call): Use internal_store_fn_p.
9119 (vectorizable_store): Handle scatter store internal functions.
9120 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9121 when deciding whether the end of the group has been reached.
9122 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9123 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9124 (mask_scatter_store<mode>): New insns.
9126 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9127 Alan Hayward <alan.hayward@arm.com>
9128 David Sherwood <david.sherwood@arm.com>
9130 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9131 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9132 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9134 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9135 Use vect_truncate_gather_scatter_offset if we can't treat the
9136 operation as a normal gather load or scatter store.
9137 (get_group_load_store_type): Take the gather_scatter_info
9138 as argument. Try using a gather load or scatter store for
9139 single-element groups.
9140 (get_load_store_type): Update calls to get_group_load_store_type
9141 and vect_use_strided_gather_scatters_p.
9143 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9144 Alan Hayward <alan.hayward@arm.com>
9145 David Sherwood <david.sherwood@arm.com>
9147 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9148 optional tree argument.
9149 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9151 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9152 but continue to use the current value as a fallback.
9153 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9154 to compare the updates.
9155 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9156 (get_load_store_type): Use it when handling a strided access.
9157 (vect_get_strided_load_store_ops): New function.
9158 (vect_get_data_ptr_increment): Likewise.
9159 (vectorizable_load): Handle strided gather loads. Always pass
9160 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9162 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9163 Alan Hayward <alan.hayward@arm.com>
9164 David Sherwood <david.sherwood@arm.com>
9166 * doc/md.texi (gather_load@var{m}): Document.
9167 (mask_gather_load@var{m}): Likewise.
9168 * genopinit.c (main): Add supports_vec_gather_load and
9169 supports_vec_gather_load_cached to target_optabs.
9170 * optabs-tree.c (init_tree_optimization_optabs): Use
9171 ggc_cleared_alloc to allocate target_optabs.
9172 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9173 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9175 * internal-fn.h (internal_load_fn_p): Declare.
9176 (internal_gather_scatter_fn_p): Likewise.
9177 (internal_fn_mask_index): Likewise.
9178 (internal_gather_scatter_fn_supported_p): Likewise.
9179 * internal-fn.c (gather_load_direct): New macro.
9180 (expand_gather_load_optab_fn): New function.
9181 (direct_gather_load_optab_supported_p): New macro.
9182 (direct_internal_fn_optab): New function.
9183 (internal_load_fn_p): Likewise.
9184 (internal_gather_scatter_fn_p): Likewise.
9185 (internal_fn_mask_index): Likewise.
9186 (internal_gather_scatter_fn_supported_p): Likewise.
9187 * optabs-query.c (supports_at_least_one_mode_p): New function.
9188 (supports_vec_gather_load_p): Likewise.
9189 * optabs-query.h (supports_vec_gather_load_p): Declare.
9190 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9191 and memory_type field.
9192 (NUM_PATTERNS): Bump to 15.
9193 * tree-vect-data-refs.c: Include internal-fn.h.
9194 (vect_gather_scatter_fn_p): New function.
9195 (vect_describe_gather_scatter_call): Likewise.
9196 (vect_check_gather_scatter): Try using internal functions for
9197 gather loads. Recognize existing calls to a gather load function.
9198 (vect_analyze_data_refs): Consider using gather loads if
9199 supports_vec_gather_load_p.
9200 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9201 (vect_get_gather_scatter_offset_type): Likewise.
9202 (vect_convert_mask_for_vectype): Likewise.
9203 (vect_add_conversion_to_patterm): Likewise.
9204 (vect_try_gather_scatter_pattern): Likewise.
9205 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9206 (vect_vect_recog_func_ptrs): Add it.
9207 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9208 internal_fn_mask_index and internal_gather_scatter_fn_p.
9209 (check_load_store_masking): Take the gather_scatter_info as an
9210 argument and handle gather loads.
9211 (vect_get_gather_scatter_ops): New function.
9212 (vectorizable_call): Check internal_load_fn_p.
9213 (vectorizable_load): Likewise. Handle gather load internal
9215 (vectorizable_store): Update call to check_load_store_masking.
9216 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9217 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9218 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9219 (aarch64_gather_scale_operand_d): New predicates.
9220 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9221 (mask_gather_load<mode>): New insns.
9223 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9224 Alan Hayward <alan.hayward@arm.com>
9225 David Sherwood <david.sherwood@arm.com>
9227 * optabs.def (fold_left_plus_optab): New optab.
9228 * doc/md.texi (fold_left_plus_@var{m}): Document.
9229 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9230 * internal-fn.c (fold_left_direct): Define.
9231 (expand_fold_left_optab_fn): Likewise.
9232 (direct_fold_left_optab_supported_p): Likewise.
9233 * fold-const-call.c (fold_const_fold_left): New function.
9234 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9235 * tree-parloops.c (valid_reduction_p): New function.
9236 (gather_scalar_reductions): Use it.
9237 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9238 (vect_finish_replace_stmt): Declare.
9239 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9240 (needs_fold_left_reduction_p): New function, split out from...
9241 (vect_is_simple_reduction): ...here. Accept reductions that
9242 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9243 (vect_force_simple_reduction): Also store the reduction type in
9244 the assignment's STMT_VINFO_REDUC_TYPE.
9245 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9246 (merge_with_identity): New function.
9247 (vect_expand_fold_left): Likewise.
9248 (vectorize_fold_left_reduction): Likewise.
9249 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9250 scalar phi in place for it. Check for target support and reject
9251 cases that would reassociate the operation. Defer the transform
9252 phase to vectorize_fold_left_reduction.
9253 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9254 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9255 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9257 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9259 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9260 call to ifc_temp_var.
9262 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9263 Alan Hayward <alan.hayward@arm.com>
9264 David Sherwood <david.sherwood@arm.com>
9266 * target.def (legitimize_address_displacement): Take the original
9267 offset as a poly_int.
9268 * targhooks.h (default_legitimize_address_displacement): Update
9270 * targhooks.c (default_legitimize_address_displacement): Likewise.
9271 * doc/tm.texi: Regenerate.
9272 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9273 as an argument, moving assert of ad->disp == ad->disp_term to...
9274 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9275 Try calling targetm.legitimize_address_displacement before expanding
9276 the address rather than afterwards, and adjust for the new interface.
9277 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9278 Match the new hook interface. Handle SVE addresses.
9279 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9282 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9284 * Makefile.in (OBJS): Add early-remat.o.
9285 * target.def (select_early_remat_modes): New hook.
9286 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9287 * doc/tm.texi: Regenerate.
9288 * targhooks.h (default_select_early_remat_modes): Declare.
9289 * targhooks.c (default_select_early_remat_modes): New function.
9290 * timevar.def (TV_EARLY_REMAT): New timevar.
9291 * passes.def (pass_early_remat): New pass.
9292 * tree-pass.h (make_pass_early_remat): Declare.
9293 * early-remat.c: New file.
9294 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9296 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9298 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9299 Alan Hayward <alan.hayward@arm.com>
9300 David Sherwood <david.sherwood@arm.com>
9302 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9303 vfm1 with a bound_epilog parameter.
9304 (vect_do_peeling): Update calls accordingly, and move the prologue
9305 call earlier in the function. Treat the base bound_epilog as 0 for
9306 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9307 this base when peeling for gaps.
9308 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9309 with fully-masked loops.
9310 (vect_estimate_min_profitable_iters): Handle the single peeled
9311 iteration in that case.
9313 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9314 Alan Hayward <alan.hayward@arm.com>
9315 David Sherwood <david.sherwood@arm.com>
9317 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9318 single-element interleaving even if the size is not a power of 2.
9319 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9320 accesses for single-element interleaving if the group size is
9323 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9324 Alan Hayward <alan.hayward@arm.com>
9325 David Sherwood <david.sherwood@arm.com>
9327 * doc/md.texi (fold_extract_last_@var{m}): Document.
9328 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9329 * optabs.def (fold_extract_last_optab): New optab.
9330 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9331 * internal-fn.c (fold_extract_direct): New macro.
9332 (expand_fold_extract_optab_fn): Likewise.
9333 (direct_fold_extract_optab_supported_p): Likewise.
9334 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9335 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9336 EXTRACT_LAST_REDUCTION.
9337 (get_initial_def_for_reduction): Do not create an initial vector
9338 for EXTRACT_LAST_REDUCTION reductions.
9339 (vectorizable_reduction): Leave the scalar phi in place for
9340 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9341 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9342 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9343 transform phase to vectorizable_condition.
9344 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9346 (vect_finish_stmt_generation): ...here.
9347 (vect_finish_replace_stmt): New function.
9348 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9349 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9351 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9353 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9354 Alan Hayward <alan.hayward@arm.com>
9355 David Sherwood <david.sherwood@arm.com>
9357 * doc/md.texi (extract_last_@var{m}): Document.
9358 * optabs.def (extract_last_optab): New optab.
9359 * internal-fn.def (EXTRACT_LAST): New internal function.
9360 * internal-fn.c (cond_unary_direct): New macro.
9361 (expand_cond_unary_optab_fn): Likewise.
9362 (direct_cond_unary_optab_supported_p): Likewise.
9363 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9364 loops using EXTRACT_LAST.
9365 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9366 (extract_last_<mode>): ...this optab.
9367 (vec_extract<mode><Vel>): Update accordingly.
9369 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9370 Alan Hayward <alan.hayward@arm.com>
9371 David Sherwood <david.sherwood@arm.com>
9373 * target.def (empty_mask_is_expensive): New hook.
9374 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9375 * doc/tm.texi: Regenerate.
9376 * targhooks.h (default_empty_mask_is_expensive): Declare.
9377 * targhooks.c (default_empty_mask_is_expensive): New function.
9378 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9379 if the target says that empty masks are expensive.
9380 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9382 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9384 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9385 Alan Hayward <alan.hayward@arm.com>
9386 David Sherwood <david.sherwood@arm.com>
9388 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9389 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9390 (vect_use_loop_mask_for_alignment_p): New function.
9391 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9392 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9393 niters_skip argument. Make sure that the first niters_skip elements
9394 of the first iteration are inactive.
9395 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9396 Update call to vect_set_loop_masks_directly.
9397 (get_misalign_in_elems): New function, split out from...
9398 (vect_gen_prolog_loop_niters): ...here.
9399 (vect_update_init_of_dr): Take a code argument that specifies whether
9400 the adjustment should be added or subtracted.
9401 (vect_update_init_of_drs): Likewise.
9402 (vect_prepare_for_masked_peels): New function.
9403 (vect_do_peeling): Skip prologue peeling if we're using a mask
9404 instead. Update call to vect_update_inits_of_drs.
9405 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9407 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9408 alignment. Do not include the number of peeled iterations in
9409 the minimum threshold in that case.
9410 (vectorizable_induction): Adjust the start value down by
9411 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9412 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9413 Take the number of skipped iterations into account when calculating
9415 * tree-vect-stmts.c (vect_gen_while_not): New function.
9417 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9418 Alan Hayward <alan.hayward@arm.com>
9419 David Sherwood <david.sherwood@arm.com>
9421 * doc/sourcebuild.texi (vect_fully_masked): Document.
9422 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9424 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9426 (vect_analyze_loop_2): ...here. Don't check the vectorization
9427 factor against the number of loop iterations if the loop is
9430 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9431 Alan Hayward <alan.hayward@arm.com>
9432 David Sherwood <david.sherwood@arm.com>
9434 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9435 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9436 (dump_groups): Update accordingly.
9437 (iv_use::mem_type): New member variable.
9438 (address_p): New function.
9439 (record_use): Add a mem_type argument and initialize the new
9441 (record_group_use): Add a mem_type argument. Use address_p.
9442 Remove obsolete null checks of base_object. Update call to record_use.
9443 (find_interesting_uses_op): Update call to record_group_use.
9444 (find_interesting_uses_cond): Likewise.
9445 (find_interesting_uses_address): Likewise.
9446 (get_mem_type_for_internal_fn): New function.
9447 (find_address_like_use): Likewise.
9448 (find_interesting_uses_stmt): Try find_address_like_use before
9449 calling find_interesting_uses_op.
9450 (addr_offset_valid_p): Use the iv mem_type field as the type
9451 of the addressed memory.
9452 (add_autoinc_candidates): Likewise.
9453 (get_address_cost): Likewise.
9454 (split_small_address_groups_p): Use address_p.
9455 (split_address_groups): Likewise.
9456 (add_iv_candidate_for_use): Likewise.
9457 (autoinc_possible_for_pair): Likewise.
9458 (rewrite_groups): Likewise.
9459 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9460 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9461 (get_alias_ptr_type_for_ptr_address): New function.
9462 (rewrite_use_address): Rewrite address uses in calls that were
9463 identified by find_address_like_use.
9465 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9466 Alan Hayward <alan.hayward@arm.com>
9467 David Sherwood <david.sherwood@arm.com>
9469 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9471 * gimple-expr.h (is_gimple_addressable: Likewise.
9472 * gimple-expr.c (is_gimple_address): Likewise.
9473 * internal-fn.c (expand_call_mem_ref): New function.
9474 (expand_mask_load_optab_fn): Use it.
9475 (expand_mask_store_optab_fn): Likewise.
9477 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9478 Alan Hayward <alan.hayward@arm.com>
9479 David Sherwood <david.sherwood@arm.com>
9481 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9482 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9483 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9484 (cond_umax@var{mode}): Document.
9485 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9486 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9487 (cond_umin_optab, cond_umax_optab): New optabs.
9488 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9489 (COND_IOR, COND_XOR): New internal functions.
9490 * internal-fn.h (get_conditional_internal_fn): Declare.
9491 * internal-fn.c (cond_binary_direct): New macro.
9492 (expand_cond_binary_optab_fn): Likewise.
9493 (direct_cond_binary_optab_supported_p): Likewise.
9494 (get_conditional_internal_fn): New function.
9495 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9496 Cope with reduction statements that are vectorized as calls rather
9498 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9499 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9500 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9501 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9502 (UNSPEC_COND_EOR): New unspecs.
9503 (optab): Add mappings for them.
9504 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9505 (sve_int_op, sve_fp_op): New int attributes.
9507 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9508 Alan Hayward <alan.hayward@arm.com>
9509 David Sherwood <david.sherwood@arm.com>
9511 * optabs.def (while_ult_optab): New optab.
9512 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9513 * internal-fn.def (WHILE_ULT): New internal function.
9514 * internal-fn.h (direct_internal_fn_supported_p): New override
9515 that takes two types as argument.
9516 * internal-fn.c (while_direct): New macro.
9517 (expand_while_optab_fn): New function.
9518 (convert_optab_supported_p): Likewise.
9519 (direct_while_optab_supported_p): New macro.
9520 * wide-int.h (wi::udiv_ceil): New function.
9521 * tree-vectorizer.h (rgroup_masks): New structure.
9522 (vec_loop_masks): New typedef.
9523 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9525 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9526 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9527 (vect_max_vf): New function.
9528 (slpeel_make_loop_iterate_ntimes): Delete.
9529 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9530 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9531 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9532 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9533 internal-fn.h, stor-layout.h and optabs-query.h.
9534 (vect_set_loop_mask): New function.
9535 (add_preheader_seq): Likewise.
9536 (add_header_seq): Likewise.
9537 (interleave_supported_p): Likewise.
9538 (vect_maybe_permute_loop_masks): Likewise.
9539 (vect_set_loop_masks_directly): Likewise.
9540 (vect_set_loop_condition_masked): Likewise.
9541 (vect_set_loop_condition_unmasked): New function, split out from
9542 slpeel_make_loop_iterate_ntimes.
9543 (slpeel_make_loop_iterate_ntimes): Rename to..
9544 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9545 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9546 (vect_do_peeling): Update call accordingly.
9547 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9549 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9550 mask_compare_type, can_fully_mask_p and fully_masked_p.
9551 (release_vec_loop_masks): New function.
9552 (_loop_vec_info): Use it to free the loop masks.
9553 (can_produce_all_loop_masks_p): New function.
9554 (vect_get_max_nscalars_per_iter): Likewise.
9555 (vect_verify_full_masking): Likewise.
9556 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9557 retries, and free the mask rgroups before retrying. Check loop-wide
9558 reasons for disallowing fully-masked loops. Make the final decision
9559 about whether use a fully-masked loop or not.
9560 (vect_estimate_min_profitable_iters): Do not assume that peeling
9561 for the number of iterations will be needed for fully-masked loops.
9562 (vectorizable_reduction): Disable fully-masked loops.
9563 (vectorizable_live_operation): Likewise.
9564 (vect_halve_mask_nunits): New function.
9565 (vect_double_mask_nunits): Likewise.
9566 (vect_record_loop_mask): Likewise.
9567 (vect_get_loop_mask): Likewise.
9568 (vect_transform_loop): Handle the case in which the final loop
9569 iteration might handle a partial vector. Call vect_set_loop_condition
9570 instead of slpeel_make_loop_iterate_ntimes.
9571 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9572 (check_load_store_masking): New function.
9573 (prepare_load_store_mask): Likewise.
9574 (vectorizable_store): Handle fully-masked loops.
9575 (vectorizable_load): Likewise.
9576 (supportable_widening_operation): Use vect_halve_mask_nunits for
9578 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9579 (vect_gen_while): New function.
9580 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9581 (aarch64_uqdec<mode>): New insn.
9583 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9584 Alan Hayward <alan.hayward@arm.com>
9585 David Sherwood <david.sherwood@arm.com>
9587 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9588 (reduc_xor_scal_optab): New optabs.
9589 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9590 (reduc_xor_scal_@var{m}): Document.
9591 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9592 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9594 * fold-const-call.c (fold_const_call): Handle them.
9595 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9596 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9597 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9598 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9599 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9600 (UNSPEC_XORV): New unspecs.
9601 (optab): Add entries for them.
9602 (BITWISEV): New int iterator.
9603 (bit_reduc_op): New int attributes.
9605 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9606 Alan Hayward <alan.hayward@arm.com>
9607 David Sherwood <david.sherwood@arm.com>
9609 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9610 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9611 * optabs.def (vec_shl_insert_optab): New optab.
9612 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9613 (duplicate_and_interleave): Likewise.
9614 * tree-vect-loop.c: Include internal-fn.h.
9615 (neutral_op_for_slp_reduction): New function, split out from
9616 get_initial_defs_for_reduction.
9617 (get_initial_def_for_reduction): Handle option 2 for variable-length
9618 vectors by loading the neutral value into a vector and then shifting
9619 the initial value into element 0.
9620 (get_initial_defs_for_reduction): Replace the code argument with
9621 the neutral value calculated by neutral_op_for_slp_reduction.
9622 Use gimple_build_vector for constant-length vectors.
9623 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9624 but the first group_size elements have a neutral value.
9625 Use duplicate_and_interleave otherwise.
9626 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9627 Update call to get_initial_defs_for_reduction. Handle SLP
9628 reductions for variable-length vectors by creating one vector
9629 result for each scalar result, with the elements associated
9630 with other scalar results stubbed out with the neutral value.
9631 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9632 Require IFN_VEC_SHL_INSERT for double reductions on
9633 variable-length vectors, or SLP reductions that have
9634 a neutral value. Require can_duplicate_and_interleave_p
9635 support for variable-length unchained SLP reductions if there
9636 is no neutral value, such as for MIN/MAX reductions. Also require
9637 the number of vector elements to be a multiple of the number of
9638 SLP statements when doing variable-length unchained SLP reductions.
9639 Update call to vect_create_epilog_for_reduction.
9640 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9641 and remove initial values.
9642 (duplicate_and_interleave): Make public.
9643 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9644 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9646 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9647 Alan Hayward <alan.hayward@arm.com>
9648 David Sherwood <david.sherwood@arm.com>
9650 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9651 (can_duplicate_and_interleave_p): New function.
9652 (vect_get_and_check_slp_defs): Take the vector of statements
9653 rather than just the current one. Remove excess parentheses.
9654 Restriction rejectinon of vect_constant_def and vect_external_def
9655 for variable-length vectors to boolean types, or types for which
9656 can_duplicate_and_interleave_p is false.
9657 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9658 (duplicate_and_interleave): New function.
9659 (vect_get_constant_vectors): Use gimple_build_vector for
9660 constant-length vectors and suitable variable-length constant
9661 vectors. Use duplicate_and_interleave for other variable-length
9662 vectors. Don't defer the update when inserting new statements.
9664 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9665 Alan Hayward <alan.hayward@arm.com>
9666 David Sherwood <david.sherwood@arm.com>
9668 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9669 min_profitable_iters doesn't go negative.
9671 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9672 Alan Hayward <alan.hayward@arm.com>
9673 David Sherwood <david.sherwood@arm.com>
9675 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9676 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9677 * optabs.def (vec_mask_load_lanes_optab): New optab.
9678 (vec_mask_store_lanes_optab): Likewise.
9679 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9680 (MASK_STORE_LANES): Likewise.
9681 * internal-fn.c (mask_load_lanes_direct): New macro.
9682 (mask_store_lanes_direct): Likewise.
9683 (expand_mask_load_optab_fn): Handle masked operations.
9684 (expand_mask_load_lanes_optab_fn): New macro.
9685 (expand_mask_store_optab_fn): Handle masked operations.
9686 (expand_mask_store_lanes_optab_fn): New macro.
9687 (direct_mask_load_lanes_optab_supported_p): Likewise.
9688 (direct_mask_store_lanes_optab_supported_p): Likewise.
9689 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9691 (vect_load_lanes_supported): Likewise.
9692 * tree-vect-data-refs.c (strip_conversion): New function.
9693 (can_group_stmts_p): Likewise.
9694 (vect_analyze_data_ref_accesses): Use it instead of checking
9695 for a pair of assignments.
9696 (vect_store_lanes_supported): Take a masked_p parameter.
9697 (vect_load_lanes_supported): Likewise.
9698 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9699 vect_store_lanes_supported and vect_load_lanes_supported.
9700 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9701 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9702 parameter. Don't allow gaps for masked accesses.
9703 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9704 and vect_load_lanes_supported.
9705 (get_load_store_type): Take a masked_p parameter and update
9706 call to get_group_load_store_type.
9707 (vectorizable_store): Update call to get_load_store_type.
9708 Handle IFN_MASK_STORE_LANES.
9709 (vectorizable_load): Update call to get_load_store_type.
9710 Handle IFN_MASK_LOAD_LANES.
9712 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9713 Alan Hayward <alan.hayward@arm.com>
9714 David Sherwood <david.sherwood@arm.com>
9716 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9718 * config/aarch64/aarch64-protos.h
9719 (aarch64_sve_struct_memory_operand_p): Declare.
9720 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9721 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9722 (VPRED, vpred): Handle SVE structure modes.
9723 * config/aarch64/constraints.md (Utx): New constraint.
9724 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9725 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9726 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9727 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9728 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9729 structure modes. Split into pieces after RA.
9730 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9731 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9733 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9734 SVE structure modes.
9735 (aarch64_classify_address): Likewise.
9736 (sizetochar): Move earlier in file.
9737 (aarch64_print_operand): Handle SVE register lists.
9738 (aarch64_array_mode): New function.
9739 (aarch64_sve_struct_memory_operand_p): Likewise.
9740 (TARGET_ARRAY_MODE): Redefine.
9742 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9743 Alan Hayward <alan.hayward@arm.com>
9744 David Sherwood <david.sherwood@arm.com>
9746 * target.def (array_mode): New target hook.
9747 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9748 * doc/tm.texi: Regenerate.
9749 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9750 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9751 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9753 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9756 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9757 Alan Hayward <alan.hayward@arm.com>
9758 David Sherwood <david.sherwood@arm.com>
9760 * fold-const.c (fold_binary_loc): Check the argument types
9761 rather than the result type when testing for a vector operation.
9763 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9765 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9766 * doc/tm.texi: Regenerate.
9768 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9769 Alan Hayward <alan.hayward@arm.com>
9770 David Sherwood <david.sherwood@arm.com>
9772 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9773 (sve): Document new AArch64 extension.
9774 * doc/md.texi (w): Extend the description of the AArch64
9775 constraint to include SVE vectors.
9776 (Upl, Upa): Document new AArch64 predicate constraints.
9777 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9779 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9780 (msve-vector-bits=): New option.
9781 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9782 SVE when these are disabled.
9783 (sve): New extension.
9784 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9785 modes. Adjust their number of units based on aarch64_sve_vg.
9786 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9787 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9788 aarch64_addr_query_type.
9789 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9790 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9791 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9792 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9793 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9794 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9795 (aarch64_simd_imm_zero_p): Delete.
9796 (aarch64_check_zero_based_sve_index_immediate): Declare.
9797 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9798 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9799 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9800 (aarch64_sve_float_mul_immediate_p): Likewise.
9801 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9803 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9804 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9805 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9806 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9807 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9808 (aarch64_regmode_natural_size): Likewise.
9809 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9810 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9812 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9813 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9814 for VG and the SVE predicate registers.
9815 (V_ALIASES): Add a "z"-prefixed alias.
9816 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9817 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9818 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9819 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9820 (REG_CLASS_NAMES): Add entries for them.
9821 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9822 and the predicate registers.
9823 (aarch64_sve_vg): Declare.
9824 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9825 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9826 (REGMODE_NATURAL_SIZE): Define.
9827 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9829 * config/aarch64/aarch64.c: Include cfgrtl.h.
9830 (simd_immediate_info): Add a constructor for series vectors,
9831 and an associated step field.
9832 (aarch64_sve_vg): New variable.
9833 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9834 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9835 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9836 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9837 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9838 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9839 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9840 (aarch64_get_mask_mode): New functions.
9841 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9842 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9843 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9844 predicate modes and predicate registers. Explicitly restrict
9845 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9846 to store a vector mode if it is recognized by
9847 aarch64_classify_vector_mode.
9848 (aarch64_regmode_natural_size): New function.
9849 (aarch64_hard_regno_caller_save_mode): Return the original mode
9851 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9852 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9853 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9854 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9856 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9857 does not overlap dest if the function is frame-related. Handle
9859 (aarch64_split_add_offset): New function.
9860 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9861 them aarch64_add_offset.
9862 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9863 and update call to aarch64_sub_sp.
9864 (aarch64_add_cfa_expression): New function.
9865 (aarch64_expand_prologue): Pass extra temporary registers to the
9866 functions above. Handle the case in which we need to emit new
9867 DW_CFA_expressions for registers that were originally saved
9868 relative to the stack pointer, but now have to be expressed
9869 relative to the frame pointer.
9870 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9872 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9873 IP0 and IP1 values for SVE frames.
9874 (aarch64_expand_vec_series): New function.
9875 (aarch64_expand_sve_widened_duplicate): Likewise.
9876 (aarch64_expand_sve_const_vector): Likewise.
9877 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9878 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9879 into the register, rather than emitting a SET directly.
9880 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9881 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9882 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9883 (offset_9bit_signed_scaled_p): New functions.
9884 (aarch64_replicate_bitmask_imm): New function.
9885 (aarch64_bitmask_imm): Use it.
9886 (aarch64_cannot_force_const_mem): Reject expressions involving
9887 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9888 (aarch64_classify_index): Handle SVE indices, by requiring
9889 a plain register index with a scale that matches the element size.
9890 (aarch64_classify_address): Handle SVE addresses. Assert that
9891 the mode of the address is VOIDmode or an integer mode.
9892 Update call to aarch64_classify_symbol.
9893 (aarch64_classify_symbolic_expression): Update call to
9894 aarch64_classify_symbol.
9895 (aarch64_const_vec_all_in_range_p): New function.
9896 (aarch64_print_vector_float_operand): Likewise.
9897 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9898 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9899 and the FP immediates 1.0 and 0.5.
9900 (aarch64_print_address_internal): Handle SVE addresses.
9901 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9902 (aarch64_regno_regclass): Handle predicate registers.
9903 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9905 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9906 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9907 (aarch64_convert_sve_vector_bits): New function.
9908 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9909 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9911 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9912 Handle SVE vector and predicate modes. Accept VL-based constants
9913 that need only one temporary register, and VL offsets that require
9914 no temporary registers.
9915 (aarch64_conditional_register_usage): Mark the predicate registers
9916 as fixed if SVE isn't available.
9917 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9918 Return true for SVE vector and predicate modes.
9919 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9920 rather than an unsigned int. Handle SVE modes.
9921 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9923 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9925 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9926 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9927 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9928 (aarch64_sve_float_mul_immediate_p): New functions.
9929 (aarch64_sve_valid_immediate): New function.
9930 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9931 Explicitly reject structure modes. Check for INDEX constants.
9932 Handle PTRUE and PFALSE constants.
9933 (aarch64_check_zero_based_sve_index_immediate): New function.
9934 (aarch64_simd_imm_zero_p): Delete.
9935 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9936 vector modes. Accept constants in the range of CNT[BHWD].
9937 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9938 ask for an Advanced SIMD mode.
9939 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9940 (aarch64_simd_vector_alignment): Handle SVE predicates.
9941 (aarch64_vectorize_preferred_vector_alignment): New function.
9942 (aarch64_simd_vector_alignment_reachable): Use it instead of
9944 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9945 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9947 (MAX_VECT_LEN): Delete.
9948 (expand_vec_perm_d): Add a vec_flags field.
9949 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9950 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9951 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9953 (aarch64_evpc_rev): Rename to...
9954 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9955 (aarch64_evpc_rev_global): New function.
9956 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9957 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9959 (aarch64_evpc_sve_tbl): New function.
9960 (aarch64_expand_vec_perm_const_1): Update after rename of
9961 aarch64_evpc_rev. Handle SVE permutes too, trying
9962 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9963 than aarch64_evpc_tbl.
9964 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9965 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9966 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9967 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9968 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9969 (aarch64_expand_sve_vcond): New functions.
9970 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9971 of aarch64_vector_mode_p.
9972 (aarch64_dwarf_poly_indeterminate_value): New function.
9973 (aarch64_compute_pressure_classes): Likewise.
9974 (aarch64_can_change_mode_class): Likewise.
9975 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9976 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9977 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9978 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9979 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9980 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9981 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9982 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9984 (Dn, Dl, Dr): Accept const as well as const_vector.
9985 (Dz): Likewise. Compare against CONST0_RTX.
9986 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9987 of "vector" where appropriate.
9988 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9989 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9990 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9991 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9992 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9993 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9994 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9995 (v_int_equiv): Extend to SVE modes.
9996 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9998 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9999 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10000 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10001 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10002 (SVE_COND_FP_CMP): New int iterators.
10003 (perm_hilo): Handle the new unpack unspecs.
10004 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10006 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10007 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10008 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10009 (aarch64_equality_operator, aarch64_constant_vector_operand)
10010 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10011 (aarch64_sve_nonimmediate_operand): Likewise.
10012 (aarch64_sve_general_operand): Likewise.
10013 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10014 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10015 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10016 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10017 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10018 (aarch64_sve_float_arith_immediate): Likewise.
10019 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10020 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10021 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10022 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10023 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10024 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10025 (aarch64_sve_float_arith_operand): Likewise.
10026 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10027 (aarch64_sve_float_mul_operand): Likewise.
10028 (aarch64_sve_vec_perm_operand): Likewise.
10029 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10030 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10031 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10032 as well as const_vector.
10033 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10034 in file. Use CONST0_RTX and CONSTM1_RTX.
10035 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10036 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10037 Use aarch64_simd_imm_zero.
10038 * config/aarch64/aarch64-sve.md: New file.
10039 * config/aarch64/aarch64.md: Include it.
10040 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10041 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10042 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10043 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10044 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10045 (sve): New attribute.
10046 (enabled): Disable instructions with the sve attribute unless
10048 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10049 aarch64_expand_mov_immediate.
10050 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10051 CNT[BHSD] immediates.
10052 (movti): Split CONST_POLY_INT moves into two halves.
10053 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10054 Split additions that need a temporary here if the destination
10055 is the stack pointer.
10056 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10057 (*add<mode>3_poly_1): New instruction.
10058 (set_clobber_cc): New expander.
10060 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10062 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10063 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10064 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10065 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10066 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10067 Change innermode from fixed_mode_size to machine_mode.
10068 (simplify_subreg): Update call accordingly. Handle a constant-sized
10069 subreg of a variable-length CONST_VECTOR.
10071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10072 Alan Hayward <alan.hayward@arm.com>
10073 David Sherwood <david.sherwood@arm.com>
10075 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10076 (add_offset_to_base): New function, split out from...
10077 (create_mem_ref): ...here. When handling a scale other than 1,
10078 check first whether the address is valid without the offset.
10079 Add it into the base if so, leaving the index and scale as-is.
10081 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10084 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10085 fold_for_warn before checking if arg2 is INTEGER_CST.
10087 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10089 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10090 (store_multiple_operation): Delete.
10091 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10092 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10093 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10094 guarded by TARGET_STRING.
10095 (rs6000_output_load_multiple): Delete.
10096 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10097 OPTION_MASK_STRING / TARGET_STRING handling.
10098 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10099 (const rs6000_opt_masks) <"string">: Change mask to 0.
10100 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10101 (MASK_STRING): Delete.
10102 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10104 (load_multiple): Delete.
10111 (store_multiple): Delete.
10118 (movmemsi_8reg): Delete.
10119 (corresponding unnamed define_insn): Delete.
10120 (movmemsi_6reg): Delete.
10121 (corresponding unnamed define_insn): Delete.
10122 (movmemsi_4reg): Delete.
10123 (corresponding unnamed define_insn): Delete.
10124 (movmemsi_2reg): Delete.
10125 (corresponding unnamed define_insn): Delete.
10126 (movmemsi_1reg): Delete.
10127 (corresponding unnamed define_insn): Delete.
10128 * config/rs6000/rs6000.opt (mno-string): New.
10129 (mstring): Replace by deprecation warning stub.
10130 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10132 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10134 * regrename.c (regrename_do_replace): If replacing the same
10135 reg multiple times, try to reuse last created gen_raw_REG.
10138 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10139 main to workaround a bug in GDB.
10141 2018-01-12 Tom de Vries <tom@codesourcery.com>
10144 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10146 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10148 PR rtl-optimization/80481
10149 * ira-color.c (get_cap_member): New function.
10150 (allocnos_conflict_by_live_ranges_p): Use it.
10151 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10152 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10154 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10157 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10158 (*saddl_se_1): Ditto.
10159 (*ssubsi_1): Ditto.
10160 (*ssubl_se_1): Ditto.
10162 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10164 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10165 rather than wi::to_widest for DR_INITs.
10166 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10167 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10168 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10170 (vect_analyze_group_access_1): Note that here.
10172 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10174 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10175 polynomial type sizes.
10177 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10179 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10180 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10181 (gimple_add_tmp_var): Likewise.
10183 2018-01-12 Martin Liska <mliska@suse.cz>
10185 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10186 (gimple_alloc_sizes): Likewise.
10187 (dump_gimple_statistics): Use PRIu64 in printf format.
10188 * gimple.h: Change uint64_t to int.
10190 2018-01-12 Martin Liska <mliska@suse.cz>
10192 * tree-core.h: Use uint64_t instead of int.
10193 * tree.c (tree_node_counts): Likewise.
10194 (tree_node_sizes): Likewise.
10195 (dump_tree_statistics): Use PRIu64 in printf format.
10197 2018-01-12 Martin Liska <mliska@suse.cz>
10199 * Makefile.in: As qsort_chk is implemented in vec.c, add
10200 vec.o to linkage of gencfn-macros.
10201 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10202 passing the info to record_node_allocation_statistics.
10203 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10205 * ggc-common.c (struct ggc_usage): Add operator== and use
10206 it in operator< and compare function.
10207 * mem-stats.h (struct mem_usage): Likewise.
10208 * vec.c (struct vec_usage): Remove operator< and compare
10209 function. Can be simply inherited.
10211 2018-01-12 Martin Jambor <mjambor@suse.cz>
10214 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10215 * tree-ssa-math-opts.c: Include domwalk.h.
10216 (convert_mult_to_fma_1): New function.
10217 (fma_transformation_info): New type.
10218 (fma_deferring_state): Likewise.
10219 (cancel_fma_deferring): New function.
10220 (result_of_phi): Likewise.
10221 (last_fma_candidate_feeds_initial_phi): Likewise.
10222 (convert_mult_to_fma): Added deferring logic, split actual
10223 transformation to convert_mult_to_fma_1.
10224 (math_opts_dom_walker): New type.
10225 (math_opts_dom_walker::after_dom_children): New method, body moved
10226 here from pass_optimize_widening_mul::execute, added deferring logic
10228 (pass_optimize_widening_mul::execute): Moved most of code to
10229 math_opts_dom_walker::after_dom_children.
10230 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10231 * config/i386/i386.c (ix86_option_override_internal): Added
10232 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10234 2018-01-12 Richard Biener <rguenther@suse.de>
10237 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10238 inline instance vars.
10240 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10243 * config/rx/rx.c (rx_is_restricted_memory_address):
10244 Handle SUBREG case.
10246 2018-01-12 Richard Biener <rguenther@suse.de>
10248 PR tree-optimization/80846
10249 * target.def (split_reduction): New target hook.
10250 * targhooks.c (default_split_reduction): New function.
10251 * targhooks.h (default_split_reduction): Declare.
10252 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10253 target requests first reduce vectors by combining low and high
10255 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10256 (get_vectype_for_scalar_type_and_size): Export.
10257 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10258 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10259 * doc/tm.texi: Regenerate.
10260 * config/i386/i386.c (ix86_split_reduction): Implement
10261 TARGET_VECTORIZE_SPLIT_REDUCTION.
10263 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10266 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10267 in PIC mode except for TARGET_VXWORKS_RTP.
10268 * config/sparc/sparc.c: Include cfgrtl.h.
10269 (TARGET_INIT_PIC_REG): Define.
10270 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10271 (sparc_pic_register_p): New predicate.
10272 (sparc_legitimate_address_p): Use it.
10273 (sparc_legitimize_pic_address): Likewise.
10274 (sparc_delegitimize_address): Likewise.
10275 (sparc_mode_dependent_address_p): Likewise.
10276 (gen_load_pcrel_sym): Remove 4th parameter.
10277 (load_got_register): Adjust call to above. Remove obsolete stuff.
10278 (sparc_expand_prologue): Do not call load_got_register here.
10279 (sparc_flat_expand_prologue): Likewise.
10280 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10281 (sparc_use_pseudo_pic_reg): New function.
10282 (sparc_init_pic_reg): Likewise.
10283 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10284 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10286 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10288 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10289 Add item for branch_cost.
10291 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10293 PR rtl-optimization/83565
10294 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10295 not extend the result to a larger mode for rotate operations.
10296 (num_sign_bit_copies1): Likewise.
10298 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10301 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10303 Use values-Xc.o for -pedantic.
10304 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10306 2018-01-12 Martin Liska <mliska@suse.cz>
10309 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10311 (possible_polymorphic_call_targets): Use it.
10312 (ipa_devirt): Likewise.
10314 2018-01-12 Martin Liska <mliska@suse.cz>
10316 * profile-count.h (enum profile_quality): Use 0 as invalid
10317 enum value of profile_quality.
10319 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10321 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10322 -mext-string options.
10324 2018-01-12 Richard Biener <rguenther@suse.de>
10326 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10327 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10328 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10330 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10332 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10334 * configure.ac (--with-long-double-format): Add support for the
10335 configuration option to change the default long double format on
10337 * config.gcc (powerpc*-linux*-*): Likewise.
10338 * configure: Regenerate.
10339 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10340 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10341 used without modification.
10343 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10345 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10346 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10347 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10348 MISC_BUILTIN_SPEC_BARRIER.
10349 (rs6000_init_builtins): Likewise.
10350 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10352 (speculation_barrier): New define_insn.
10353 * doc/extend.texi: Document __builtin_speculation_barrier.
10355 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10358 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10359 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10360 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10362 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10363 integral modes instead of "ss" and "sd".
10364 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10365 vectors with 32-bit and 64-bit elements.
10366 (vecdupssescalarmodesuffix): New mode attribute.
10367 (vec_dup<mode>): Use it.
10369 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10372 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10373 frame if argument is passed on stack.
10375 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10378 * ree.c (combine_reaching_defs): Optimize also
10379 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10380 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10382 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10384 PR middle-end/83189
10385 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10387 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10389 PR middle-end/83718
10390 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10391 after they are computed.
10393 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10395 PR tree-optimization/83695
10396 * gimple-loop-linterchange.cc
10397 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10398 reset cached scev information after interchange.
10399 (pass_linterchange::execute): Remove call to scev_reset_htab.
10401 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10403 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10404 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10405 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10406 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10407 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10408 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10409 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10410 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10411 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10412 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10413 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10414 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10415 (V_lane_reg): Likewise.
10416 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10418 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10419 (vfmal_lane_low<mode>_intrinsic,
10420 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10421 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10422 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10423 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10424 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10425 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10427 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10429 * config/arm/arm-cpus.in (fp16fml): New feature.
10430 (ALL_SIMD): Add fp16fml.
10431 (armv8.2-a): Add fp16fml as an option.
10432 (armv8.3-a): Likewise.
10433 (armv8.4-a): Add fp16fml as part of fp16.
10434 * config/arm/arm.h (TARGET_FP16FML): Define.
10435 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10437 * config/arm/arm-modes.def (V2HF): Define.
10438 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10439 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10440 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10441 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10442 vfmsl_low, vfmsl_high): New set of builtins.
10443 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10444 (vfml_op): New code attribute.
10445 (VFMLHALVES): New int iterator.
10446 (VFML, VFMLSEL): New mode attributes.
10447 (V_reg): Define mapping for V2HF.
10448 (V_hi, V_lo): New mode attributes.
10449 (VF_constraint): Likewise.
10450 (vfml_half, vfml_half_selector): New int attributes.
10451 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10453 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10454 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10456 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10457 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10458 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10459 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10461 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10462 Document new effective target and option set.
10464 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10466 * config/arm/arm-cpus.in (armv8_4): New feature.
10467 (ARMv8_4a): New fgroup.
10468 (armv8.4-a): New arch.
10469 * config/arm/arm-tables.opt: Regenerate.
10470 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10471 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10472 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10473 Add matching rules for -march=armv8.4-a and extensions.
10474 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10476 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10479 * config/rx/rx.md (BW): New mode attribute.
10480 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10482 2018-01-11 Richard Biener <rguenther@suse.de>
10484 PR tree-optimization/83435
10485 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10486 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10487 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10489 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10490 Alan Hayward <alan.hayward@arm.com>
10491 David Sherwood <david.sherwood@arm.com>
10493 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10495 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10496 (aarch64_print_address_internal): Use it to check for a zero offset.
10498 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10499 Alan Hayward <alan.hayward@arm.com>
10500 David Sherwood <david.sherwood@arm.com>
10502 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10503 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10504 Return a poly_int64 rather than a HOST_WIDE_INT.
10505 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10506 rather than a HOST_WIDE_INT.
10507 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10508 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10509 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10510 final_offset from HOST_WIDE_INT to poly_int64.
10511 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10512 to_constant when getting the number of units in an Advanced SIMD
10514 (aarch64_builtin_vectorized_function): Check for a constant number
10516 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10518 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10519 attribute instead of GET_MODE_NUNITS.
10520 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10521 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10522 GET_MODE_SIZE for fixed-size registers.
10523 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10524 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10525 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10526 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10527 (aarch64_print_operand, aarch64_print_address_internal)
10528 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10529 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10530 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10531 Handle polynomial GET_MODE_SIZE.
10532 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10533 wider than SImode without modification.
10534 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10535 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10536 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10537 passing and returning SVE modes.
10538 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10539 rather than GEN_INT.
10540 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10541 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10542 (aarch64_allocate_and_probe_stack_space): Likewise.
10543 (aarch64_layout_frame): Cope with polynomial offsets.
10544 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10545 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10546 polynomial offsets.
10547 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10548 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10549 poly_int64 rather than a HOST_WIDE_INT.
10550 (aarch64_get_separate_components, aarch64_process_components)
10551 (aarch64_expand_prologue, aarch64_expand_epilogue)
10552 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10553 (aarch64_anchor_offset): New function, split out from...
10554 (aarch64_legitimize_address): ...here.
10555 (aarch64_builtin_vectorization_cost): Handle polynomial
10556 TYPE_VECTOR_SUBPARTS.
10557 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10559 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10560 number of elements from the PARALLEL rather than the mode.
10561 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10562 rather than GET_MODE_BITSIZE.
10563 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10564 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10565 (aarch64_expand_vec_perm_const_1): Handle polynomial
10566 d->perm.length () and d->perm elements.
10567 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10568 Apply to_constant to d->perm elements.
10569 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10570 polynomial CONST_VECTOR_NUNITS.
10571 (aarch64_move_pointer): Take amount as a poly_int64 rather
10573 (aarch64_progress_pointer): Avoid temporary variable.
10574 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10575 the mode attribute instead of GET_MODE.
10577 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10578 Alan Hayward <alan.hayward@arm.com>
10579 David Sherwood <david.sherwood@arm.com>
10581 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10582 x exists before using it.
10583 (aarch64_add_constant_internal): Rename to...
10584 (aarch64_add_offset_1): ...this. Replace regnum with separate
10585 src and dest rtxes. Handle the case in which they're different,
10586 including when the offset is zero. Replace scratchreg with an rtx.
10587 Use 2 additions if there is no spare register into which we can
10588 move a 16-bit constant.
10589 (aarch64_add_constant): Delete.
10590 (aarch64_add_offset): Replace reg with separate src and dest
10591 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10592 Use aarch64_add_offset_1.
10593 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10594 an rtx rather than an int. Take the delta as a poly_int64
10595 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10596 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10597 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10598 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10599 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10600 and aarch64_add_sp.
10601 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10602 aarch64_add_constant.
10604 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10606 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10607 Use scalar_float_mode.
10609 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10611 * config/aarch64/aarch64-simd.md
10612 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10613 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10614 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10615 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10616 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10617 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10618 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10619 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10620 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10621 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10623 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10626 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10627 targ_options->x_arm_arch_string is non NULL.
10629 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10631 * config/aarch64/aarch64.h
10632 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10634 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10637 * expmed.c (emit_store_flag_force): Swap if const op0
10638 and change VOIDmode to mode of op0.
10640 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10642 PR rtl-optimization/83761
10643 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10644 than bytes to mode_for_size.
10646 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10648 PR middle-end/83189
10649 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10650 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10653 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10655 PR middle-end/83575
10656 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10657 when in layout mode.
10658 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10659 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10662 2018-01-10 Michael Collison <michael.collison@arm.com>
10664 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10665 * config/aarch64/aarch64-option-extension.def: Add
10666 AARCH64_OPT_EXTENSION of 'fp16fml'.
10667 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10668 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10669 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10670 * config/aarch64/constraints.md (Ui7): New constraint.
10671 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10672 (VFMLA_SEL_W): Ditto.
10675 (VFMLA16_LOW): New int iterator.
10676 (VFMLA16_HIGH): Ditto.
10677 (UNSPEC_FMLAL): New unspec.
10678 (UNSPEC_FMLSL): Ditto.
10679 (UNSPEC_FMLAL2): Ditto.
10680 (UNSPEC_FMLSL2): Ditto.
10681 (f16mac): New code attribute.
10682 * config/aarch64/aarch64-simd-builtins.def
10683 (aarch64_fmlal_lowv2sf): Ditto.
10684 (aarch64_fmlsl_lowv2sf): Ditto.
10685 (aarch64_fmlalq_lowv4sf): Ditto.
10686 (aarch64_fmlslq_lowv4sf): Ditto.
10687 (aarch64_fmlal_highv2sf): Ditto.
10688 (aarch64_fmlsl_highv2sf): Ditto.
10689 (aarch64_fmlalq_highv4sf): Ditto.
10690 (aarch64_fmlslq_highv4sf): Ditto.
10691 (aarch64_fmlal_lane_lowv2sf): Ditto.
10692 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10693 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10694 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10695 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10696 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10697 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10698 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10699 (aarch64_fmlal_lane_highv2sf): Ditto.
10700 (aarch64_fmlsl_lane_highv2sf): Ditto.
10701 (aarch64_fmlal_laneq_highv2sf): Ditto.
10702 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10703 (aarch64_fmlalq_lane_highv4sf): Ditto.
10704 (aarch64_fmlsl_lane_highv4sf): Ditto.
10705 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10706 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10707 * config/aarch64/aarch64-simd.md:
10708 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10709 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10710 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10711 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10712 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10713 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10714 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10715 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10716 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10717 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10718 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10719 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10720 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10721 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10722 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10723 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10724 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10725 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10726 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10727 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10728 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10729 (vfmlsl_low_u32): Ditto.
10730 (vfmlalq_low_u32): Ditto.
10731 (vfmlslq_low_u32): Ditto.
10732 (vfmlal_high_u32): Ditto.
10733 (vfmlsl_high_u32): Ditto.
10734 (vfmlalq_high_u32): Ditto.
10735 (vfmlslq_high_u32): Ditto.
10736 (vfmlal_lane_low_u32): Ditto.
10737 (vfmlsl_lane_low_u32): Ditto.
10738 (vfmlal_laneq_low_u32): Ditto.
10739 (vfmlsl_laneq_low_u32): Ditto.
10740 (vfmlalq_lane_low_u32): Ditto.
10741 (vfmlslq_lane_low_u32): Ditto.
10742 (vfmlalq_laneq_low_u32): Ditto.
10743 (vfmlslq_laneq_low_u32): Ditto.
10744 (vfmlal_lane_high_u32): Ditto.
10745 (vfmlsl_lane_high_u32): Ditto.
10746 (vfmlal_laneq_high_u32): Ditto.
10747 (vfmlsl_laneq_high_u32): Ditto.
10748 (vfmlalq_lane_high_u32): Ditto.
10749 (vfmlslq_lane_high_u32): Ditto.
10750 (vfmlalq_laneq_high_u32): Ditto.
10751 (vfmlslq_laneq_high_u32): Ditto.
10752 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10753 (AARCH64_FL_FOR_ARCH8_4): New.
10754 (AARCH64_ISA_F16FML): New ISA flag.
10755 (TARGET_F16FML): New feature flag for fp16fml.
10756 (doc/invoke.texi): Document new fp16fml option.
10758 2018-01-10 Michael Collison <michael.collison@arm.com>
10760 * config/aarch64/aarch64-builtins.c:
10761 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10762 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10763 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10764 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10765 (AARCH64_ISA_SHA3): New ISA flag.
10766 (TARGET_SHA3): New feature flag for sha3.
10767 * config/aarch64/iterators.md (sha512_op): New int attribute.
10768 (CRYPTO_SHA512): New int iterator.
10769 (UNSPEC_SHA512H): New unspec.
10770 (UNSPEC_SHA512H2): Ditto.
10771 (UNSPEC_SHA512SU0): Ditto.
10772 (UNSPEC_SHA512SU1): Ditto.
10773 * config/aarch64/aarch64-simd-builtins.def
10774 (aarch64_crypto_sha512hqv2di): New builtin.
10775 (aarch64_crypto_sha512h2qv2di): Ditto.
10776 (aarch64_crypto_sha512su0qv2di): Ditto.
10777 (aarch64_crypto_sha512su1qv2di): Ditto.
10778 (aarch64_eor3qv8hi): Ditto.
10779 (aarch64_rax1qv2di): Ditto.
10780 (aarch64_xarqv2di): Ditto.
10781 (aarch64_bcaxqv8hi): Ditto.
10782 * config/aarch64/aarch64-simd.md:
10783 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10784 (aarch64_crypto_sha512su0qv2di): Ditto.
10785 (aarch64_crypto_sha512su1qv2di): Ditto.
10786 (aarch64_eor3qv8hi): Ditto.
10787 (aarch64_rax1qv2di): Ditto.
10788 (aarch64_xarqv2di): Ditto.
10789 (aarch64_bcaxqv8hi): Ditto.
10790 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10791 (vsha512h2q_u64): Ditto.
10792 (vsha512su0q_u64): Ditto.
10793 (vsha512su1q_u64): Ditto.
10794 (veor3q_u16): Ditto.
10795 (vrax1q_u64): Ditto.
10796 (vxarq_u64): Ditto.
10797 (vbcaxq_u16): Ditto.
10798 * config/arm/types.md (crypto_sha512): New type attribute.
10799 (crypto_sha3): Ditto.
10800 (doc/invoke.texi): Document new sha3 option.
10802 2018-01-10 Michael Collison <michael.collison@arm.com>
10804 * config/aarch64/aarch64-builtins.c:
10805 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10806 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10807 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10808 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10809 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10810 (AARCH64_ISA_SM4): New ISA flag.
10811 (TARGET_SM4): New feature flag for sm4.
10812 * config/aarch64/aarch64-simd-builtins.def
10813 (aarch64_sm3ss1qv4si): Ditto.
10814 (aarch64_sm3tt1aq4si): Ditto.
10815 (aarch64_sm3tt1bq4si): Ditto.
10816 (aarch64_sm3tt2aq4si): Ditto.
10817 (aarch64_sm3tt2bq4si): Ditto.
10818 (aarch64_sm3partw1qv4si): Ditto.
10819 (aarch64_sm3partw2qv4si): Ditto.
10820 (aarch64_sm4eqv4si): Ditto.
10821 (aarch64_sm4ekeyqv4si): Ditto.
10822 * config/aarch64/aarch64-simd.md:
10823 (aarch64_sm3ss1qv4si): Ditto.
10824 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10825 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10826 (aarch64_sm4eqv4si): Ditto.
10827 (aarch64_sm4ekeyqv4si): Ditto.
10828 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10829 (sm3part_op): Ditto.
10830 (CRYPTO_SM3TT): Ditto.
10831 (CRYPTO_SM3PART): Ditto.
10832 (UNSPEC_SM3SS1): New unspec.
10833 (UNSPEC_SM3TT1A): Ditto.
10834 (UNSPEC_SM3TT1B): Ditto.
10835 (UNSPEC_SM3TT2A): Ditto.
10836 (UNSPEC_SM3TT2B): Ditto.
10837 (UNSPEC_SM3PARTW1): Ditto.
10838 (UNSPEC_SM3PARTW2): Ditto.
10839 (UNSPEC_SM4E): Ditto.
10840 (UNSPEC_SM4EKEY): Ditto.
10841 * config/aarch64/constraints.md (Ui2): New constraint.
10842 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10843 * config/arm/types.md (crypto_sm3): New type attribute.
10844 (crypto_sm4): Ditto.
10845 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10846 (vsm3tt1aq_u32): Ditto.
10847 (vsm3tt1bq_u32): Ditto.
10848 (vsm3tt2aq_u32): Ditto.
10849 (vsm3tt2bq_u32): Ditto.
10850 (vsm3partw1q_u32): Ditto.
10851 (vsm3partw2q_u32): Ditto.
10852 (vsm4eq_u32): Ditto.
10853 (vsm4ekeyq_u32): Ditto.
10854 (doc/invoke.texi): Document new sm4 option.
10856 2018-01-10 Michael Collison <michael.collison@arm.com>
10858 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10859 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10860 (AARCH64_FL_FOR_ARCH8_4): New.
10861 (AARCH64_FL_V8_4): New flag.
10862 (doc/invoke.texi): Document new armv8.4-a option.
10864 2018-01-10 Michael Collison <michael.collison@arm.com>
10866 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10867 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10868 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10869 * config/aarch64/aarch64-option-extension.def: Add
10870 AARCH64_OPT_EXTENSION of 'sha2'.
10871 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10872 (crypto): Disable sha2 and aes if crypto disabled.
10873 (crypto): Enable aes and sha2 if enabled.
10874 (simd): Disable sha2 and aes if simd disabled.
10875 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10877 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10878 (TARGET_SHA2): New feature flag for sha2.
10879 (TARGET_AES): New feature flag for aes.
10880 * config/aarch64/aarch64-simd.md:
10881 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10882 conditional on TARGET_AES.
10883 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10884 (aarch64_crypto_sha1hsi): Make pattern conditional
10886 (aarch64_crypto_sha1hv4si): Ditto.
10887 (aarch64_be_crypto_sha1hv4si): Ditto.
10888 (aarch64_crypto_sha1su1v4si): Ditto.
10889 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10890 (aarch64_crypto_sha1su0v4si): Ditto.
10891 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10892 (aarch64_crypto_sha256su0v4si): Ditto.
10893 (aarch64_crypto_sha256su1v4si): Ditto.
10894 (doc/invoke.texi): Document new aes and sha2 options.
10896 2018-01-10 Martin Sebor <msebor@redhat.com>
10898 PR tree-optimization/83781
10899 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10902 2018-01-11 Martin Sebor <msebor@gmail.com>
10903 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10905 PR tree-optimization/83501
10906 PR tree-optimization/81703
10908 * tree-ssa-strlen.c (get_string_cst): Rename...
10909 (get_string_len): ...to this. Handle global constants.
10910 (handle_char_store): Adjust.
10912 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10913 Jim Wilson <jimw@sifive.com>
10915 * config/riscv/riscv-protos.h (riscv_output_return): New.
10916 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10917 (riscv_attribute_table, riscv_output_return),
10918 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10919 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10920 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10922 (riscv_expand_prologue): Add early return for naked function.
10923 (riscv_expand_epilogue): Likewise.
10924 (riscv_function_ok_for_sibcall): Return false for naked function.
10925 (riscv_set_current_function): New.
10926 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10927 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10928 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10929 * doc/extend.texi (RISC-V Function Attributes): New.
10931 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10933 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10934 check for 128-bit long double before checking TCmode.
10935 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10936 128-bit long doubles before checking TFmode or TCmode.
10937 (FLOAT128_IBM_P): Likewise.
10939 2018-01-10 Martin Sebor <msebor@redhat.com>
10941 PR tree-optimization/83671
10942 * builtins.c (c_strlen): Unconditionally return zero for the empty
10944 Use -Warray-bounds for warnings.
10945 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10946 for non-constant array indices with COMPONENT_REF, arrays of
10947 arrays, and pointers to arrays.
10948 (gimple_fold_builtin_strlen): Determine and set length range for
10949 non-constant character arrays.
10951 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10953 PR middle-end/81897
10954 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10957 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10959 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10961 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10964 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10965 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10966 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10967 indexed_or_indirect_operand predicate.
10968 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10969 (*vsx_le_perm_load_v8hi): Likewise.
10970 (*vsx_le_perm_load_v16qi): Likewise.
10971 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10972 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10973 (*vsx_le_perm_store_v8hi): Likewise.
10974 (*vsx_le_perm_store_v16qi): Likewise.
10975 (eight unnamed splitters): Likewise.
10977 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10979 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10980 * config/rs6000/emmintrin.h: Likewise.
10981 * config/rs6000/mmintrin.h: Likewise.
10982 * config/rs6000/xmmintrin.h: Likewise.
10984 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10987 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10989 * tree.c (tree_nop_conversion): Return true for location wrapper
10991 (maybe_wrap_with_location): New function.
10992 (selftest::check_strip_nops): New function.
10993 (selftest::test_location_wrappers): New function.
10994 (selftest::tree_c_tests): Call it.
10995 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10996 (maybe_wrap_with_location): New decl.
10997 (EXPR_LOCATION_WRAPPER_P): New macro.
10998 (location_wrapper_p): New inline function.
10999 (tree_strip_any_location_wrapper): New inline function.
11001 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11004 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11005 stack_realign_offset for the largest alignment of stack slot
11007 (ix86_find_max_used_stack_alignment): New function.
11008 (ix86_finalize_stack_frame_flags): Use it. Set
11009 max_used_stack_alignment if we don't realign stack.
11010 * config/i386/i386.h (machine_function): Add
11011 max_used_stack_alignment.
11013 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11015 * config/arm/arm.opt (-mbranch-cost): New option.
11016 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11019 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11022 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11023 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11025 2018-01-10 Richard Biener <rguenther@suse.de>
11028 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11029 early out so it also covers the case where we have a non-NULL
11032 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11034 PR tree-optimization/83753
11035 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11036 for non-strided grouped accesses if the number of elements is 1.
11038 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11041 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11042 * i386.h (TARGET_USE_GATHER): Define.
11043 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11045 2018-01-10 Martin Liska <mliska@suse.cz>
11048 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11049 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11051 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11052 CLEANUP_NO_PARTITIONING is not set.
11054 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11056 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11057 for vectors, as a partial revert of r254296.
11058 * rtl.h (const_vec_p): Delete.
11059 (const_vec_duplicate_p): Don't test for vector CONSTs.
11060 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11061 * expmed.c (make_tree): Likewise.
11064 * common.md (E, F): Use CONSTANT_P instead of checking for
11066 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11067 checking for CONST_VECTOR.
11069 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11071 PR middle-end/83575
11072 * predict.c (force_edge_cold): Handle in more sane way edges
11073 with no prediction.
11075 2018-01-09 Carl Love <cel@us.ibm.com>
11077 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11079 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11080 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11081 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11082 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11083 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11084 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11085 * config/rs6000/rs6000-protos.h: Add extern defition for
11086 rs6000_generate_float2_double_code.
11087 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11089 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11090 (float2_v2df): Add define_expand.
11092 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11095 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11096 op_mode in the force_to_mode call.
11098 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11100 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11101 instead of checking each element individually.
11102 (aarch64_evpc_uzp): Likewise.
11103 (aarch64_evpc_zip): Likewise.
11104 (aarch64_evpc_ext): Likewise.
11105 (aarch64_evpc_rev): Likewise.
11106 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11107 instead of checking each element individually. Return true without
11109 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11110 whether all selected elements come from the same input, instead of
11111 checking each element individually. Remove calls to gen_rtx_REG,
11112 start_sequence and end_sequence and instead assert that no rtl is
11115 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11117 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11118 order of HIGH and CONST checks.
11120 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11122 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11123 if the destination isn't an SSA_NAME.
11125 2018-01-09 Richard Biener <rguenther@suse.de>
11127 PR tree-optimization/83668
11128 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11130 (canonicalize_loop_form): ... here, renamed from ...
11131 (canonicalize_loop_closed_ssa_form): ... this and amended to
11132 swap successor edges for loop exit blocks to make us use
11133 the RPO order we need for initial schedule generation.
11135 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11137 PR tree-optimization/64811
11138 * match.pd: When optimizing comparisons with Inf, avoid
11139 introducing or losing exceptions from comparisons with NaN.
11141 2018-01-09 Martin Liska <mliska@suse.cz>
11144 * asan.c (shadow_mem_size): Add gcc_assert.
11146 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11148 Don't save registers in main().
11151 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11152 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11153 * config/avr/avr.c (avr_set_current_function): Don't error if
11154 naked, OS_task or OS_main are specified at the same time.
11155 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11157 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11159 * common/config/avr/avr-common.c (avr_option_optimization_table):
11160 Switch on -mmain-is-OS_task for optimizing compilations.
11162 2018-01-09 Richard Biener <rguenther@suse.de>
11164 PR tree-optimization/83572
11165 * graphite.c: Include cfganal.h.
11166 (graphite_transform_loops): Connect infinite loops to exit
11167 and remove fake edges at the end.
11169 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11171 * ipa-inline.c (edge_badness): Revert accidental checkin.
11173 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11176 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11177 symbols; not inline clones.
11179 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11182 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11183 hard registers. Formatting fixes.
11185 PR preprocessor/83722
11186 * gcc.c (try_generate_repro): Pass
11187 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11188 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11191 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11192 Kito Cheng <kito.cheng@gmail.com>
11194 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11195 (riscv_leaf_function_p): Delete.
11196 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11198 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11200 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11202 (do_ifelse): New function.
11203 (do_isel): New function.
11204 (do_sub3): New function.
11205 (do_add3): New function.
11206 (do_load_mask_compare): New function.
11207 (do_overlap_load_compare): New function.
11208 (expand_compare_loop): New function.
11209 (expand_block_compare): Call expand_compare_loop() when appropriate.
11210 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11211 option description.
11212 (-mblock-compare-inline-loop-limit): New option.
11214 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11217 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11218 Reverse order of second and third operands in first alternative.
11219 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11220 of first and second elements in UNSPEC_VPERMR vector.
11221 (altivec_expand_vec_perm_le): Likewise.
11223 2018-01-08 Jeff Law <law@redhat.com>
11225 PR rtl-optimizatin/81308
11226 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11227 (process_switch): If group_case_labels makes a change, then set
11229 (pass_convert_switch::execute): If a switch is converted, then
11230 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11232 PR rtl-optimization/81308
11233 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11236 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11238 PR target/83663 - Revert r255946
11239 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11240 generation for cases where splatting a value is not useful.
11241 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11242 across a vec_duplicate and a paradoxical subreg forming a vector
11243 mode to a vec_concat.
11245 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11247 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11248 -march=armv8.3-a variants.
11249 * config/arm/t-multilib: Likewise.
11250 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11252 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11254 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11256 (cceq_ior_compare_complement): Give it a name so I can use it, and
11257 change boolean_or_operator predicate to boolean_operator so it can
11258 be used to generate a crand.
11259 (eqne): New code iterator.
11260 (bd/bd_neg): New code_attrs.
11261 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11262 a single define_insn.
11263 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11264 decrement (bdnzt/bdnzf/bdzt/bdzf).
11265 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11266 with the new names of the branch decrement patterns, and added the
11267 names of the branch decrement conditional patterns.
11269 2018-01-08 Richard Biener <rguenther@suse.de>
11271 PR tree-optimization/83563
11272 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11275 2018-01-08 Richard Biener <rguenther@suse.de>
11277 PR middle-end/83713
11278 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11280 2018-01-08 Richard Biener <rguenther@suse.de>
11282 PR tree-optimization/83685
11283 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11284 references to abnormals.
11286 2018-01-08 Richard Biener <rguenther@suse.de>
11289 * dwarf2out.c (output_indirect_strings): Handle empty
11290 skeleton_debug_str_hash.
11291 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11293 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11295 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11296 (emit_store_direct): Likewise.
11297 (arc_trampoline_adjust_address): Likewise.
11298 (arc_asm_trampoline_template): New function.
11299 (arc_initialize_trampoline): Use asm_trampoline_template.
11300 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11301 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11302 * config/arc/arc.md (flush_icache): Delete pattern.
11304 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11306 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11307 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11310 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11313 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11314 by not USED_FOR_TARGET.
11315 (make_pass_resolve_sw_modes): Likewise.
11317 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11319 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11322 2018-01-08 Richard Biener <rguenther@suse.de>
11324 PR middle-end/83580
11325 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11327 2018-01-08 Richard Biener <rguenther@suse.de>
11329 PR middle-end/83517
11330 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11332 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11334 PR middle-end/81897
11335 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11336 basic blocks with a small number of successors.
11337 (convert_control_dep_chain_into_preds): Improve handling of
11339 (dump_predicates): Split apart into...
11340 (dump_pred_chain): ...here...
11341 (dump_pred_info): ...and here.
11342 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11343 (can_chain_union_be_invalidated_p): Improve check for invalidation
11345 (uninit_uses_cannot_happen): Avoid unnecessary if
11346 convert_control_dep_chain_into_preds yielded nothing.
11348 2018-01-06 Martin Sebor <msebor@redhat.com>
11350 PR tree-optimization/83640
11351 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11352 subtracting negative offset from size.
11353 (builtin_access::overlap): Adjust offset bounds of the access to fall
11354 within the size of the object if possible.
11356 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11358 PR rtl-optimization/83699
11359 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11360 extract_bit_field_as_subreg to cases in which the extracted
11361 value is also a vector.
11363 * lra-constraints.c (process_alt_operands): Test for the equivalence
11364 substitutions when detecting a possible reload cycle.
11366 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11369 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11370 by default if flag_selective_schedling{,2}. Formatting fixes.
11372 PR rtl-optimization/83682
11373 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11374 if it has non-VECTOR_MODE element mode.
11375 (vec_duplicate_p): Likewise.
11377 PR middle-end/83694
11378 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11379 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11381 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11384 * config/i386/i386-builtin.def
11385 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11386 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11387 Require also OPTION_MASK_ISA_AVX512F in addition to
11388 OPTION_MASK_ISA_GFNI.
11389 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11390 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11391 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11392 to OPTION_MASK_ISA_GFNI.
11393 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11394 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11395 OPTION_MASK_ISA_AVX512BW.
11396 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11397 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11398 addition to OPTION_MASK_ISA_GFNI.
11399 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11400 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11401 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11402 to OPTION_MASK_ISA_GFNI.
11403 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11404 a requirement for all ISAs rather than any of them with a few
11406 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11408 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11409 bitmasks to be enabled with 3 exceptions, instead of requiring any
11410 enabled ISA with lots of exceptions.
11411 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11412 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11413 Change avx512bw in isa attribute to avx512f.
11414 * config/i386/sgxintrin.h: Add license boilerplate.
11415 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11416 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11417 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11418 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11420 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11421 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11422 temporarily sse2 rather than sse if not enabled already.
11425 * config/i386/sse.md (VI248_VLBW): Rename to ...
11426 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11427 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11428 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11429 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11430 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11431 mode iterator instead of VI248_VLBW.
11433 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11435 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11436 (record_modified): Skip clobbers; add debug output.
11437 (param_change_prob): Use sreal frequencies.
11439 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11441 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11442 punt for user-aligned variables.
11444 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11446 * tree-chrec.c (chrec_contains_symbols): Return true for
11449 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11452 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11453 of (x|y) == x for BICS pattern.
11455 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11457 PR tree-optimization/83605
11458 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11459 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11462 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11464 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11465 * config/epiphany/rtems.h: New file.
11467 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11468 Uros Bizjak <ubizjak@gmail.com>
11471 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11472 QIreg_operand instead of register_operand predicate.
11473 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11474 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11475 comments instead of -fmitigate[-_]rop.
11477 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11480 * cgraphunit.c (symbol_table::compile): Switch to text_section
11481 before calling assembly_start debug hook.
11482 * run-rtl-passes.c (run_rtl_passes): Likewise.
11485 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11487 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11488 range_int_cst_p rather than !symbolic_range_p before calling
11489 extract_range_from_multiplicative_op_1.
11491 2018-01-04 Jeff Law <law@redhat.com>
11493 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11494 redundant test in assertion.
11496 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11498 * doc/rtl.texi: Document machine_mode wrapper classes.
11500 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11502 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11503 using tree_to_uhwi.
11505 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11507 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11508 the VEC_PERM_EXPR fold to fail.
11510 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11513 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11514 to switched_sections.
11516 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11519 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11520 test for d.testing.
11522 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11525 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11526 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11528 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11531 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11532 is BLKmode and bitpos not zero or mode change is needed.
11534 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11537 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11540 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11543 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11544 instead of MULT rtx. Update all corresponding splitters.
11545 (*saddl_se): Ditto.
11546 (*ssub<modesuffix>): Ditto.
11547 (*ssubl_se): Ditto.
11548 (*cmp_sadd_di): Update split patterns.
11549 (*cmp_sadd_si): Ditto.
11550 (*cmp_sadd_sidi): Ditto.
11551 (*cmp_ssub_di): Ditto.
11552 (*cmp_ssub_si): Ditto.
11553 (*cmp_ssub_sidi): Ditto.
11554 * config/alpha/predicates.md (const23_operand): New predicate.
11555 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11556 Look for ASHIFT, not MULT inner operand.
11557 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11559 2018-01-04 Martin Liska <mliska@suse.cz>
11561 PR gcov-profile/83669
11562 * gcov.c (output_intermediate_file): Add version to intermediate
11564 * doc/gcov.texi: Document new field 'version' in intermediate
11565 file format. Fix location of '-k' option of gcov command.
11567 2018-01-04 Martin Liska <mliska@suse.cz>
11570 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11572 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11574 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11576 2018-01-03 Martin Sebor <msebor@redhat.com>
11578 PR tree-optimization/83655
11579 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11580 checking calls with invalid arguments.
11582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11584 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11585 (vectorizable_mask_load_store): Delete.
11586 (vectorizable_call): Return false for masked loads and stores.
11587 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11588 instead of gimple_assign_rhs1.
11589 (vectorizable_load): Handle IFN_MASK_LOAD.
11590 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11594 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11596 (vectorizable_mask_load_store): ...here.
11597 (vectorizable_load): ...and here.
11599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11601 * tree-vect-stmts.c (vect_build_all_ones_mask)
11602 (vect_build_zero_merge_argument): New functions, split out from...
11603 (vectorizable_load): ...here.
11605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11607 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11609 (vectorizable_mask_load_store): ...here.
11610 (vectorizable_store): ...and here.
11612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11614 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11616 (vectorizable_mask_load_store): ...here.
11618 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11620 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11621 (vect_model_store_cost): Take a vec_load_store_type instead of a
11623 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11624 (vect_model_store_cost): Take a vec_load_store_type instead of a
11626 (vectorizable_mask_load_store): Update accordingly.
11627 (vectorizable_store): Likewise.
11628 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11632 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11633 IFN_MASK_LOAD calls here rather than...
11634 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11637 Alan Hayward <alan.hayward@arm.com>
11638 David Sherwood <david.sherwood@arm.com>
11640 * expmed.c (extract_bit_field_1): For vector extracts,
11641 fall back to extract_bit_field_as_subreg if vec_extract
11644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11645 Alan Hayward <alan.hayward@arm.com>
11646 David Sherwood <david.sherwood@arm.com>
11648 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11649 they are variable or constant sized.
11650 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11651 slots for constant-sized data.
11653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11654 Alan Hayward <alan.hayward@arm.com>
11655 David Sherwood <david.sherwood@arm.com>
11657 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11658 handling COND_EXPRs with boolean comparisons, try to find a better
11659 basis for the mask type than the boolean itself.
11661 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11663 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11664 is calculated and how it can be overridden.
11665 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11666 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11668 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11672 Alan Hayward <alan.hayward@arm.com>
11673 David Sherwood <david.sherwood@arm.com>
11675 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11676 Remove the mode argument.
11677 (aarch64_simd_valid_immediate): Remove the mode and inverse
11679 * config/aarch64/iterators.md (bitsize): New iterator.
11680 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11681 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11682 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11683 aarch64_simd_valid_immediate.
11684 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11685 (aarch64_reg_or_bic_imm): Likewise.
11686 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11687 with an insn_type enum and msl with a modifier_type enum.
11688 Replace element_width with a scalar_mode. Change the shift
11689 to unsigned int. Add constructors for scalar_float_mode and
11690 scalar_int_mode elements.
11691 (aarch64_vect_float_const_representable_p): Delete.
11692 (aarch64_can_const_movi_rtx_p)
11693 (aarch64_simd_scalar_immediate_valid_for_move)
11694 (aarch64_simd_make_constant): Update call to
11695 aarch64_simd_valid_immediate.
11696 (aarch64_advsimd_valid_immediate_hs): New function.
11697 (aarch64_advsimd_valid_immediate): Likewise.
11698 (aarch64_simd_valid_immediate): Remove mode and inverse
11699 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11700 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11701 and aarch64_float_const_representable_p on the result.
11702 (aarch64_output_simd_mov_immediate): Remove mode argument.
11703 Update call to aarch64_simd_valid_immediate and use of
11704 simd_immediate_info.
11705 (aarch64_output_scalar_simd_mov_immediate): Update call
11708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11709 Alan Hayward <alan.hayward@arm.com>
11710 David Sherwood <david.sherwood@arm.com>
11712 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11713 (mode_nunits): Likewise CONST_MODE_NUNITS.
11714 * machmode.def (ADJUST_NUNITS): Document.
11715 * genmodes.c (mode_data::need_nunits_adj): New field.
11716 (blank_mode): Update accordingly.
11717 (adj_nunits): New variable.
11718 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11720 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11721 listed in adj_nunits.
11722 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11723 listed in adj_nunits. Don't emit case statements for such modes.
11724 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11725 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11726 nothing if adj_nunits is nonnull.
11727 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11728 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11729 (emit_mode_fbit): Update use of print_maybe_const_decl.
11730 (emit_move_size): Likewise. Treat the array as non-const
11732 (emit_mode_adjustments): Handle adj_nunits.
11734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11736 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11737 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11738 (VECTOR_MODES): Use it.
11739 (make_vector_modes): Take the prefix as an argument.
11741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11742 Alan Hayward <alan.hayward@arm.com>
11743 David Sherwood <david.sherwood@arm.com>
11745 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11746 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11747 for MODE_VECTOR_BOOL.
11748 * machmode.def (VECTOR_BOOL_MODE): Document.
11749 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11750 (make_vector_bool_mode): New function.
11751 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11753 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11754 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11756 * stor-layout.c (int_mode_for_mode): Likewise.
11757 * tree.c (build_vector_type_for_mode): Likewise.
11758 * varasm.c (output_constant_pool_2): Likewise.
11759 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11760 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11761 for MODE_VECTOR_BOOL.
11762 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11763 of mode class checks.
11764 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11765 instead of a list of mode class checks.
11766 (expand_vector_scalar_condition): Likewise.
11767 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11770 Alan Hayward <alan.hayward@arm.com>
11771 David Sherwood <david.sherwood@arm.com>
11773 * machmode.h (mode_size): Change from unsigned short to
11775 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11776 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11777 or if measurement_type is not polynomial.
11778 (fixed_size_mode::includes_p): Check for constant-sized modes.
11779 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11780 return a poly_uint16 rather than an unsigned short.
11781 (emit_mode_size): Change the type of mode_size from unsigned short
11782 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11783 (emit_mode_adjustments): Cope with polynomial vector sizes.
11784 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11786 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11788 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11789 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11790 * caller-save.c (setup_save_areas): Likewise.
11791 (replace_reg_with_saved_mem): Likewise.
11792 * calls.c (emit_library_call_value_1): Likewise.
11793 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11794 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11795 (gen_lowpart_for_combine): Likewise.
11796 * convert.c (convert_to_integer_1): Likewise.
11797 * cse.c (equiv_constant, cse_insn): Likewise.
11798 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11799 (cselib_subst_to_values): Likewise.
11800 * dce.c (word_dce_process_block): Likewise.
11801 * df-problems.c (df_word_lr_mark_ref): Likewise.
11802 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11803 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11804 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11805 (rtl_for_decl_location): Likewise.
11806 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11807 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11808 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11809 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11810 (expand_expr_real_1): Likewise.
11811 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11812 (pad_below): Likewise.
11813 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11814 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11815 * ira.c (get_subreg_tracking_sizes): Likewise.
11816 * ira-build.c (ira_create_allocno_objects): Likewise.
11817 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11818 (ira_sort_regnos_for_alter_reg): Likewise.
11819 * ira-costs.c (record_operand_costs): Likewise.
11820 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11821 (resolve_simple_move): Likewise.
11822 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11823 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11824 (lra_constraints): Likewise.
11825 (CONST_POOL_OK_P): Reject variable-sized modes.
11826 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11827 (add_pseudo_to_slot, lra_spill): Likewise.
11828 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11829 * optabs-query.c (get_best_extraction_insn): Likewise.
11830 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11831 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11832 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11833 * recog.c (offsettable_address_addr_space_p): Likewise.
11834 * regcprop.c (maybe_mode_change): Likewise.
11835 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11836 * regrename.c (build_def_use): Likewise.
11837 * regstat.c (dump_reg_info): Likewise.
11838 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11839 (find_reloads, find_reloads_subreg_address): Likewise.
11840 * reload1.c (eliminate_regs_1): Likewise.
11841 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11842 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11843 (simplify_binary_operation_1, simplify_subreg): Likewise.
11844 * targhooks.c (default_function_arg_padding): Likewise.
11845 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11846 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11847 (verify_gimple_assign_ternary): Likewise.
11848 * tree-inline.c (estimate_move_cost): Likewise.
11849 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11850 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11851 (get_address_cost_ainc): Likewise.
11852 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11853 (vect_supportable_dr_alignment): Likewise.
11854 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11855 (vectorizable_reduction): Likewise.
11856 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11857 (vectorizable_operation, vectorizable_load): Likewise.
11858 * tree.c (build_same_sized_truth_vector_type): Likewise.
11859 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11860 * var-tracking.c (emit_note_insn_var_location): Likewise.
11861 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11862 (ADDR_VEC_ALIGN): Likewise.
11864 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11865 Alan Hayward <alan.hayward@arm.com>
11866 David Sherwood <david.sherwood@arm.com>
11868 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11870 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11871 or if measurement_type is polynomial.
11872 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11873 * combine.c (make_extraction): Likewise.
11874 * dse.c (find_shift_sequence): Likewise.
11875 * dwarf2out.c (mem_loc_descriptor): Likewise.
11876 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11877 (extract_bit_field, extract_low_bits): Likewise.
11878 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11879 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11880 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11881 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11882 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11883 * reload.c (find_reloads): Likewise.
11884 * reload1.c (alter_reg): Likewise.
11885 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11886 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11887 * tree-if-conv.c (predicate_mem_writes): Likewise.
11888 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11889 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11890 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11891 * valtrack.c (dead_debug_insert_temp): Likewise.
11892 * varasm.c (mergeable_constant_section): Likewise.
11893 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11896 Alan Hayward <alan.hayward@arm.com>
11897 David Sherwood <david.sherwood@arm.com>
11899 * expr.c (expand_assignment): Cope with polynomial mode sizes
11900 when assigning to a CONCAT.
11902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11903 Alan Hayward <alan.hayward@arm.com>
11904 David Sherwood <david.sherwood@arm.com>
11906 * machmode.h (mode_precision): Change from unsigned short to
11908 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11910 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11911 or if measurement_type is not polynomial.
11912 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11913 in which the mode is already known to be a scalar_int_mode.
11914 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11915 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11917 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11918 for GET_MODE_PRECISION.
11919 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11920 for GET_MODE_PRECISION.
11921 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11923 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11924 (expand_field_assignment, make_extraction): Likewise.
11925 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11926 (get_last_value): Likewise.
11927 * convert.c (convert_to_integer_1): Likewise.
11928 * cse.c (cse_insn): Likewise.
11929 * expr.c (expand_expr_real_1): Likewise.
11930 * lra-constraints.c (simplify_operand_subreg): Likewise.
11931 * optabs-query.c (can_atomic_load_p): Likewise.
11932 * optabs.c (expand_atomic_load): Likewise.
11933 (expand_atomic_store): Likewise.
11934 * ree.c (combine_reaching_defs): Likewise.
11935 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11936 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11937 * tree.h (type_has_mode_precision_p): Likewise.
11938 * ubsan.c (instrument_si_overflow): Likewise.
11940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11941 Alan Hayward <alan.hayward@arm.com>
11942 David Sherwood <david.sherwood@arm.com>
11944 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11945 polynomial numbers of units.
11946 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11947 (valid_vector_subparts_p): New function.
11948 (build_vector_type): Remove temporary shim and take the number
11949 of units as a poly_uint64 rather than an int.
11950 (build_opaque_vector_type): Take the number of units as a
11951 poly_uint64 rather than an int.
11952 * tree.c (build_vector_from_ctor): Handle polynomial
11953 TYPE_VECTOR_SUBPARTS.
11954 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11955 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11956 (build_vector_from_val): If the number of units is variable,
11957 use build_vec_duplicate_cst for constant operands and
11958 VEC_DUPLICATE_EXPR otherwise.
11959 (make_vector_type): Remove temporary is_constant ().
11960 (build_vector_type, build_opaque_vector_type): Take the number of
11961 units as a poly_uint64 rather than an int.
11962 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11964 * cfgexpand.c (expand_debug_expr): Likewise.
11965 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11966 (store_constructor, expand_expr_real_1): Likewise.
11967 (const_scalar_mask_from_tree): Likewise.
11968 * fold-const-call.c (fold_const_reduction): Likewise.
11969 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11970 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11971 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11972 (fold_relational_const): Likewise.
11973 (native_interpret_vector): Likewise. Change the size from an
11974 int to an unsigned int.
11975 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11976 TYPE_VECTOR_SUBPARTS.
11977 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11978 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11979 duplicating a non-constant operand into a variable-length vector.
11980 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11981 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11982 * ipa-icf.c (sem_variable::equals): Likewise.
11983 * match.pd: Likewise.
11984 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11985 * print-tree.c (print_node): Likewise.
11986 * stor-layout.c (layout_type): Likewise.
11987 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11988 * tree-cfg.c (verify_gimple_comparison): Likewise.
11989 (verify_gimple_assign_binary): Likewise.
11990 (verify_gimple_assign_ternary): Likewise.
11991 (verify_gimple_assign_single): Likewise.
11992 * tree-pretty-print.c (dump_generic_node): Likewise.
11993 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11994 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11995 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11996 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11997 (vect_shift_permute_load_chain): Likewise.
11998 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11999 (expand_vector_condition, optimize_vector_constructor): Likewise.
12000 (lower_vec_perm, get_compute_type): Likewise.
12001 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12002 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12003 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12004 (vect_recog_mask_conversion_pattern): Likewise.
12005 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12006 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12007 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12008 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12009 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12010 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12011 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12012 (supportable_widening_operation): Likewise.
12013 (supportable_narrowing_operation): Likewise.
12014 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12016 * varasm.c (output_constant): Likewise.
12018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12019 Alan Hayward <alan.hayward@arm.com>
12020 David Sherwood <david.sherwood@arm.com>
12022 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12023 so that both the length == 3 and length != 3 cases set up their
12024 own permute vectors. Add comments explaining why we know the
12025 number of elements is constant.
12026 (vect_permute_load_chain): Likewise.
12028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12029 Alan Hayward <alan.hayward@arm.com>
12030 David Sherwood <david.sherwood@arm.com>
12032 * machmode.h (mode_nunits): Change from unsigned char to
12034 (ONLY_FIXED_SIZE_MODES): New macro.
12035 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12036 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12037 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12039 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12040 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12041 or if measurement_type is not polynomial.
12042 * genmodes.c (ZERO_COEFFS): New macro.
12043 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12045 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12046 Use ZERO_COEFFS when emitting initializers.
12047 * data-streamer.h (bp_pack_poly_value): New function.
12048 (bp_unpack_poly_value): Likewise.
12049 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12050 for GET_MODE_NUNITS.
12051 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12052 for GET_MODE_NUNITS.
12053 * tree.c (make_vector_type): Remove temporary shim and make
12054 the real function take the number of units as a poly_uint64
12055 rather than an int.
12056 (build_vector_type_for_mode): Handle polynomial nunits.
12057 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12058 * emit-rtl.c (const_vec_series_p_1): Likewise.
12059 (gen_rtx_CONST_VECTOR): Likewise.
12060 * fold-const.c (test_vec_duplicate_folding): Likewise.
12061 * genrecog.c (validate_pattern): Likewise.
12062 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12063 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12064 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12065 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12066 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12067 * rtlanal.c (subreg_get_info): Likewise.
12068 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12069 (vect_grouped_load_supported): Likewise.
12070 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12071 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12072 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12073 (simplify_const_unary_operation, simplify_binary_operation_1)
12074 (simplify_const_binary_operation, simplify_ternary_operation)
12075 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12076 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12077 instead of CONST_VECTOR_NUNITS.
12078 * varasm.c (output_constant_pool_2): Likewise.
12079 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12080 explicit-encoded elements in the XVEC for variable-length vectors.
12082 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12084 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
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 * coretypes.h (fixed_size_mode): Declare.
12091 (fixed_size_mode_pod): New typedef.
12092 * builtins.h (target_builtins::x_apply_args_mode)
12093 (target_builtins::x_apply_result_mode): Change type to
12094 fixed_size_mode_pod.
12095 * builtins.c (apply_args_size, apply_result_size, result_vector)
12096 (expand_builtin_apply_args_1, expand_builtin_apply)
12097 (expand_builtin_return): Update accordingly.
12099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12101 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12102 * cselib.c (cselib_hash_rtx): Likewise.
12103 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12104 CONST_VECTOR encoding.
12106 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12107 Jeff Law <law@redhat.com>
12110 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12111 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12112 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12113 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12116 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12117 explicitly probe *sp in a noreturn function if there were any callee
12118 register saves or frame pointer is needed.
12120 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12123 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12124 BLKmode for ternary, binary or unary expressions.
12127 * var-tracking.c (delete_vta_debug_insn): New inline function.
12128 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12129 insns from get_insns () to NULL instead of each bb separately.
12130 Use delete_vta_debug_insn. No longer static.
12131 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12132 delete_vta_debug_insns callers.
12133 * rtl.h (delete_vta_debug_insns): Declare.
12134 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12135 instead of variable_tracking_main.
12137 2018-01-03 Martin Sebor <msebor@redhat.com>
12139 PR tree-optimization/83603
12140 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12141 arguments past the endof the argument list in functions declared
12142 without a prototype.
12143 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12144 Avoid checking when arguments are null.
12146 2018-01-03 Martin Sebor <msebor@redhat.com>
12149 * doc/extend.texi (attribute const): Fix a typo.
12150 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12151 issuing -Wsuggest-attribute for void functions.
12153 2018-01-03 Martin Sebor <msebor@redhat.com>
12155 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12156 offset_int::from instead of wide_int::to_shwi.
12157 (maybe_diag_overlap): Remove assertion.
12158 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12159 * gimple-ssa-sprintf.c (format_directive): Same.
12160 (parse_directive): Same.
12161 (sprintf_dom_walker::compute_format_length): Same.
12162 (try_substitute_return_value): Same.
12164 2018-01-03 Jeff Law <law@redhat.com>
12166 PR middle-end/83654
12167 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12168 non-constant residual for zero at runtime and avoid probing in
12169 that case. Reorganize code for trailing problem to mirror handling
12172 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12174 PR tree-optimization/83501
12175 * tree-ssa-strlen.c (get_string_cst): New.
12176 (handle_char_store): Call get_string_cst.
12178 2018-01-03 Martin Liska <mliska@suse.cz>
12180 PR tree-optimization/83593
12181 * tree-ssa-strlen.c: Include tree-cfg.h.
12182 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12183 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12184 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12186 (strlen_dom_walker::before_dom_children): Call
12187 gimple_purge_dead_eh_edges. Dump tranformation with details
12189 (strlen_dom_walker::before_dom_children): Update call by adding
12190 new argument cleanup_eh.
12191 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12193 2018-01-03 Martin Liska <mliska@suse.cz>
12196 * cif-code.def (VARIADIC_THUNK): New enum value.
12197 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12200 2018-01-03 Jan Beulich <jbeulich@suse.com>
12202 * sse.md (mov<mode>_internal): Tighten condition for when to use
12203 vmovdqu<ssescalarsize> for TI and OI modes.
12205 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12207 Update copyright years.
12209 2018-01-03 Martin Liska <mliska@suse.cz>
12212 * ipa-visibility.c (function_and_variable_visibility): Skip
12213 functions with noipa attribure.
12215 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12217 * gcc.c (process_command): Update copyright notice dates.
12218 * gcov-dump.c (print_version): Ditto.
12219 * gcov.c (print_version): Ditto.
12220 * gcov-tool.c (print_version): Ditto.
12221 * gengtype.c (create_file): Ditto.
12222 * doc/cpp.texi: Bump @copying's copyright year.
12223 * doc/cppinternals.texi: Ditto.
12224 * doc/gcc.texi: Ditto.
12225 * doc/gccint.texi: Ditto.
12226 * doc/gcov.texi: Ditto.
12227 * doc/install.texi: Ditto.
12228 * doc/invoke.texi: Ditto.
12230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12232 * vector-builder.h (vector_builder::m_full_nelts): Change from
12233 unsigned int to poly_uint64.
12234 (vector_builder::full_nelts): Update prototype accordingly.
12235 (vector_builder::new_vector): Likewise.
12236 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12237 (vector_builder::operator ==): Likewise.
12238 (vector_builder::finalize): Likewise.
12239 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12240 Take the number of elements as a poly_uint64 rather than an
12242 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12243 from unsigned int to poly_uint64.
12244 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12245 (vec_perm_indices::new_vector): Likewise.
12246 (vec_perm_indices::length): Likewise.
12247 (vec_perm_indices::nelts_per_input): Likewise.
12248 (vec_perm_indices::input_nelts): Likewise.
12249 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12250 number of elements per input as a poly_uint64 rather than an
12251 unsigned int. Use the original encoding for variable-length
12252 vectors, rather than clamping each individual element.
12253 For the second and subsequent elements in each pattern,
12254 clamp the step and base before clamping their sum.
12255 (vec_perm_indices::series_p): Handle polynomial element counts.
12256 (vec_perm_indices::all_in_range_p): Likewise.
12257 (vec_perm_indices_to_tree): Likewise.
12258 (vec_perm_indices_to_rtx): Likewise.
12259 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12260 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12261 (tree_vector_builder::new_binary_operation): Handle polynomial
12262 element counts. Return false if we need to know the number
12263 of elements at compile time.
12264 * fold-const.c (fold_vec_perm): Punt if the number of elements
12265 isn't known at compile time.
12267 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12269 * vec-perm-indices.h (vec_perm_builder): Change element type
12270 from HOST_WIDE_INT to poly_int64.
12271 (vec_perm_indices::element_type): Update accordingly.
12272 (vec_perm_indices::clamp): Handle polynomial element_types.
12273 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12274 (vec_perm_indices::all_in_range_p): Likewise.
12275 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12277 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12278 polynomial vec_perm_indices element types.
12279 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12280 * fold-const.c (fold_vec_perm): Likewise.
12281 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12282 * tree-vect-generic.c (lower_vec_perm): Likewise.
12283 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12284 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12285 element type to HOST_WIDE_INT.
12287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12288 Alan Hayward <alan.hayward@arm.com>
12289 David Sherwood <david.sherwood@arm.com>
12291 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12292 rather than an int. Use plus_constant.
12293 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12294 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12297 Alan Hayward <alan.hayward@arm.com>
12298 David Sherwood <david.sherwood@arm.com>
12300 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12301 a HOST_WIDE_INT to a poly_int64.
12303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12304 Alan Hayward <alan.hayward@arm.com>
12305 David Sherwood <david.sherwood@arm.com>
12307 * calls.c (load_register_parameters): Cope with polynomial
12308 mode sizes. Require a constant size for BLKmode parameters
12309 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12310 forces a parameter to be padded at the lsb end in order to
12311 fill a complete number of words, require the parameter size
12312 to be ordered wrt UNITS_PER_WORD.
12314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12315 Alan Hayward <alan.hayward@arm.com>
12316 David Sherwood <david.sherwood@arm.com>
12318 * reload1.c (spill_stack_slot_width): Change element type
12319 from unsigned int to poly_uint64_pod.
12320 (alter_reg): Treat mode sizes as polynomial.
12322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12323 Alan Hayward <alan.hayward@arm.com>
12324 David Sherwood <david.sherwood@arm.com>
12326 * reload.c (complex_word_subreg_p): New function.
12327 (reload_inner_reg_of_subreg, push_reload): Use it.
12329 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12330 Alan Hayward <alan.hayward@arm.com>
12331 David Sherwood <david.sherwood@arm.com>
12333 * lra-constraints.c (process_alt_operands): Reject matched
12334 operands whose sizes aren't ordered.
12335 (match_reload): Refer to this check here.
12337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12338 Alan Hayward <alan.hayward@arm.com>
12339 David Sherwood <david.sherwood@arm.com>
12341 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12342 that the mode size is in the set {1, 2, 4, 8, 16}.
12344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12345 Alan Hayward <alan.hayward@arm.com>
12346 David Sherwood <david.sherwood@arm.com>
12348 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12349 Use plus_constant instead of gen_rtx_PLUS.
12351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12352 Alan Hayward <alan.hayward@arm.com>
12353 David Sherwood <david.sherwood@arm.com>
12355 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12356 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12357 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12358 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12359 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12360 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12361 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12362 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12363 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12364 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12366 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12367 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12368 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12369 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12370 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12371 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12372 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12373 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12374 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12375 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12377 * expr.c (emit_move_resolve_push): Treat the input and result
12378 of PUSH_ROUNDING as a poly_int64.
12379 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12380 (emit_push_insn): Likewise.
12381 * lra-eliminations.c (mark_not_eliminable): Likewise.
12382 * recog.c (push_operand): Likewise.
12383 * reload1.c (elimination_effects): Likewise.
12384 * rtlanal.c (nonzero_bits1): Likewise.
12385 * calls.c (store_one_arg): Likewise. Require the padding to be
12386 known at compile time.
12388 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12389 Alan Hayward <alan.hayward@arm.com>
12390 David Sherwood <david.sherwood@arm.com>
12392 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12393 Use plus_constant instead of gen_rtx_PLUS.
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 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12400 rather than an int.
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 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12407 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12408 via stack temporaries. Treat the mode size as polynomial too.
12410 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12411 Alan Hayward <alan.hayward@arm.com>
12412 David Sherwood <david.sherwood@arm.com>
12414 * expr.c (expand_expr_real_2): When handling conversions involving
12415 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12416 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12417 as a poly_uint64 too.
12419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12420 Alan Hayward <alan.hayward@arm.com>
12421 David Sherwood <david.sherwood@arm.com>
12423 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12425 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12426 Alan Hayward <alan.hayward@arm.com>
12427 David Sherwood <david.sherwood@arm.com>
12429 * combine.c (can_change_dest_mode): Handle polynomial
12430 REGMODE_NATURAL_SIZE.
12431 * expmed.c (store_bit_field_1): Likewise.
12432 * expr.c (store_constructor): Likewise.
12433 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12434 and polynomial REGMODE_NATURAL_SIZE.
12435 (gen_lowpart_common): Likewise.
12436 * reginfo.c (record_subregs_of_mode): Likewise.
12437 * rtlanal.c (read_modify_subreg_p): Likewise.
12439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12440 Alan Hayward <alan.hayward@arm.com>
12441 David Sherwood <david.sherwood@arm.com>
12443 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12444 numbers of elements.
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 * match.pd: Cope with polynomial numbers of vector elements.
12452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12453 Alan Hayward <alan.hayward@arm.com>
12454 David Sherwood <david.sherwood@arm.com>
12456 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12457 in a POINTER_PLUS_EXPR.
12459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12460 Alan Hayward <alan.hayward@arm.com>
12461 David Sherwood <david.sherwood@arm.com>
12463 * omp-simd-clone.c (simd_clone_subparts): New function.
12464 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12465 (ipa_simd_modify_function_body): Likewise.
12467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12468 Alan Hayward <alan.hayward@arm.com>
12469 David Sherwood <david.sherwood@arm.com>
12471 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12472 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12473 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12474 (expand_vector_condition, vector_element): Likewise.
12475 (subparts_gt): New function.
12476 (get_compute_type): Use subparts_gt.
12477 (count_type_subparts): Delete.
12478 (expand_vector_operations_1): Use subparts_gt instead of
12479 count_type_subparts.
12481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12482 Alan Hayward <alan.hayward@arm.com>
12483 David Sherwood <david.sherwood@arm.com>
12485 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12486 (vect_compile_time_alias): ...this new function. Do the calculation
12487 on poly_ints rather than trees.
12488 (vect_prune_runtime_alias_test_list): Update call accordingly.
12490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12491 Alan Hayward <alan.hayward@arm.com>
12492 David Sherwood <david.sherwood@arm.com>
12494 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12496 (vect_schedule_slp_instance): Likewise.
12498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12499 Alan Hayward <alan.hayward@arm.com>
12500 David Sherwood <david.sherwood@arm.com>
12502 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12503 constant and extern definitions for variable-length vectors.
12504 (vect_get_constant_vectors): Note that the number of units
12505 is known to be constant.
12507 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12508 Alan Hayward <alan.hayward@arm.com>
12509 David Sherwood <david.sherwood@arm.com>
12511 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12512 of units as polynomial. Choose between WIDE and NARROW based
12515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12516 Alan Hayward <alan.hayward@arm.com>
12517 David Sherwood <david.sherwood@arm.com>
12519 * tree-vect-stmts.c (simd_clone_subparts): New function.
12520 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12523 Alan Hayward <alan.hayward@arm.com>
12524 David Sherwood <david.sherwood@arm.com>
12526 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12527 vectors as polynomial. Use build_index_vector for
12528 IFN_GOMP_SIMD_LANE.
12530 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12531 Alan Hayward <alan.hayward@arm.com>
12532 David Sherwood <david.sherwood@arm.com>
12534 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12535 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12536 for variable-length vectors.
12537 (vectorizable_mask_load_store): Treat the number of units as
12538 polynomial, asserting that it is constant if the condition has
12539 already been enforced.
12540 (vectorizable_store, vectorizable_load): Likewise.
12542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12543 Alan Hayward <alan.hayward@arm.com>
12544 David Sherwood <david.sherwood@arm.com>
12546 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12547 of units as polynomial. Punt if we can't tell at compile time
12548 which vector contains the final result.
12550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12551 Alan Hayward <alan.hayward@arm.com>
12552 David Sherwood <david.sherwood@arm.com>
12554 * tree-vect-loop.c (vectorizable_induction): Treat the number
12555 of units as polynomial. Punt on SLP inductions. Use an integer
12556 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12557 cast of such a series for variable-length floating-point
12560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12561 Alan Hayward <alan.hayward@arm.com>
12562 David Sherwood <david.sherwood@arm.com>
12564 * tree.h (build_index_vector): Declare.
12565 * tree.c (build_index_vector): New function.
12566 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12567 of units as polynomial, forcibly converting it to a constant if
12568 vectorizable_reduction has already enforced the condition.
12569 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12570 to create a {1,2,3,...} vector.
12571 (vectorizable_reduction): Treat the number of units as polynomial.
12572 Choose vectype_in based on the largest scalar element size rather
12573 than the smallest number of units. Enforce the restrictions
12576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12577 Alan Hayward <alan.hayward@arm.com>
12578 David Sherwood <david.sherwood@arm.com>
12580 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12581 number of units as polynomial.
12583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12584 Alan Hayward <alan.hayward@arm.com>
12585 David Sherwood <david.sherwood@arm.com>
12587 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12588 * target.def (autovectorize_vector_sizes): Return the vector sizes
12589 by pointer, using vector_sizes rather than a bitmask.
12590 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12591 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12592 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12594 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12595 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12596 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12597 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12598 * omp-general.c (omp_max_vf): Likewise.
12599 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12600 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12601 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12602 * tree-vect-slp.c (vect_slp_bb): Likewise.
12603 * doc/tm.texi: Regenerate.
12604 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12606 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12607 the vector size as a poly_uint64 rather than an unsigned int.
12608 (current_vector_size): Change from an unsigned int to a poly_uint64.
12609 (get_vectype_for_scalar_type): Update accordingly.
12610 * tree.h (build_truth_vector_type): Take the size and number of
12611 units as a poly_uint64 rather than an unsigned int.
12612 (build_vector_type): Add a temporary overload that takes
12613 the number of units as a poly_uint64 rather than an unsigned int.
12614 * tree.c (make_vector_type): Likewise.
12615 (build_truth_vector_type): Take the number of units as a poly_uint64
12616 rather than an unsigned int.
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 * target.def (get_mask_mode): Take the number of units and length
12623 as poly_uint64s rather than unsigned ints.
12624 * targhooks.h (default_get_mask_mode): Update accordingly.
12625 * targhooks.c (default_get_mask_mode): Likewise.
12626 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12627 * doc/tm.texi: Regenerate.
12629 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12630 Alan Hayward <alan.hayward@arm.com>
12631 David Sherwood <david.sherwood@arm.com>
12633 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12634 * omp-general.c (omp_max_vf): Likewise.
12635 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12636 (expand_omp_simd): Handle polynomial safelen.
12637 * omp-low.c (omplow_simd_context): Add a default constructor.
12638 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12639 (lower_rec_simd_input_clauses): Update accordingly.
12640 (lower_rec_input_clauses): Likewise.
12642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12643 Alan Hayward <alan.hayward@arm.com>
12644 David Sherwood <david.sherwood@arm.com>
12646 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12647 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12648 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12649 (vect_analyze_slp_cost): Likewise.
12650 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12651 (vect_model_load_cost): Likewise.
12653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12654 Alan Hayward <alan.hayward@arm.com>
12655 David Sherwood <david.sherwood@arm.com>
12657 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12658 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12659 from an unsigned int * to a poly_uint64_pod *.
12660 (calculate_unrolling_factor): New function.
12661 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12664 Alan Hayward <alan.hayward@arm.com>
12665 David Sherwood <david.sherwood@arm.com>
12667 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12668 from an unsigned int to a poly_uint64.
12669 (_loop_vec_info::slp_unrolling_factor): Likewise.
12670 (_loop_vec_info::vectorization_factor): Change from an int
12672 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12673 (vect_get_num_vectors): New function.
12674 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12675 (vect_get_num_copies): Use vect_get_num_vectors.
12676 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12677 to an unsigned int *.
12678 (vect_analyze_data_refs): Change min_vf from an int * to a
12680 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12681 than an unsigned HOST_WIDE_INT.
12682 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12683 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12684 to an unsigned int *.
12685 (vect_analyze_data_ref_dependences): Likewise.
12686 (vect_compute_data_ref_alignment): Handle polynomial vf.
12687 (vect_enhance_data_refs_alignment): Likewise.
12688 (vect_prune_runtime_alias_test_list): Likewise.
12689 (vect_shift_permute_load_chain): Likewise.
12690 (vect_supportable_dr_alignment): Likewise.
12691 (dependence_distance_ge_vf): Take the vectorization factor as a
12692 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12693 (vect_analyze_data_refs): Change min_vf from an int * to a
12695 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12696 vfm1 as a poly_uint64 rather than an int. Make the same change
12697 for the returned bound_scalar.
12698 (vect_gen_vector_loop_niters): Handle polynomial vf.
12699 (vect_do_peeling): Likewise. Update call to
12700 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12701 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12703 * tree-vect-loop.c (vect_determine_vectorization_factor)
12704 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12705 (vect_get_known_peeling_cost): Likewise.
12706 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12707 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12708 (vect_transform_loop): Likewise. Use the lowest possible VF when
12709 updating the upper bounds of the loop.
12710 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12711 rather than an int.
12712 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12713 polynomial unroll factors.
12714 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12715 (vect_make_slp_decision): Likewise.
12716 (vect_supported_load_permutation_p): Likewise, and polynomial
12718 (vect_analyze_slp_cost): Handle polynomial vf.
12719 (vect_slp_analyze_node_operations): Likewise.
12720 (vect_slp_analyze_bb_1): Likewise.
12721 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12722 than an unsigned HOST_WIDE_INT.
12723 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12724 (vectorizable_load): Handle polynomial vf.
12725 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12727 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12730 Alan Hayward <alan.hayward@arm.com>
12731 David Sherwood <david.sherwood@arm.com>
12733 * match.pd: Handle bit operations involving three constants
12734 and try to fold one pair.
12736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12738 * tree-vect-loop-manip.c: Include gimple-fold.h.
12739 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12740 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12741 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12742 Add a path that uses a step of VF instead of 1, but disable it
12744 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12745 and niters_no_overflow parameters. Update calls to
12746 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12747 Create a new SSA name if the latter choses to use a ste other
12748 than zero, and return it via niters_vector_mult_vf_var.
12749 * tree-vect-loop.c (vect_transform_loop): Update calls to
12750 vect_do_peeling, vect_gen_vector_loop_niters and
12751 slpeel_make_loop_iterate_ntimes.
12752 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12753 (vect_gen_vector_loop_niters): Update declarations after above changes.
12755 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12757 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12758 128-bit round to integer instructions.
12759 (ceil<mode>2): Likewise.
12760 (btrunc<mode>2): Likewise.
12761 (round<mode>2): Likewise.
12763 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12765 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12766 unaligned VSX load/store on P8/P9.
12767 (expand_block_clear): Allow the use of unaligned VSX
12768 load/store on P8/P9.
12770 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12772 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12774 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12775 swap associated with both a load and a store.
12777 2018-01-02 Andrew Waterman <andrew@sifive.com>
12779 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12780 * config/riscv/riscv.md (clear_cache): Use it.
12782 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12784 * web.c: Remove out-of-date comment.
12786 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12788 * expr.c (fixup_args_size_notes): Check that any existing
12789 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12790 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12791 (emit_single_push_insn): ...here.
12793 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12795 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12796 (const_vector_encoded_nelts): New function.
12797 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12798 (const_vector_int_elt, const_vector_elt): Declare.
12799 * emit-rtl.c (const_vector_int_elt_1): New function.
12800 (const_vector_elt): Likewise.
12801 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12802 of CONST_VECTOR_ELT.
12804 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12806 * expr.c: Include rtx-vector-builder.h.
12807 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12808 directly on the tree encoding.
12809 (const_vector_from_tree): Likewise.
12810 * optabs.c: Include rtx-vector-builder.h.
12811 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12812 sequence of "u" values.
12813 * vec-perm-indices.c: Include rtx-vector-builder.h.
12814 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12815 directly on the vec_perm_indices encoding.
12817 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12819 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12820 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12821 * rtx-vector-builder.h: New file.
12822 * rtx-vector-builder.c: Likewise.
12823 * rtl.h (rtx_def::u2): Add a const_vector field.
12824 (CONST_VECTOR_NPATTERNS): New macro.
12825 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12826 (CONST_VECTOR_DUPLICATE_P): Likewise.
12827 (CONST_VECTOR_STEPPED_P): Likewise.
12828 (CONST_VECTOR_ENCODED_ELT): Likewise.
12829 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12830 (unwrap_const_vec_duplicate): Likewise.
12831 (const_vec_series_p): Check for a non-duplicated vector encoding.
12832 Say that the function only returns true for integer vectors.
12833 * emit-rtl.c: Include rtx-vector-builder.h.
12834 (gen_const_vec_duplicate_1): Delete.
12835 (gen_const_vector): Call gen_const_vec_duplicate instead of
12836 gen_const_vec_duplicate_1.
12837 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12838 (gen_const_vec_duplicate): Use rtx_vector_builder.
12839 (gen_const_vec_series): Likewise.
12840 (gen_rtx_CONST_VECTOR): Likewise.
12841 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12842 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12843 Build a new vector rather than modifying a CONST_VECTOR in-place.
12844 (handle_special_swappables): Update call accordingly.
12845 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12846 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12847 Build a new vector rather than modifying a CONST_VECTOR in-place.
12848 (handle_special_swappables): Update call accordingly.
12850 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12852 * simplify-rtx.c (simplify_const_binary_operation): Use
12853 CONST_VECTOR_ELT instead of XVECEXP.
12855 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12857 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12858 the selector elements to be different from the data elements
12859 if the selector is a VECTOR_CST.
12860 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12861 ssizetype for the selector.
12863 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12865 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12866 before testing each element individually.
12867 * tree-vect-generic.c (lower_vec_perm): Likewise.
12869 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12871 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12872 * selftest-run-tests.c (selftest::run_tests): Call it.
12873 * vector-builder.h (vector_builder::operator ==): New function.
12874 (vector_builder::operator !=): Likewise.
12875 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12876 (vec_perm_indices::all_from_input_p): New function.
12877 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12878 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12879 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12880 instead of reading the VECTOR_CST directly. Detect whether both
12881 vector inputs are the same before constructing the vec_perm_indices,
12882 and update the number of inputs argument accordingly. Use the
12883 utility functions added above. Only construct sel2 if we need to.
12885 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12887 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12888 the broadcast of the low byte.
12889 (expand_mult_highpart): Use an explicit encoding for the permutes.
12890 * optabs-query.c (can_mult_highpart_p): Likewise.
12891 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12892 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12893 (vectorizable_bswap): Likewise.
12894 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12895 explicit encoding for the power-of-2 permutes.
12896 (vect_permute_store_chain): Likewise.
12897 (vect_grouped_load_supported): Likewise.
12898 (vect_permute_load_chain): Likewise.
12900 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12902 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12903 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12904 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12905 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12906 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12907 (vect_gen_perm_mask_any): Likewise.
12909 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12911 * int-vector-builder.h: New file.
12912 * vec-perm-indices.h: Include int-vector-builder.h.
12913 (vec_perm_indices): Redefine as an int_vector_builder.
12914 (auto_vec_perm_indices): Delete.
12915 (vec_perm_builder): Redefine as a stand-alone class.
12916 (vec_perm_indices::vec_perm_indices): New function.
12917 (vec_perm_indices::clamp): Likewise.
12918 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12919 (vec_perm_indices::new_vector): New function.
12920 (vec_perm_indices::new_expanded_vector): Update for new
12921 vec_perm_indices class.
12922 (vec_perm_indices::rotate_inputs): New function.
12923 (vec_perm_indices::all_in_range_p): Operate directly on the
12924 encoded form, without computing elided elements.
12925 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12926 encoding. Update for new vec_perm_indices class.
12927 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12928 the given vec_perm_builder.
12929 (expand_vec_perm_var): Update vec_perm_builder constructor.
12930 (expand_mult_highpart): Use vec_perm_builder instead of
12931 auto_vec_perm_indices.
12932 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12933 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12934 or double series encoding as appropriate.
12935 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12936 vec_perm_indices instead of auto_vec_perm_indices.
12937 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12938 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12939 (vect_permute_store_chain): Likewise.
12940 (vect_grouped_load_supported): Likewise.
12941 (vect_permute_load_chain): Likewise.
12942 (vect_shift_permute_load_chain): Likewise.
12943 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12944 (vect_transform_slp_perm_load): Likewise.
12945 (vect_schedule_slp_instance): Likewise.
12946 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12947 (vectorizable_mask_load_store): Likewise.
12948 (vectorizable_bswap): Likewise.
12949 (vectorizable_store): Likewise.
12950 (vectorizable_load): Likewise.
12951 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12952 vec_perm_indices instead of auto_vec_perm_indices. Use
12953 tree_to_vec_perm_builder to read the vector from a tree.
12954 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12955 vec_perm_builder instead of a vec_perm_indices.
12956 (have_whole_vector_shift): Use vec_perm_builder and
12957 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12958 truncation to calc_vec_perm_mask_for_shift.
12959 (vect_create_epilog_for_reduction): Likewise.
12960 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12961 from auto_vec_perm_indices to vec_perm_indices.
12962 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12963 instead of changing individual elements.
12964 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12965 the vector in d.perm.
12966 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12967 from auto_vec_perm_indices to vec_perm_indices.
12968 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12969 instead of changing individual elements.
12970 (arm_vectorize_vec_perm_const): Use new_vector to install
12971 the vector in d.perm.
12972 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12973 Update vec_perm_builder constructor.
12974 (rs6000_expand_interleave): Likewise.
12975 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12976 (rs6000_expand_interleave): Likewise.
12978 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12980 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12981 to qimode could truncate the indices.
12982 * optabs.c (expand_vec_perm_var): Likewise.
12984 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12986 * Makefile.in (OBJS): Add vec-perm-indices.o.
12987 * vec-perm-indices.h: New file.
12988 * vec-perm-indices.c: Likewise.
12989 * target.h (vec_perm_indices): Replace with a forward class
12991 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12992 * optabs.h: Include vec-perm-indices.h.
12993 (expand_vec_perm): Delete.
12994 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12995 (expand_vec_perm_const): Declare.
12996 * target.def (vec_perm_const_ok): Replace with...
12997 (vec_perm_const): ...this new hook.
12998 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12999 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13000 * doc/tm.texi: Regenerate.
13001 * optabs.def (vec_perm_const): Delete.
13002 * doc/md.texi (vec_perm_const): Likewise.
13003 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13004 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13005 expand_vec_perm for constant permutation vectors. Assert that
13006 the mode of variable permutation vectors is the integer equivalent
13007 of the mode that is being permuted.
13008 * optabs-query.h (selector_fits_mode_p): Declare.
13009 * optabs-query.c: Include vec-perm-indices.h.
13010 (selector_fits_mode_p): New function.
13011 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13012 is defined, instead of checking whether the vec_perm_const_optab
13013 exists. Use targetm.vectorize.vec_perm_const instead of
13014 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13015 fit in the vector mode before using a variable permute.
13016 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13017 vec_perm_indices instead of an rtx.
13018 (expand_vec_perm): Replace with...
13019 (expand_vec_perm_const): ...this new function. Take the selector
13020 as a vec_perm_indices rather than an rtx. Also take the mode of
13021 the selector. Update call to shift_amt_for_vec_perm_mask.
13022 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13023 Use vec_perm_indices::new_expanded_vector to expand the original
13024 selector into bytes. Check whether the indices fit in the vector
13025 mode before using a variable permute.
13026 (expand_vec_perm_var): Make global.
13027 (expand_mult_highpart): Use expand_vec_perm_const.
13028 * fold-const.c: Includes vec-perm-indices.h.
13029 * tree-ssa-forwprop.c: Likewise.
13030 * tree-vect-data-refs.c: Likewise.
13031 * tree-vect-generic.c: Likewise.
13032 * tree-vect-loop.c: Likewise.
13033 * tree-vect-slp.c: Likewise.
13034 * tree-vect-stmts.c: Likewise.
13035 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13037 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13038 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13039 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13040 (aarch64_vectorize_vec_perm_const): ...this new function.
13041 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13042 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13043 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13044 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13045 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13046 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13047 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13049 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13050 check for NEON modes.
13051 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13052 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13053 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13054 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13056 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13057 the old VEC_PERM_CONST conditions.
13058 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13059 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13060 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13061 (ia64_vectorize_vec_perm_const_ok): Merge into...
13062 (ia64_vectorize_vec_perm_const): ...this new function.
13063 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13064 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13065 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13066 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13067 * config/mips/mips.c (mips_expand_vec_perm_const)
13068 (mips_vectorize_vec_perm_const_ok): Merge into...
13069 (mips_vectorize_vec_perm_const): ...this new function.
13070 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13071 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13072 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13073 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13074 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13075 (rs6000_expand_vec_perm_const): Delete.
13076 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13078 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13079 (altivec_expand_vec_perm_const_le): Take each operand individually.
13080 Operate on constant selectors rather than rtxes.
13081 (altivec_expand_vec_perm_const): Likewise. Update call to
13082 altivec_expand_vec_perm_const_le.
13083 (rs6000_expand_vec_perm_const): Delete.
13084 (rs6000_vectorize_vec_perm_const_ok): Delete.
13085 (rs6000_vectorize_vec_perm_const): New function.
13086 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13087 an element count and rtx array.
13088 (rs6000_expand_extract_even): Update call accordingly.
13089 (rs6000_expand_interleave): Likewise.
13090 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13091 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13092 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13093 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13094 (rs6000_expand_vec_perm_const): Delete.
13095 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13096 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13097 (altivec_expand_vec_perm_const_le): Take each operand individually.
13098 Operate on constant selectors rather than rtxes.
13099 (altivec_expand_vec_perm_const): Likewise. Update call to
13100 altivec_expand_vec_perm_const_le.
13101 (rs6000_expand_vec_perm_const): Delete.
13102 (rs6000_vectorize_vec_perm_const_ok): Delete.
13103 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13104 reference to the SPE evmerge intructions.
13105 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13106 an element count and rtx array.
13107 (rs6000_expand_extract_even): Update call accordingly.
13108 (rs6000_expand_interleave): Likewise.
13109 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13110 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13112 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13114 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13116 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13117 vector mode and that that mode matches the mode of the data
13119 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13120 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13121 directly using expand_vec_perm_1 when forcing selectors into
13123 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13125 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13127 * optabs-query.h (can_vec_perm_p): Delete.
13128 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13129 * optabs-query.c (can_vec_perm_p): Split into...
13130 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13131 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13132 particular selector is valid.
13133 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13134 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13135 (vect_grouped_load_supported): Likewise.
13136 (vect_shift_permute_load_chain): Likewise.
13137 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13138 (vect_transform_slp_perm_load): Likewise.
13139 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13140 (vectorizable_bswap): Likewise.
13141 (vect_gen_perm_mask_checked): Likewise.
13142 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13143 implementations of variable permutation vectors into account
13144 when deciding which selector to use.
13145 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13146 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13147 with a false third argument.
13148 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13149 to test whether the constant selector is valid and can_vec_perm_var_p
13150 to test whether a variable selector is valid.
13152 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13154 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13155 * optabs-query.c (can_vec_perm_p): Likewise.
13156 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13157 instead of vec_perm_indices.
13158 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13159 (vect_gen_perm_mask_checked): Likewise,
13160 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13161 (vect_gen_perm_mask_checked): Likewise,
13163 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13165 * optabs-query.h (qimode_for_vec_perm): Declare.
13166 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13167 (qimode_for_vec_perm): ...this new function.
13168 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13170 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13172 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13173 does not have a conditional at the top.
13175 2018-01-02 Richard Biener <rguenther@suse.de>
13177 * ipa-inline.c (big_speedup_p): Fix expression.
13179 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13182 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13185 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13189 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13190 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13191 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13192 cond_taken_branch_cost 3->4.
13194 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13196 PR tree-optimization/83581
13197 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13198 TODO_cleanup_cfg if any changes have been made.
13200 PR middle-end/83608
13201 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13202 convert_modes if target mode has the right side, but different mode
13205 PR middle-end/83609
13206 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13207 last argument when extracting from CONCAT. If either from_real or
13208 from_imag is NULL, use expansion through memory. If result is not
13209 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13210 the parts directly to inner mode, if even that fails, use expansion
13213 PR middle-end/83623
13214 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13215 check for bswap in mode rather than HImode and use that in expand_unop
13218 Copyright (C) 2018 Free Software Foundation, Inc.
13220 Copying and distribution of this file, with or without modification,
13221 are permitted in any medium without royalty provided the copyright
13222 notice and this notice are preserved.